bulletin - wordpress.comspeakers jianqing fan at princeton university, zhil-iang ying at columbia...
TRANSCRIPT
Bulletin
ISSN: 2226-2393
www.icsa.orgJuly, 2019
Volume 31, Issue 2
AReminiscence of Dr. Margaret C.Wu’s Sta-tistical Innovationat theNationalHeart, Lungand Blood InstituteJoseph F. HeyseYi’s FDA Story: Where Statistics Met Regu-lation: 1988 & 1989Impact and CitationsGive Industry a ChanceAFundamental LinkbetweenStatistics andHumor
2019 ICSA Awards2019StudentPaperAwardsandTravelGrantsRecipients
Board Meeting at 2019 ICSA Applied Statistics
Symposium
2019 ICSA Applied Statistics Symposium Executive Committee
ICSABulletin
Volume 31/2, July, 2019ISSN 2226-2393
Editorial StaffEditor-in-Chief
Editorial Assistant
Xiaoyu [email protected]
Executive Committee
PresidentHeping [email protected]
Past PresidentAiyi [email protected]
President-ElectJianguo (Tony) [email protected]
Executive DirectorGang [email protected]
TreasurerRochelle [email protected]
From the Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
From the 2019 President, ICSA . . . . . . . . . . . . . . . . . . . 55
From the Executive Director 2017-2019 . . . . . . . . . . . . . . . 56
Results of 2019 ICSA Election . . . . . . . . . . . . . . . . . . . . 57
2019 ICSA Awards . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
New Fellows of ASA . . . . . . . . . . . . . . . . . . . . . . . . . 59
New Fellows of IMS . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Report from Statistics in Biosciences (SIBS) . . . . . . . . . . . . 60
Statistica Sinica Co-Editors’ Report . . . . . . . . . . . . . . . . . 61
Report from the Program Committee . . . . . . . . . . . . . . . . 64
Report from the 28th ICSA Applied Statistics Symposium . . . 65
2019 Student Paper Awards and Travel Grants Recipients . . . . 66
A Reminiscence of Dr. Margaret C. Wu’s Statistical Innovation
at the National Heart, Lung and Blood Institute . . . . . . . . 69
Joseph F. Heyse . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
1988 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
1989 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Impact and Citations . . . . . . . . . . . . . . . . . . . . . . . . . 81
Give Industry a Chance . . . . . . . . . . . . . . . . . . . . . . . . 82
A Fundamental Link between Statistics and Humor . . . . . . . 83
Upcoming Events . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Editorial ICSA Bulletin July 2019 Vol.31/2
From the EditorYi Huang
Dear ICSA Members,It is my great pleasure to welcome you to the 2019Jul issue of the ICSA Bulletin. This issue focuseson “Statistics: Making an Impact”, same themeas JSM2019.This issue has two featured articles showing theprofound impacts of two distinguished statisticianson government statistics and pharmaceutical statis-tics respectively. The first one is written by Dr.Colin Wu, our new ICSA 2020 President-Elect, high-lighting the tremendous influence and innovationby Dr. Margaret C. Wu on modern statisticalmethodology and data science research conductedat the National Heart, Lung and Blood Institute. Allher work focused on providing simple and scien-tifically interpretable results for complex problemsdeeply rooted in real biomedical studies, which wassummarized in four areas, such as informative cen-soring in longitudinal data, tracking and predictionfrom serial measurements, and etc.The second feature article is written by Dr. Jie Chenand Dr. Lisa Lupinacci, on Dr. Joseph F. Heyse, oneof the most respected statisticians and influential in-dustry leaders“who helped cultivate the culture ofstatistical research at Merck and in the biopharma-ceutical industry”(Jie Chen). Dr. John Tukey saidJoe was one of the most creative statisticians he evermet. He had touched many people’s lives and in-spired hundreds of statisticians through work andmentoring and collaborations in his life. Luck-ily, I was one of them. His profound influence inbiopharmaceutical statistics covers health economicstatistics (cost-effectiveness), vaccine biostatistics(design studies and analysis for vaccine develop-ments, FDA approval, and post-market safety as-sessment), early development statistics, and more.This issue features five column articles. Two arein Dr. Yi Tsong’s column ”Yi’s FDA Story: WhereStatistics Met Regulation”. With more than 30yrsof FDA experience as a statistician and leader, Yishares his remarkable personal experience and in-sider stories about important FDA events with agreat sense of humor. In 1988 article, Yi shared hispersonal story of 1988 AIDS ACT-UP demonstrationand the critical thoughts on Food and Drug Admin-istration Act and Spontaneous Reporting System.In 1989 article, he shared his personal experience of
the safety assessment of Accutane (Tretinoin) usingthe adverse event reporting system database. Doyou know the connection between the book “TheHot Zone”and Reston city in VA and Ebola virus?The 1989 article has the answer. Dr. Hans RudolfKünsch shared his critical thoughts and inspiringcomments on how to think about impact and cita-tions of research publications in his column“Hintsfrom Hans”. Dr. Terry Speed promoted researchersto join the industry and make more direct impact inreal world through employment, or research collab-orations, or graduate trainings, under his column“Terence’s Stuff”. Dr. Xiao-Li Meng explains a fun-
damental link between statistics and humor in“XL-Files”. Want to know why a student asked Prof.Casella about standard deviation always being six?Check out this column article, please.Turning to ICSA business, this issue includes theletter from 2019 ICSA President, Dr. Heping Zhangand the letter from the executive director, Dr. GangLi, highlighting selective conferences and profes-sional activities of ICSA under their leadership.After ICSA election this July, the results on 2020President-Elect and directors of ICSA board (2020-2012) are posted here. Additionally, this issue in-cludes recipients of the 2019 ICSA awards; new fel-lows of ASA and IMS in ICSA family; Statistics inBiosciences (SIBS) co-editors’ report; Statistica Sinicaco-editors’ report; reports from the 2019 programcommittee; reports from the 28th ICSA Applied Sta-tistical Symposium; 2019 student paper awards andtravel grant recipients, and the announcements ofmultiple upcoming meetings/ conferences at theend of this issue, especially the 2020 ICSA AppliedStatistics Symposium.I would like to thank all the authors and contrib-utors, ICSA executives and committees for theirstrong support and enthusiasm in the ICSA Bulletin,and Xiaoyu Cai for putting it together. Enjoy.
Yi Huang, Ph.D.Editor-in-Chief, ICSA BulletinAssociate ProfessorDepartiment of Mathematics andStatisticsUniversity of Maryland, BaltimoreCounty
54 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 From the ICSA Executives
From the 2019 President, ICSAHeping Zhang
Dear ICSA Members,It has been a great pleasure
and honor to serve as the pres-ident of the ICSA, and this op-portunity enables me to wit-ness and experience the dy-namic and exciting activitiesthat have been happening in
our community. Being established in 1987, theICSA now has over 1,500 members including a fewhundreds of students who we offer free member-ship.
This year began with the successful data sci-ence conference that took place on January 11-13,2019, in Xishuangbanna, Yunnan, China. It was co-organized by ICSA, Yunan Applied Statistical Asso-ciation, and Shanghai Jiaotong University. I wouldlike to thank Professor Ming Yuan at Columbia Uni-versity, the program committee that he chaired, andthe local committee for their hard work and contri-bution.
I had the pleasure to attend and speak briefly inthe highly successful 2019 Applied Statistics Sym-posium and 2019 ICSA China Conference. The2019 Applied Statistics Symposium was held onJune 9-12, 2019 at the Raleigh Convention Centerin Raleigh, NC, and the organization committeewas led by Professor Wenbin Lu, at the North Car-olina State University. This Symposium featuredthree keynote speakers: Marie Davidian at NorthCarolina State University, Steve Ruberg at AnalytixThinking, and Lilly Yue at U.S. Food and DrugAdministration, representing academia, industry,and government agencies. The 2019 ICSA ChinaConference took place on July 1-July 4, 2019 onthe campus of Nankai University in Tianjin, China.This Conference was organized jointly by the ICSA,Nankai University and Shanghai Jiaotong Univer-sity. Professor Zhezhen Jin at Columbia Universitychaired the organization committee. Hongyu Zhaoat Yale University and Lixin Zhang at Zhejiang Uni-versity gave keynote lectures on“Statistical Meth-ods for Genetic Risk Prediction”and “Adaptive-randomization—Models and Theory,”respectively.In both meetings, prizes were given to a number ofstudent participants for their excellent research andpresentations. I wish to thank Professors Lu and Jinand their committees for the amazing successes.
To promote next-generation leaders in our com-
munity, the ICSA established an OutstandingYoung Investigator award this year. The ICSAaward committee chaired by Professor Yufeng Liuat the University of North Carolina, Chapel Hill,selected Professor Xi Chen at New York Univer-sity, Xiaohui Chen at University of Illinois Urbana-Champaign, and Gongjun Xu at University ofMichigan as the 2019 recipients. In addition, Pro-fessor Tony Cai at University of Pennsylvania is the2019 ICSA Achievement Award winner. I wouldlike to thank Professor Liu for leading the ICSAaward committee, and congratulate these outstand-ing award winners.
The 11th ICSA International Conference willbe held at Zhejiang University, Hangzhou, Zhe-jiang, China, from December 20, 2019 to Decem-ber 22, 2019. The organizing committee is chairedby Hongzhe Li at University of Pennsylvania. Thisinternational conference will feature three keynotespeakers Jianqing Fan at Princeton University, Zhil-iang Ying at Columbia University, and HongyuZhao at Yale University. Professor Fan will deliverthe inaugural Peter Hall lecture, and Professor Zhaowill be the recipient of 2020 Pao-Li Hsu Award. I amexcited and look forward to the conclusion of 2019on a high note.
The establishment and growth of ICSA wouldnot be possible without the dedication of many vol-unteers, especially Dr. Gang Li, the ICSA Execu-tive Director, and his successor, Professor MenglingLiu. It has been my privilege and great pleasure tohave been working with Dr. Li, and more recentlywith Professor Liu. The ICSA has been unfortu-nate to have them leading the daily operation. Alsoimportantly, the ICSA committees and their chairshave done outstanding jobs. For example, Profes-sor Jianqing Fan led the Special Lectures and hiscommittee selected a list of outstanding ICSA mem-bers who will deliver special lectures in future ICSAconferences. Professor Annie Qu at University ofIllinois Urbana-Champaign chaired the nominationcommittee that recommended well-qualified candi-dates for the leadership of the ICSA in the next fewyears. Professor Peter Song at University of Michi-gan chairs the Program Committee that coordinatesall of the exciting activities related to the ICSA meet-ings. I also wish to mention that it has been arewarding experience to serve in the ICSA Execu-tive Committee and have the opportunity to workclosely with Professors Tony Cai (a former ICSApresident), Aiyi Liu (the ICSA past-president), and
© www.icsa.org 55
From the ICSA Executives ICSA Bulletin July 2019 Vol.31/2
Tony Sun (the ICSA president-elect). There are toomany individuals whom I feel grateful for theircontribution to the ICSA, including all editors andthose working at the ICSA office.
Thanks to the outstanding and dedicated contri-bution of our members, ICSA is now a major asso-ciation in the statistical society. We are strong andthriving, but we also face challenges like every othersociety. With your support, we will be stronger!
Thank you all!
Heping Zhang, Ph.D.2019 President, ICSASusan Dwight Bliss Professor of BiostatisticsProfessor in the Child Study Center and Professor ofStatistics and Data ScienceYale School of Public HealthYale University
From the Executive Director 2017-2019Gang Li
Dear ICSA members,We have a very productive
2019—all our planned projectshave progressed smoothly.The three ICSA conference
were held with great success. The 2019 ICSA Con-ference on Data Science was held on January 11-13, 2019, at Xishuangbanna, Yunnan, China, withabout 100 attendees. The 2019 ICSA Applied Statis-tics Symposium was held on June 9-12, 2019 atRaleigh, NC. The theme of this conference was Mod-ernizing Statistics for Emerging Data Science, in recog-nition of a new era for statisticians with differentchallenges and opportunities with more than 500attendees. The 2019 ICSA China Conference tookplace on July 1-4, 2019, on the campus of NankaiUniversity in Tianjin, China. The Conference wasorganized jointly by ICSA, Nankai University, andShanghai Jiaotong University. The conference pro-vided a platform for the exchange of recent researchand developments in modern statistical methods,to create collaboration opportunities and to iden-tify new directions for further research. We ap-preciate our organization committee members andpartners in China, including Yunnan University,Nankai University, and Shanghai Jiaotong Univer-sity, for their tireless devotions to the preparationfor conferences and collaborations.
The 11th ICSA International Conference isscheduled on December 19 22, 2019, in Hangzhou,China. The ICSA International Conference is themost prestigious conference of ICSA and heldevery 3 years. The theme of this conferenceis Innovation with Statistics and Data Science.We envision that this meeting will attract indus-trial statisticians and many international statisti-cians as well as statisticians working in govern-ment and academia. Details on the conferencecan be found at http://cds.zju.edu.cn/ICSA2019.aspx?k1=4&k2=79&k3=80.
To support young statisticians in their career de-velopment, ICSA gives 1 2 scholars the Young Re-searcher Awards each year. The ICSA Young Re-searcher Awards are given to three scholars in 2019for the first time. The awards will be presented tothem at the ICSA Member Meeting at the Joint Sta-tistical Meeting in Denver, CO. Now it is time foryou to consider the nomination for the candidatesnext year.
Let us work together for the continued successof ICSA. Please let me know if you have any ideasor suggestions.
Gang Li, Ph.D.ICSA Executive Director (2017-2019)Director, Statistics and Decision SciencesJanssen Research & Development, LLC
56 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 ICSA Reports
Results of 2019 ICSA Election2020 President-ElectColin Wu (National Institutes of Health)
Directors of ICSA Board (2020-2022)(alphabetical order)
• Jason Liao (Merck)
• Bin Nan (University of California)
• Peihua Qiu (University of Florida)
• Jane Zhang (Allergan)
• Yichuan Zhao (Georgia University)
2019 ICSA AwardsDistinguished AchievementAwardIn recognition of the distinguished achievement in statis-tical research and unselfish support of the association.
Tony Cai Ph.D., University of Pennsylvania,Philadelphia, PA
For fundamental contri-butions to high-dimensionalstatistical inference, adaptivenonparametric function, andminimax optimality, for out-standing mentorship, and forexceptional leadership andservice to our profession.
Outstanding Service AwardIn recognition and with sincere appreciation for the ded-icated effort, unselfish support and outstanding service.
Tony Cai Ph.D., University of Pennsylvania,Philadelphia, PA
In recognition ofhis outstanding serviceand leadership as 2017ICSA President and asa member of the ICSABoard of Directors (2009-2011).
Gang Li Ph.D., University of California at Los An-gles, CA
For his outstanding ser-vices to ICSA including or-ganizing multiple highly suc-cessful ICSA conferences in-cluding the 10th ICSA Inter-national Conference and the2018 ICSA China Conference,promoting ICSA and its mis-
sion in Asia, and dedicated ICSA committee ser-vices.
Hongliang Shi Blueprint Medicines Inc., Cam-bridge, MA
In recognition of her ded-icated service as the ICSATreasurer 2016-2018, the ICSASymposium Treasurer 2013-2015, and the local committeeChair of ICSA 2012 AppliedSymposium.
President’s CitationIn grateful appreciation of the generosity, dedication anddevoted effort for ICSA.
© www.icsa.org 57
ICSA Reports ICSA Bulletin July 2019 Vol.31/2
Chengyong Tang Ph.D., Temple University,Philadelphia, PA
In recognition and ap-preciation of his ded-icated and outstand-ing service and lead-ership as the Chair ofthe Program Committeefor the 2018 ICSA Ap-plied Statistics Sympo-sium.
Min-ge Xie Ph.D., Rutgers University, Piscat-away, NJ
In recognition of his dedi-cated services to ICSA, includ-ing serving as the co-chair ofthe Organizing Committee forthe 2018 ICSA Applied Statis-tics Symposium.
Outstanding Young ResearcherAward
In recognition of the outstanding research in statisticaltheory, methodology, and/or applications.
Xi Chen Ph.D., New York University, New York,NY
For his impact on bridg-ing between machine learn-ing, statistics, and optimiza-tion, in particular, his novelcontributions to statistical in-ference based on stochasticoptimization, distributed datainference, sequential learning,and their applications to rev-
enue management, and crowdsourcing.
Xiaohui Chen Ph.D., University of Illinois atUrbana-Champaign, IL
For fundamental contri-butions to high-dimensionalstatistics, time series anal-ysis and statistical ma-chine learning, in partic-ular to Gaussian approxi-mation to U-statistics andcovariance matrix estima-tion.
Gongjun Xu Ph.D., University of Michigan, AnnArbor, MI
For his fundamental con-tributions to latent-class mod-els and statistical modelingand inference for educationalmeasurements.
58 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 ICSA Announcements
New Fellows of ASA• Huiman X. Barnhart, Professor of Biostatis-
tics, Duke University
• Jinbo Chen, Professor, University of Pennsyl-vania
• Haoda Fu, Senior Research Adviser, Eli Lillyand Company
• Jianhua Hu, Professor, University ofColumbia
• Hongkai Ji, Professor, Johns HopkinsBloomberg School of Public Health
• Jiashun Jin, Professor, Carnegie Mellon Uni-versity
• Katerina Kechris, Professor, Colorado Schoolof Public Health
• Jia Li, Professor, Penn State University
• Yehua Li, Professor, University of Californiaat Riverside
• Pei Wang, Professor, Icahn School of Medicineat Mount Sinai
• Min Yang, Professor, University of Illinois atChicago
• Xiangrong Yin , Professor, University of Ken-tucky
• Menggang Yu, Professor, University ofWisconsin-Madison
• Lanju Zhang, Director and Research Fellow,AbbVie
• Hui Zou, Professor, University of Minnesota
New Fellows of IMSJeng-Min Chiou, Academia Sinica
For contributions to methodology for clustering,classification, and prediction with functional data.
Cynthia Rudin, Duke University
For contributions to interpretable machine learn-ing algorithms, prediction in large scale medicaldatabases, and theoretical properties of ranking al-gorithms.
Xiaofeng Shao, University of Illinois, Urbana-Champaign
For contributions to non-parametric statistical in-ference for multivariate time series, in particular tothe asymptotic theory for time series analysis via
moments and cumulants.
Yuedong Wang, University of California, SantaBarbara
For contributions to non-parametric regression andcomputational statistics, in particular smoothingspline methodology for dependent observationsand applications to bioinformatics and biomedicalmodeling.
Hongquan Xu, University of California, Los An-geles
For contributions to experimental design, computerexperiments, and functional data analysis, in partic-ular to nonregular fractional factorial designs andspacefilling designs.
© www.icsa.org 59
ICSA Announcements ICSA Bulletin July 2019 Vol.31/2
Report from Statistics in Biosciences(SIBS)Hongzhe Li and Mei-Cheng Wang
Statistics in Biosciences (SIBS) is one of the twoofficial journals established by ICSA. The journalwas established about 10 years ago and has fo-cus on development and application of statisticalmethods and their interface with other quantitativemethods, such as computational and mathematicalmethods, in biological and life science, health sci-ence, and biopharmaceutical and biotechnologicalscience. ICSA members can find more informationof the journal from the following website: http://www.springer.com/statistics/journal/12561?detailsPage=editorialBoard
SIBS publishes both regular articles and topic-oriented papers in Special Issues. The last 5 Spe-cial Issues that SIBS published includes ‘SAMSI-Beyond Bioinformatics,’ ‘Statistical Methods in Or-gan Failure and Transplantation,’ ‘Statistics and Ge-nomics: Emerging Issues and Solutions,’ ‘Statisticalmethods for clinical trials and precision medicine’and ‘Challenges in Computational Neuroscience.’In July 2019 the journal will publish a special issueon ‘Medical Device Data,’ which is guest edited by
Jaroslaw Harezlak and Chongzhi Di. As the ICSAhas become the 4th largest statistical association inthe world, we expect SIBS to grow into a leadingjournal over time.
Prof. Hongzhe Li/Hongzhe LeeProfessor of Biostatistics and Statis-ticsDirector, Center for Statistics in BigDataVice Chair for Integrative ResearchDepartment of Biostatistics, Epi-demiology and InformaticsUniversity of Pennsylvania
Mei-Cheng Wang, Ph.D.ProfessorDepartment of Biostatistics JohnsHopkins Bloomberg School of PublicHealth
60 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 ICSA Announcements
Statistica Sinica Co-Editors’ ReportYuan-chin Ivan Chang, Hans-Georg Muller and Yazhen Wang
Submissions and Acceptance Statistics
In the past 12 months (August 1, 2018 to July 31, 2019), 450 manuscripts were submitted to Statistica Sinica,which include one submission to a special issue and nine comments to a discussion paper (see Table 1). Thenumbers of manuscripts submitted and accepted for the past six years are shown in Table 2. The submissionand acceptance rates during 2015 to 2017 is higher because of the special issues, and the rate in the past 11months is almost the same as the rate before 2015. The review status for the past three years is displayed inTable 3, and the top 10 countries with the highest submissions for the past three years are shown in Table 4.
© www.icsa.org 61
ICSA Announcements ICSA Bulletin July 2019 Vol.31/2
Manuscript Processing TimeTable 5 shows the turnaround statistics of initial decision for the past three years, with the decision timescensored on July 2, 2019. About 75% of the editorial decisions during 2018-2019 take less than 81 days, but5% take over 138 days.
Backlog for PublicationIn the past year, we have published five issues containing 133 articles. There remain 156 accepted manuscriptswaiting to be published. Among them, 32 will appear in two special issues in Oct 2017. The backlog is about16 months from acceptance to publication.
Rankings and Impact FactorsTable 6 shows the ranks of Statistica Sinica based on the 2-Year Impact Factor and the 5-Year Impact Factorprovided by the Journal Citation Reports (JCR) in the area of Statistics and Probability from 2010 to 2018.Table 7 shows the ranks of Statistica Sinica in Scimago Journal Rankings among all journals of Statistics and
62 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 ICSA Announcements
Probability in the Scopus database from 2010-2018. The ranking is performed using the algorithm GooglePageRank.
© www.icsa.org 63
ICSA Announcements ICSA Bulletin July 2019 Vol.31/2
Report from the Program CommitteePeter X.K. Song
During the period of January to July, 2019, ICSA hassuccessfully organized two major conferences. Oneis the ICSA Applied Statistical Symposium held atthe Raleigh Convention Center in Raleigh, NC dur-ing June 9–12, 2019 and the other is the 2019 ICSAChina Conference held in Tianjin, China during July1–4, 2019.
Chaired by Dr. Wenbin Lu, Professor at theDepartment of Statistics of North Carolina StateUniversity, the Applied Statistical Symposium wasthemed as “Modernizing Statistics for EmergingData Science”in recognition of a new era for statis-ticians with different challenges and opportunities.It attracted more than 400 participants and offered 9short courses and 103 scientific sessions, includingthree keynote speeches, one featured session andtwo student paper award sessions, as well as excit-ing social events. The second was hosted jointly byICSA, Nankai University and Shanhai Jiaotong Uni-versity. It took place on the campus of Nankai Uni-versity on July 1-4, 2019. Chaired by Dr. ZhezhenJin (Chair), Professor at the Department of Biostatis-tics of Columbia University, the scientific programof the conference focused on creating collabora-tion opportunities and identifying new directionsfor further research. It also attracted over 400 par-ticipants, offered over 80 invited sessions and onelarge invited poster session, including two plenaryspeakers, and gave out five Junior Researcher PaperAwards and one Junior Researcher Poster Award.Both conferences were big successes! In addition totimely and high-quality scientific programs assem-bled by the respective program committees, the suc-cess was also built critically upon the great contri-butions and support of the large numbers of volun-teers and co-sponsors.
In addition to the two major conferences above,ICSA has also organized various other activitiesin the first half of year 2019, including some co-sponsored conferences. Among them, the ICSAConference on Data Science was held during Jan-uary 11-13, 2019 in Xishuangbanna, Yunnan, China,the Duke Industrial Statistics Symposium in thetheme of Innovative Design and Analysis for Com-
plex Clinical Trials for Drug and Device Develop-ments took place in Durham, NC in April 10-12,2019, and the 9th IMS-FIPS Workshop was held onJune 15-16, 2019 in Shanghai, China.
By the time this comes out, ICSA has alreadygeared up with additional exciting events in thesecond half of year 2019. The 11th ICSA Interna-tional Conference will be held Hangzhou, China onDecember 20-22, 2019. The theme of this confer-ence concerns Innovation with Statistics and DataScience, and the scientific program committee ischaired by Dr. Hongzhe Li. The ICSA-CanadaChapter will hold a conference“Advances and In-novations in Statistics and Data Science”on August9-11, 2019 in Kingston, ON, Canada. In addition,several ICSA-sponsored international conferencesinclude a Conference on Current Trends in SurveyStatistics on August 13-16, 2019 in Singapore, the6th International Symposium on Biopharmaceuti-cal Statistics (ISBS2019) on August 26-30, 2019 in Ky-oto, Japan, and the 11th International Conferenceon Multiple Comparison Procedures (MCP) on De-cember 12-15, 2019 in Taipei.
While the ICSA program committee is workingwith several hosts to finalize both scientific pro-grams and logistic matters regarding ICSA’s majorsymposia and conferences to be held in year 2020-2021, we would like to invite organizations whoare interested in holding “2022 ICSA-China con-ference”and “the 12th ICSA International Con-ference”to contact the committee chair Peter Song([email protected]). Your support to ICSA ishighly appreciated!
Peter X.K. Song, Ph.D.ProfessorAssociate Chair, ResearchDepartment of BiostatisticsSchool of Public HealthUniversity of Michigan, Ann Arbor
64 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 ICSA Announcements
Report from the 28th ICSA AppliedStatistics SymposiumWenbin Lu and Donglin Zeng
The 28th ICSA Applied Statistics Symposium wassuccessfully held at Raleigh Convention Center,North Carolina, during June 9-12, 2019. The themeof the conference is Modernizing Statistics forEmerging Data Science, in recognition of a new erafor statisticians with different challenges and op-portunities. Lilly Yue of U.S. Food and Drug Ad-ministration, Stephen Ruberg of Analytix Think-ing, and Marie Davidian of North Carolina StateUniversity delivered keynote presentations. Therewere 99 invited sessions, 1 student paper award ses-sion and 3 contributed sessions. In addition, therewere 9 short courses including 6 half day and 3full day courses, covering topics from clinical tri-als, causal inference, precision medicine to networkdata analysis and deep learning for big data. Over500 participants working in academia, government,and industry from all over the world attended theconference. Over 150 participants attended the con-ference banquet. David Banks of Duke Universitydelivered an inspiring banquet speech. The topic ofDavid’s speech was about how Chinese researchershave changed the world of statistics.
The conference received 35 applications for thestudent paper award competition. Zhou Lan ofNorth Carolina State University, Jitong Lou of Uni-versity of North Carolina at Chapel Hill, Tao Sun ofUniversity of Pittsburgh, Xiangyu Liu of Universityof Texas Health Science Center at Houston and Jen-nifer Starling of University of Texas at Austin wereselected for the student paper award and XizhouGuo of University of Michigan and Ann Arbor wasselected for the Jiann-Ping Hsu Pharmaceutical and
Regulatory Sciences Student Paper Award. The sixawardees received the awards during the confer-ence banquet and presented their papers in a spe-cial student paper award session. In addition, to en-courage more students to participate, the programcommittee offered six student paper contributionawards so that they can present their works in a con-tributed session of the conference.
The program committee of the 28th ICSA Ap-plied Statistics Symposium would like to thankour sponsors: AbbVie, Boehringer Ingelheim, FMDK&L, Merck, Pfizer, Sanofi Pasteur, SAS and Tech-Data for the generous support. Special thanksalso go to the student volunteers of Duke Univer-sity, North Carolina State University, University ofNorth Carolina at Chapel Hill for their great effortsand hard work for providing the excellent servicesto the conference.
Wenbin Lu, PhDChair of Organizing CommitteeProfessor of StatisticsNorth Carolina State University
Donglin Zeng, PhDChair of Scientific Program Com-mitteeProfessor of BiostatisticsUniversity of North Carolina atChapel Hill
© www.icsa.org 65
ICSA Announcements ICSA Bulletin July 2019 Vol.31/2
2019 Student Paper Awards and TravelGrants RecipientsA student paper award competition was held forthe 2019 ICSA Applied Statistics Symposium. Theaward committee consists of
• Dr. Luo Xiao, Chair, North Carolina State Uni-versity
• Dr. Gen Li, Columbia University
• Dr. Meng Li, Rice University
• Dr. Suyu Liu, University of Texas MD Ander-son Cancer Center
• Dr. Min Qian, Columbia University
• Dr. Rui Song, North Carolina State University
• Dr. Yang Song, Vertex Pharmaceuticals Inc.
• Dr. Yanxun Xu, Johns Hopkins University
• Dr. Lingzhou Xue, Pennsylvania State Univer-sity
• Dr. Shu Yang, North Carolina State University
• Dr. Jiajia Zhang, University of South Carolina
• Dr. Yichuan Zhao, Georgia State University
• Dr. Yingqi Zhao, Fred Hutchinson Cancer Re-search Center
• Dr. Ruoqing Zhu, University of Illinois atUrbana-Champaign
• Dr. Yunzhang Zhu, Ohio State University
Congratulations to the following 2019 ICSAAwards Recipients!
ICSA Applied Statistical Symposium Student PaperAwards
• Zhou Lan, North Carolina State University
• Xiangyu Liu, University of Texas Health Sci-ence Center at Houston
• Jitong Lou, University of North Carolina atChapel Hill
• Jennifer Starling, University of Texas at Austin
• Tao Sun, University of Pittsburgh
Jiann-Ping Hsu Pharmaceutical and Regulatory Sci-ences Student Paper Award
• Xinzhou Guo, University of Michigan
We heartily thank all of the participants for theircontributions. We congratulate the winners fortheir great achievements. The abstracts of the win-ning papers are as follows.
Authors: Zhou Lan, Brian J Reich, Joseph Guinnessand Dipankar BandyopadhyayTitle: Geostatistical Modeling of Positive DefiniteMatrices Using the Spatial Wishart ProcessAbstract: Geostatistical modeling for continuouspoint-referenced data has been extensively appliedto neuroimaging because it produces efficient andvalid statistical inference. However, diffusion ten-sor imaging (DTI), a neuroimaging characteriz-ing the brain structure produces a positive defi-nite (p.d.) matrix for each voxel. Current geo-statistical modeling has not been extended to p.d.matrices because introducing spatial dependenceamong positive definite matrices properly is chal-lenging. In this paper, we use the spatial Wishartprocess, a spatial stochastic process (random field)where each p.d. matrix-variate marginally followsa Wishart distribution, and spatial dependence be-tween random matrices is induced by latent Gaus-sian processes. This process is valid on an uncount-able collection of spatial locations and is almostsurely continuous, leading to a reasonable meansof modeling spatial dependence. Motivated by aDTI dataset of cocaine users, we propose a spatialmatrix-variate regression model based on the spa-tial Wishart process. A problematic issue is thatthe spatial Wishart process has no closed-form den-sity function. Hence, we propose approximationmethods to obtain a feasible working model. Alocal likelihood approximation method is also ap-plied to achieve fast computation. The simulationstudies and real data analysis demonstrate that theworking model produces reliable inference and im-proved performance compared to other methods.
Authors: Xiangyu Liu, Jing Ning, Xuming He andRuosha LiTitle: Semiparametric Modelling and Estimation of
66 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 ICSA Announcements
the Global Percentile OutcomeAbstract: Biomedical studies are often challenged bythe lack of a single primary outcome that can com-prehensively capture the multidimensional impair-ments and symptoms of a disease. To address this,global composite outcomes are commonly formu-lated to integrate multiple individual outcomes andto achieve comprehensive assessments of the over-all disease severity and burden. The global rank-sum outcome and the corresponding K-sample testprocedures have been successfully applied in manyclinical studies, but existing methods do not lendthemselves to regression modeling. In this work,we consider sensible regression strategies to eval-uate covariate effects on the global percentile out-come (GPO), under the transformed linear modeland the monotonic index model respectively. Pos-ing minimal and realistic assumptions, we developestimation and inference procedures that accountfor the special features of the GPO. Asymptotic re-sults were established rigorously using Ustatisticand U-process techniques. Simulation studies sug-gest that the proposed methods perform satisfacto-rily under realistic sample sizes. An application to aParkinson’s disease dataset illustrates the practicalutilities of the proposed methods
Authors: Jitong Lou, Yuanjia Wang, Lang Li andDonglin ZengTitle: Integrative Analysis of Irregularly Measuredand Mixed-Type Biomarkers in Electronic HealthRecords Method with Statistical GuaranteesAbstract: Electronic health records (EHRs) has in-creasingly become an important data source for per-sonalized medicine. In EHRs, disease biomark-ers from the same patient are recorded longitudi-nally at clinical encounters. In order to compre-hensively assess patient’s disease comorbidity andsusceptibility, it is necessary to characterize thesebiomarkers over time in an integrative way. How-ever, there exist some challenges such as, in EHRs,the biomarkers are measured sparsely at irregularand informative clinical encounters. In this paper,we propose multivariate generalized linear mod-els to analyze mixed-type biomarkers, where latentmultivariate Gaussian temporal processes are intro-duced to capture between-marker dependence overtime. We allow the covariate effects and covariancematrix of the latent processes to be time-dependent.For inference, we apply kernel-weighted estimatingequations based on the method of moments wherekernel weights account for the heterogeneous in-tensity of measurement times. We investigate theasymptotic properties of the derived estimators.
Under the multivariate models, we integrate theirregularly measured biomarkers of mixed modesinto composite scores that reflect patients’under-lying health status. We illustrate the finite-sampleperformance of our method through extensive sim-ulation studies. Lastly, we apply our method to an-alyze a large sample of EHRs of Type 2 Diabetes(T2D) patients and show its utility to character-ize patients’comorbidity and disease progressionwhile accounting for challenges of EHRs.Keywords:Electronic health records, latent process, kernelsmoothing, generalized linear models, method ofmoments, type 2 diabetes
Authors: Jennifer StarlingTitle: BART with Targeted Smoothing: an Analysisof Patient-Specific Stillbirth RiskAbstract: We introduce BART with TargetedSmoothing, or tsBART, a new Bayesian tree-basedmodel for nonparametric regression. The goal oftsBART is to introduce smoothness over a singletarget covariate t, while not necessarily requiringsmoothness over other covariates x. TsBART isbased on the Bayesian Additive Regression Trees(BART) model, an ensemble of regression trees. Ts-BART extends BART by parameterizing each tree’s terminal nodes with smooth functions of t, ratherthan independent scalars. Like BART, tsBART cap-tures complex nonlinear relationships and interac-tions among the predictors. But unlike BART, ts-BART guarantees that the response surface will besmooth in the target covariate. This improves inter-pretability and helps regularize the estimate. Afterintroducing and benchmarking the tsBART model,we apply it to our motivating example: pregnancyoutcomes data from the National Center for HealthStatistics. Our aim is to provide patient-specific es-timates of stillbirth risk across gestational age (t),based on maternal and fetal risk factors (x). Obste-tricians expect stillbirth risk to vary smoothly overgestational age, but not necessarily over other co-variates, and tsBART has been designed preciselyto reflect this structural knowledge. The results ofour analysis show the clear superiority of the ts-BART model for quantifying stillbirth risk, therebyproviding patients and doctors with better infor-mation for managing the risk of fetal mortality. Allmethods described here are implemented in the Rpackage tsbart.
Authors: Tao Sun, Wei Chen and Ying DingTitle: GWAS-based AMD Progression Using a Cop-ula Semiparametric Model
© www.icsa.org 67
ICSA Announcements ICSA Bulletin July 2019 Vol.31/2
Abstract: The genome-wide association studies(GWAS) of Age-related Macular Degeneration(AMD), a progressive bilateral eye disease, is thefirst and most successful GWAS research, wherethe massive GWAS data provide unprecedented op-portunities to study disease risk and progression.This research is motivated by discovering geneticcauses and making accurate prediction for AMDprogression. For genetic variant identification, wedevelop a copula-based semiparametric approachfor modeling and testing bivariate censored data.Specifically, the joint likelihood is modeled througha two-parameter Archimedean copula, which canflexibly characterize the dependence between twomargins. The marginal distributions are modeledthrough a semiparametric transformation modelusing sieves, with the proportional hazards or oddsmodel being a special case. We propose a sieve max-imum likelihood estimation procedure and developa generalized score test for testing the regression pa-rameter(s). We apply our method to a genome-wideanalysis of AMD progression to identify suscepti-ble risk variants for the disease progression. Lastly,we build a novel GWAS-based survival neural net-work prediction model for AMD progression. Our
results demonstrate how the synergy of wealthyGWAS data and deep learning can effectively pre-dict survival probabilities.
Authors: Xinzhou Guo, Ruosha Li and Xuming HeTitle: A Quantitative Assessment of Risk for Sub-group Pursuit in Clinical TrialsAbstract: In clinical studies, when to recommend ordecide further pursuit of the most promising sub-group that we have observed from an existing trialis a very important question. It is well recognizedthat the observed treatment effect size of the bestidenti ed subgroup tends to be too optimistic and,therefore, any subgroup pursuit needs to be madewith a careful statistical consideration. In this pa-per, we address the issue of bias in subgroup pur-suit and provide a quantitative screening measurethat can be used in the decision-making of sub-group pursuit. In addition, we propose a con dencebound on the best subgroup treatment effect fromclinical data. The proposed quantitative analysis ismodel-free, transparent and easy to compute, andcan help make a better-informed decision of sub-group pursuit in clinical trials.
68 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 Invited Articles
A Reminiscence of Dr. Margaret C. Wu’sStatistical Innovation at the NationalHeart, Lung and Blood InstituteColin O. Wu
In today’s era of “Big Data”and “Real WorldData/Real World Evidence”(RWD/RWE), inno-vative methods for exploratory statistical analysishave been well recognized as one of the key compo-nents of data science–a multidisciplinary field thatis becoming increasingly popular across the scien-tific community. For example, Goal 3 of the NIHStrategic Plan for Data Science is“Support the De-velopment and Dissemination of Advanced DataManagement, Analytics, and Visualization Tools,”which includes “Extracting understanding fromlarge-scale or complex biomedical research data re-quires algorithms, software, models, statistics, vi-sualization tools, and other advanced approachessuch as machine learning, deep learning, and arti-ficial intelligence.”(https://datascience.nih.gov/strategicplan) The National Heart, Lungand Blood Institute (NHLBI) has also identified“leveraging emerging opportunities in data science
to open new frontiers”as one of the objectives of itsStrategic Vision (https://www.nhlbi.nih.gov/about/strategic-vision). As a statistician,it is exciting for me to see that these important objec-tives have finally garnered the tremendous amountof attention they truly deserve in the biomedical re-search community. But, I am also constantly re-minded that the field of statistics as a critical com-ponent of data science for“large-scale or complexbiomedical research data”is not new. Many earlyvisionary statisticians have made significant con-tributions to the development of critical analyticaltools that have become the integral parts of today’sdata science.
Dr. Margaret C. Wu (not related to me), a bio-statistician who spent her career at the NHLBI Of-fice of Biostatistics Research between 1973 and 2001,was one of these early visionary statisticians whoseseminal work from the late 1970’s to the early 2000shad tremendous influence on the modern statisti-cal methodology in data science. It is most no-table that Dr. Margaret Wu’s methodological con-tributions were almost entirely motivated by themany complex clinical trials and long-term observa-tional studies in which she participated at NHLBI.
Her work was primarily motivated by real scientificproblems encountered by the project investigatorsand biomedical researchers. Her statistical resultswere focused on providing simple and scientificallyinterpretable solutions with sufficient mathemati-cal rigor. This was a period when many of the com-plex data structures, practical issues and shortfallsof classical statistical techniques in large biomedi-cal studies were not widely known to the generalstatistical community. Much of the statistical termi-nology with which we are so familiar today, suchas “error-in-variable regression,” “unbalancedrepeated measurements,” “informative censor-ing,”“nonrandom missing data,”“nonignorabledropout,”and“longitudinal tracking”were non-existent prior to Dr. Margaret Wu’s work. In thisrespect, Dr. Margaret Wu is a pioneer who spear-headed many of the statistical analysis tools that wetake for granted today.
Figure 1: Dr. Margaret C. Wu
I joined NHLBI in early 2002, shortly after Dr.Margaret Wu’s retirement in September 2001. My
© www.icsa.org 69
Invited Articles ICSA Bulletin July 2019 Vol.31/2
appreciation of Dr. Margaret Wu and her NHLBIcolleagues’ innovative work began when I startedparticipating in several high impact NHLBI long-term epidemiological studies. The complexity ofpractical data and potential mechanisms of biolog-ical processes made me understand that seeminglyelegant theoretical results could be worthless andsoon forgotten if they could not provide meaning-ful solutions to real problems. The beauty of thetrailblazing work by Dr. Margaret Wu and her col-laborators rests on its applicability to real practicalsettings(something close to today’s“Big Data”and“RWD”). The corresponding theoretical develop-
ments are all based on practically meaningful as-sumptions and meant to provide useful insight andinterpretations. To make a connection between Dr.Margaret Wu’s publications and potential researchquestions which we may face today, I briefly com-ment on the following four areas:
Informative Censoring in Longitudinal Data. In arecent special issue paper in Statistics in Medicine,Albert (2019) [1] provided a systematic review ofone of Dr. Margaret Wu’s most influential contri-butions on statistical methods: correcting nonig-norable dropout(or nonrandom missing data) biasin longitudinal data analysis with the shared ran-dom parameter models. These methods were mo-tivated by the realization that most high impactand long-term biomedical studies have complex re-peated measurements, and missing data in thesestudies often are not “missing at random.”Al-though most of Dr. Margaret Wu’s publicationsin this area appeared in the 1980’s and 1990’s [3,8, 9, 10, 11, 12, 13, 14] , the methods developedin these papers are still the standard approachestoday for large studies with complex longitudinaldata structures. One reason that Dr. Margaret Wuand her colleagues’ methodological contributionshave stood the test of time is that these methodsare interpretable and tractable. Indeed, these meth-ods can provide useful insights into the underlyingrelationships between longitudinal variables withcomplex data structures. Moreover, they rely ontractable mathematical structures which are com-putationally feasible. Although some of the “BigData”and“RWD”problems we are facing todaymay be more complicated than the data structuresdiscussed in Dr. Margaret Wu’s publications, suchas the potential selection bias caused by incorpo-rating datasets from multiple different longitudinalstudies, I believe that the basic framework devel-oped in her work lays the foundation for future re-search in developing innovative“RWE”analyticalmethods.
Tracking and Prediction from Serial Measurements.A main advantage of a long-term longitudinalstudy over a simpler cross-sectional study is that,with repeated measurements over time, it is possi-ble to estimate and predict a subject’s specific out-comes in the future without solely resorting to thepopulation averages. This type of“subject-specificprediction”(or simply termed“tracking”) is par-ticularly useful, for example, in pediatric studies,where it is unlikely to observe hard time-to-eventhealth outcomes, such as death or hospitalization.In Dr. Margaret Wu’s work with Professor JamesH. Ware in 1981 [5], a practical solution for individ-ual outcome tracking and prediction is provided us-ing the now well-known linear mixed effects model.Unfortunately, it took several years for the originalwork of Ware and Wu (1981) to attract widespreadattention. We now recognize its potential for min-ing big longitudinal data or electronic health recorddata for health status and disease risk predictions.
Clinical Trial Designs with Repeated Measurements.Since almost every major clinical trial conducted atNHLBI has some repeated measurements, design-ing clinical trials with repeatedly measured out-comes is also a major theme in Dr. Margaret Wu’sresearch in the 1980’s and 1990’s [6, 7, 15]. Theseare all clever trial designs because they take intoaccount the challenges of subject withdrawal, stag-gered entry, informative censoring, and sequentialmonitoring. In my opinion, these ideas for trial de-signs were far ahead of their time, because the con-cepts of informative censoring and other potentialcomplications were not well understood by manyclinical trial practitioners in the 1980s and 1990s.However, the push for “RWD/RWE”and prag-matic clinical trials demands pragmatic trial de-signs which take into account these challenges. Ibelieve these early ideas of Dr. Margaret Wu andher colleagues will continue to be quite useful forfuture clinical trial designs.
Influential Statistical Practice. In addition to hermany methodological contributions, Dr. MargaretWu’s work on statistical practice has also motivatedmany influential statistical methods developed byother well-known statisticians. A notable exampleis Halperin, Wu and Gordon (1979) [4] in which theauthors discovered through an application to the fa-mous Framingham Heart Study that the differencesof baseline risk variables between cases and non-cases in a cohort study should be evaluated throughan appropriate mathematical model. This findingwas noted in Carroll et al (1984) [2] and served as aprecursor for developing the well-known“error-in-variable regression,”which is now a major branch
70 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 Invited Articles
of statistics that is widely used in real applications.Although it is impossible to comment on the full
impact of Dr. Margaret Wu and her colleagues’ sta-tistical research, I hope that this essay serves as anappreciation of their trailblazing work. Many oftheir innovative ideas were ahead of their time andprovide creative insight into current statistical re-search and data science. We are indebted to theirefforts.
References
1. Albert PS. (2019) Shared random parametermodels: A legacy of the biostatistics programat the National Heart, Lung and Blood Insti-tute. Statistics in Medicine, 38(4):501-511.
2. Carroll RJ, Spiegelman CH, Lan GKK, Bai-ley KT and Abbott RD. (1984) On errors-in-variables for binary regression models.Biometrika, 71(1):19-25.
3. Follmann D and Wu MC. (1995) An approx-imate generalized linear model with randomeffects for informative missing data. Biomet-rics, 51(1): 151-168.
4. Halperin M, Wu MC and Gordon T. (1979)Genesis and interpretation of differences indistributions of baseline characteristics be-tween cases and non-cases in cohort studies.Journal of Chronic Diseases, 32:483-491.
5. Ware JH and Wu MC. (1981) Tracking: pre-diction of future values from serial measure-ments. Biometrics, 37:427-437.
6. Wu MC. (1988) Sample size for comparisonof changes in the presence of right censoringcaused by death, withdrawal, and staggeredentry. Controlled Clinical Trials, 9(1):32-46.
7. Wu MC. (1994) Design aspects for clinical tri-als with repeated measures in the presence ofinformative censoring. Journal of StatisticalPlanning and Inference, 42(1-2):109-122.
8. Wu MC, Albert PS and Wu BU. (2001) Adjust-ing for drop-out in clinical trials with repeatedmeasures: design and analysis issues. Statis-tics in Medicine, 20(1):93-108.
9. Wu MC and Bailey K. (1988) Analysingchanges in the presence of informative rightcensoring caused by death and withdrawal.Statistics in Medicine, 7(1-2):337-346.
10. Wu MC and Bailey KR. (1989) Estimation andcomparison of changes in the presence of in-formative right censoring: conditional linearmodel. Biometrics, 45(3):939-955.
11. Wu MC, Borkowf CB and Albert PS. (2001)Analysis of longitudinal quality of life datawith informative dropout. In book: Statis-tics Methods for Quality of Life Studies.DOI:10.1007/978-1-4757-3625-0_28.
12. Wu MC and Carroll RJ. (1988) Estimation andcomparison of changes in the presence of in-formative right censoring by modeling thecensoring process. Biometrics, 44:175-188.
13. Wu MC and Follmann DA. (1999) Use of sum-mary measures to adjust for informative miss-ingness in repeated measures data with ran-dom effects. Biometrics, 55(1):75-84.
14. Wu MC, Hunsberger S and Zucker DM. (1994)Testing for differences in changes in the pres-ence of censoring: Parametric and nonpara-metric methods. Statistics in Medicine, 13(5-7)635-646.
15. Wu MC and Lan GKK. (1992) Sequential mon-itoring for comparison of changes in a re-sponse variable in clinical studies. Biometrics,48(3):756-779.
16. Wu MC, Lan GKK and Connett JE. (1994) Useof surrogate information time for monitoringthe effect of treatment on the change in a re-sponse variable in clinical trials. Statistics inMedicine, 13(9):945-953.
17. Zerbe GO, Wu MC and Zucker DM. (1994)Studying the relationship between changeand initial value in longitudinal studies.Statistics in Medicine, 13(5-7):759-768.
18. Zucker DM, Zerbe GO and Wu MC. (1995)Inference for the association between coeffi-cients in a multivariate growth curve model.Biometrics, 51(2):413-424.
Acknowledgement: The views expressed in thisessay are those the author’s and do not necessarilyrepresent the views of the NHLBI, the National In-stitutes of Health, or the US Department of Healthand Human Services. I would like to thank Drs.Nancy L. Geller, Eric Leifer, Xin Tian, James Troen-dle and Myron Waclawiw for comments and sug-gestions.
© www.icsa.org 71
Invited Articles ICSA Bulletin July 2019 Vol.31/2
Colin O. Wu, Ph.D.Office of Biostatistics Research,National Heart, Lung and Blood In-stitute,National Institutes of Health
Joseph F. HeyseJie Chen and Lisa Lupinacci
A veteran statistician who helped cultivate the culture ofstatistical research at Merck and in the biopharmaceuti-cal industry
Joseph F. Heyse, one of the most respected andinfluential leaders in biopharmaceutical statistics,passed away Friday, May 31, 2019. He was 67.
Joe joined Merck & Co., Inc., in 1976 as a statis-tician supporting preclinical research and immedi-ately began to have a profound influence on thestatistics profession by developing statistical meth-ods that subsequently became industry standards.
Figure 1: Dr. Joseph F. Heyse
In 1987, Joe began supporting clinical pharma-ceutical research, providing technical oversight forhealth economics studies to support the market-ing of Merck products. In 1990, he was namedthe director of a newly established department ofHealth Economic Statistics, which furnished cost-effectiveness evaluations of all Merck products, andproceeded to become a well-recognized leader inthe field of health economic statistics.
In 1993, Joe assumed responsibility for the Clini-cal Biostatistics group at Merck’s Pennsylvania site,including Vaccine Biostatistics which was under-going unprecedented growth. His impact on hu-man health in this role is best summarized by Se-nior Vice President of Medical Affairs, Eliav Barr,who said, ”Joe led the [clinical] statistics group inan era of extraordinary challenges and productiv-ity at Merck. At that time, Merck developed vac-cines against rotavirus, a leading cause of infantmorbidity and mortality; human papillomavirus,which causes most cases of cervical, genital, andhead/neck cancer; and varicella zoster virus, whichcauses chickenpox in children and often debilitat-ing herpes zoster in adults. Each of these programsrequired novel, large, complicated studies to ad-dress key efficacy and safety questions. Joe’s in-novative spirit, keen understanding of clinical re-search and the underlying disease, deep fount ofstatistics knowledge, and can-do attitude was in-strumental in the success of these studies and thesubsequent availability of these vaccines. In short,his work was instrumental to the availability of vac-cines that have saved hundreds of thousands oflives and prevented untold misery.”
In 2009, Joe transitioned to the head of the EarlyDevelopment Statistics group at Merck, and mostrecently, he was in his fifth year as a Scientific As-sociate Vice President and Head of the BiostatisticsMethodology Research group which provides sta-tistical support to research design, analysis and re-porting relating to product development throughthe application of existing and innovative statisticalmethods.
Throughout his career, Joe published books andpeer-reviewed journal papers extensively. JohnTukey, one of the most distinguished individuals inthe field of statistics, and one of Joe’s earliest men-tors, once noted that Joe was one of the most cre-ative statisticians he ever met.
Joe received his M.S. in statistics from Vil-lanova University in 1975 and an MBA in Economicsfrom Temple University in 1979. He was selected
72 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 Invited Articles
to participate in the Merck Research LaboratoriesDoctoral Program in 1984 and received his Ph.D.in statistics from Temple University the followingyear. He was an elected Fellow of the American Sta-tistical Association (ASA) in 1997 and of the Amer-ican Association for the Advancement of Science(AAAS) in 2006. Joe was the founding editor ofStatistics in Biopharmaceutical Research and edi-tor of Statistical Methods in Medical Research. Heserved as Vice President and President of Philadel-phia Chapter in 1988-1989 and BiopharmaceuticalSection Program Chair of ASA in 1994-1995. Joealso served on doctoral dissertation advisory com-mittees at Temple University and the University ofMaryland Baltimore County (UMBC). ”I still re-member his encouraging words and strong supportto junior researchers ... he always encourages me topursue my dream, don’t wait. He is such an inspir-ing leader to work with and a role model for us tofollow.” written by Yi Huang, a professor of statis-tics at UMBC.
In his 42 years at Merck, Joe had an extraordi-nary impact on human health. ”Joe not only di-rectly contributed to the successful developmentof dozens of pharmaceutical products, but he per-sonally and proudly developed, mentored, coachedand collaborated with hundreds of statisticians whohave, in turn, become successful drug and vaccinedevelopers and leaders in their profession both atMerck and in the external statistics community,”said Lisa Lupinacci, Associate Vice President ofLate Development Statistics at Merck. ”Joe was in-deed the catalyst for my career at Merck [and] hisencouragement and counsel for so many years havedirectly contributed to my professional and per-sonal achievements. I will certainly miss Joe’s gen-
erosity, coaching, inspirational stories, candor, in-tegrity, wisdom and humor.” said Amy Gillespie,Associate Vice President for Statistical Program-ming at Merck.
Keaven Anderson, a distinguished scientist atMerck, recalled his interactions with Joe by noting,”His thinking always drew us into a future visionthat we could not imagine on our own! I have somuch to thank him for.”
Joe will be remembered as a humble and kindperson, who was quicker to give credit than acceptit and was most willing to share his time and talentsto help mentor others along the way.
Joe will be dearly missed by his wife Lil, theirdaughters Angelina and Gabby, and his manyfriends and colleagues at Merck and in the biophar-maceutical statistics community around the world.
Jie Chen, Ph.D.Distinguished ScientistBiostatistics and Research DecisionSciencesMerck Research LaboratoriesMerck & Co., Inc.
Lisa Lupinacci, Ph.D.AVP, BiostatisticsLate Development StatisticsMerck Research LaboratoriesMerck & Co., Inc.
© www.icsa.org 73
Yi’s FDA Story: Where Statistics Met Regulation ICSA Bulletin July 2019 Vol.31/2
1988Yi Tsong
*This article reflects the review of the author and shouldnot be construed to represent FDA’s views or policies.
Settling in Northern Virginia:As I mentioned in my 1987 report, when I joined
FDA, I moved to Rockville, Maryland, alone inMarch. By the summer of 1987, my family moved tojoin me. The four months between March and Julywas the time window for me to get a basic idea whatthe housing market looked like. I checked the hous-ing in Rockville, Maryland, and Reston/Herndonin Virginia. They are about the same distance fromInterstate 495 beltway. Meanwhile, my wife Patri-cia was approved to transfer as an employee to Boe-ing office in Reston, Virginia. Boeing rented a ho-tel room at Wolftrap hotel in Vienna, Virginia. Af-ter selling the house in Houston, my parents-in-lawtook Stephanie, then four years old, to visit Pat’s sis-ter living in Detroit. Pat brought daughter Jennifer,then eight years old, to Virginia to join me. Whenthey arrived in Dulles Airport, my housemate, Dr.Kao-Tai Tsai was kind enough to join me to pick upand escort them to the hotel. It was the first time wedrove on this part of the Virginia in dark night. Weprepared a map and decided to take toll road exitfrom Wiehle Avenue south to Sunrise Valley road.Driving east it would be Hunter Mill Road leadingto south. It connected to Lawyers Road and MapleAvenue which was the main street of Vienna, Vir-ginia. But little did we know that part of the HunterMill-Lawyers Road route was so hilly like a narrowroller coaster ride without lighting. All four of uswere driving fearfullyfor the 10 minutes distancebefore reaching the hotel. To ease myself and myfamily, I sang the Bobby McFerrin’s 1988 popularsong“don’t worry, be happy”song with words like,“Don’t worry, be happy
In every life we have some troubleBut when you worry you make it doubleDon’t worry, be happy,…”https://www.youtube.com/watch?v=d-diB65scQU
In later years, I often drove on the same roadswith much more comfort. But I did blow a tire onceby the sharp edge of the road.
A few days after checking in at the WolfTrapHotel in Vienna, we moved to an apartment in Re-ston for six months. It was time for Stephanie tocome home to join us. I still remember the sceneof Stephanie coming out of the tunnel bridge from
the airplane and seeing her sister Jennifer at the en-trance. She was so happy, opened her arms, gig-gling loud while running to meet Jennifer. Jen-nifer was equally happy to run forward to catch her.Without a camera or photo, I had the scene carvedin my mind of the love between the two sisters.
By March 1988, the home in Herndon Virginiawas built. We finally moved into the home. It ishard to believe that Pat and I raised the daughtersand lived there for more than 32 years until both ofus retired.
History of Reston (taken from Wikipedia):Reston is one of the leading ”New Town”
planned communities in the United States.Founded in 1964, Reston was influenced by the Gar-den City movement that emphasized planned, self-contained communities that intermingled greenspace, residential neighborhoods, and commer-cial development. The intent of Reston’s founder,Robert E. Simon, was to build a town that wouldrevolutionize post–World War II concepts of landuse and residential/corporate development in sub-urban America. Reston was frequently (2012, 2014,2016, 2017 and 2018) ranked in the list of “BestPlaces to Live in America”by Money magazine forits expanses of parks, lakes, golf courses, and bri-dle paths as well as the numerous shopping anddining opportunities in Reston Town Center. Be-ginning in 2017, however, high-density commer-cial and residential developments along the DullesToll Road began to spark concerns among residentsabout local government’s ability to ensure that keyinfrastructure, including roads, schools, and parks,would remain in sync with the accelerating paceof new construction. In the early days of ColonialAmerica, the land on which Reston sits was part ofthe Northern Neck Proprietary, a vast grant by KingCharles II to Lord Thomas Fairfax that extendedfrom the Potomac River to the Rappahannock. Theproperty remained in the Fairfax family until theysold it in 1852. Carl A. Wiehle and William Dunnbought 6,449 acres in northern Fairfax County alongthe Washington & Old Dominion (W&OD) Railroadline in 1886, later dividing the land between them,with Wiehle retaining the acreage north of the rail-road line. Wiehle envisioned founding a town onthe property, including a hotel, parks, and commu-nity center, but completed only a handful of homesbefore his death in 1901.
The apartment we lived in Reston was locatedclose to Lake Ann, one of the two lakes received wa-
74 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 Yi’s FDA Story: Where Statistics Met Regulation
ter from Potomac River. Though the lake is small,it has a very peaceful scenery. I drove on the roadover the lake to Dulles toll road to work every morn-ing. When I drove over the lake I watched the watervaporizing on the surface of Lake Ann, it gave me asense of calm and happiness feeling.
1988 Summer Olympic Games:By the summer, the 1988 Summer Olympics
were held in Seoul, South Korea. This was the lastOlympia Games for Soviet Union and East Germanybefore they ceased to exist. This is the year Flo-rence Griffith Joyner set the Olympic record in the100-metre and 200-metre dashes and received bothgold medals. Furthermore, she added a gold in the4x100 relay and a silver in the 4x400 relay. Afterlosing consecutively the gold medals in years, 1988Olympic is the last year that the US Olympic basket-ball team was represented by only non-professionalplayers. This is the Olympic Game that SeoulOlympic Committee produced and distributed anofficial song to publicize the Games. The song“Hand in Hand”was written by Italian composer
Giorgio Moroder and American songwriter TomWhitlock. It was with lyrics like“…Hand in hand we stand
All across the landWe can make this worldA better place in which to live,…”https://www.youtube.com/watch?v=h3tRATI99ik
However, USA Team also has its own Olympicsong sang by Whitney Houston, called “One mo-ment in time”with lyrics like“Give me one moment in time
When I’m more than I thought I could beWhen all of my dreams are a heartbeat awayAnd the answers are all up to meGive me one moment in timeWhen I’m racing with destinyThen in that one moment of timeI will feelI will feel eternity”https://www.youtube.com/watch?v=96aAx0kxVSA“One moment in time”became Houston’s third
song reached UK number-1 hit and second number-1 song in Germany hits. It reached number 5 of theUS Billboard chart. It further became her all-timebest performance at the 1989 Grammy ceremony.
Food and Drug Administration Act and Spon-taneous Reporting System:
A sponsor seeking to market a new drug prod-uct is required by law to have an application (NewDrug Application) reviewed and approved by FDA.
The principal requirement of an NDA is that thesafety and efficacy of the drug is supported by datacollected through phase I to III trials. Often the clin-ical trial comprises most patients with age lowerthan sixty and no complication of other disease.The trial is typically designed with a short studyduration for adverse event since many adverse re-actions has a incidence rate of lower than 0.1% withlong exposure. Once the drug is marketed, thesame drug product is available to a patient pop-ulation that is broader and not well studied. Thedrug may be used for indications of different dis-ease states; for longer duration; patients, such aselderly, children and women not studied in pre-marketing studies; in patients exposed also to othertreatments; used off-label (use for treatment notstated in the label), etc. Postmarket adverse drugevent (ADE) surveillance is designed for exactly thepurpose of identifying the low frequency reactionsthat are not caught in pre-marketing; the high-riskgroup; the drug-drug interaction; the long-termeffect of the drug; increased severity and/or fre-quency of known reactions.
FDA started the Spontaneous Reporting System(SRS) in 1968 for collecting, processing and ana-lyzing adverse drug event (ADE) reports. Most ofthem were reported on the ADE reporting formdesigned by FDA (Form 3500 which was later re-placed by the MedWatch Form in 1992) and submit-ted by manufacturers of prescription drugs as a con-dition for marketing in the United States, while oth-ers are submitted voluntarily by concerned healthprofessionals and consumers. Each report contains:patient’s demographic information including ageand gender; adverse event information, includingdate and outcome of event; description of event,evidence, and existing medical history; informa-tion of suspect and concomitant medicine(s), in-cluding drug name, dose and frequency, diagnosisfor use, whether the event abated after drug use wasstopped, and whether the event reappeared afterthe drug was reintroduced; and the reporter’s nameand address. Each report was reviewed by a trainedprofessional in the FDA to assure that the form iscomplete and the information is correct before itwas entered into the computer database. Together,these reports in the FDA database make up theSpontaneous Reporting System (SRS) of the UnitedStates. In 1997, FDA overhauled the postmarket ad-verse reaction reporting system in order to accom-modate electronic submission and unify the cod-ing with MedDRA (a dictionary of adverse drug re-actions) and renamed the system AERS (AdverseEvent Reporting System). In comparison to SRS,
© www.icsa.org 75
Yi’s FDA Story: Where Statistics Met Regulation ICSA Bulletin July 2019 Vol.31/2
the post-1996 AERS is more structured, with morecapacity, and uses more detailed coding terminolo-gies. The FDA now receives over 300,000 adversedrug event reports annually, and currently has over1.6 million reports in its computerized database. Itforms a major database for postmarketing adversedrug event monitoring and screening.
However, the information in the AERS data baseis rich but cannot be used for computation of inci-dence rates of specific adverse event and drug com-binations. The difficulties result from the followingfour reasons:
1. The method of ascertainment of an adverseevent is not uniform among the reporters andnon-reporters, and the reporters have theirown criteria for identifying an AE. It is gen-erally believed that the number of reportsis much smaller than the actual number ofevents.
2. Although all the reports are reviewed care-fully in the FDA, the quality of the informa-tion collected for a particular adverse eventvaries with the reporter.
3. The population at risk is unknown, hence thenumber of reports is not necessarily closely re-lated to the number of individuals taking thedrug.
4. Although there are courses offered in medi-cal schools and other medical institutes for allthe students, it is, however, not mandatory toreport ADEs by the US medical laws. Thisis different from ADR reporting systems incountries with social welfare systems whichrequires reporting by every health care han-dler in the government system.
In spite of these limitations, AERS is a valuabledatabase to be used for signaling the possibility ofexcessive or outlying ADE and for generating hy-potheses.
Epidemiological and statistical applicationswere often used in the signaling process. The drug-adverse event (ADE) signaling procedures used inUnited States have evolved in the last three decades.
As part of my work on postmarketing safety as-sessment, I was responsible for development andmodified statistical methods for using the SRS(AERS) and other database such as the Medicaiddatabase and third party reporting systems (includ-ing most of the insurance company systems). De-tails of the many developments will be documentedin a journal of the later years.
1988 AIDS ACT-UP demonstration:1986: One by one, we strolled into the con-
ference room of The University of Texas MedicalBranch and took our seats. Finally, Professor TomBranowski, the professor in Family Medicine andthe principal investigator of this research team ofFamily Health Project walked in with his assis-tant Ms. Henske. Professor Branowski looked sosolemn and stood by the table without sitting downand finally said with a trembling voice“Paul won’tbe able to come. He checked in the hospital after di-agnosed for infection of HIV virus.”I sucked in mybreath and looked around and saw five faces dis-played with shock. This was an August morningof 1986. Paul Hook was a professor of sociologyand the co-principal investigator of this researchproject. The objective of Family Health Project istwo-fold. The first is to investigate the health statusand awareness of the black families in Galveston Is-land and southern Texas. Typically, those familieswere in the ghetto areas. Paul was a tall black andattractive professor. His responsibility was to visitand interview the families and collect all the datathrough interview and questionnaire survey Thosewere the years when HIV meant the same as AIDSand unavoidable death.
Six months later, I left Houston and moved toRockville, Maryland and joined FDA. Two moremonths later, I received a package of the reprints ofthe papers published and a letter informed me thatPaul was gone.
1988: is the year when the incidence of AIDSreached its high point. In the summer, FDA in-creased the size of the AIDS drug review team try-ing to speed up the AIDS drug review process. OnOctober 11, those of us working at the FDA Park-lawn building in Rockville were notified that therewould be a demonstration of AIDS patients andfriends. We were strongly advised to avoid any con-frontation with the demonstrators. We should becautious if yelled at when entering and leaving theoffice. Since it was likely to have obstacles block-ing the entrance of the building, we were allowed toturn back home instead to forcing our way in. Therewere actually about a few hundred demonstratorswho showed up in an orderly demonstration in themorning about 8:00 o’clock. We learned later onthat the number increased to more than 2,000. Theprotesting about the concerns about the length ofthe drug approval process was brought to the cen-ter in the AIDS epidemic. In the mid- and late 1980s,ACT-UP and other HIV activist organizations ac-cused the FDA of unnecessarily delaying the ap-proval of medications to fight HIV and opportunis-
76 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 Yi’s FDA Story: Where Statistics Met Regulation
tic infections. ”Shame! Shame! Shame!” and ”Nomore deaths!” were yelled by the protesters. Thisprotest was led by John Thomas, head of the AIDSResource Center in Dallas. He claimed ”I’m herebecause the FDA is holding up drugs that are avail-able, and because people are dying.” John Thomasspoke through a loudspeaker as police motorcyclescircled the building and crowd behind him.
”I am here today – we are all here today – be-cause we all have AIDS,” Thomas said. ”Some ofus have AIDS in our bloodstream. And some of ushave AIDS in our minds. We look into the mirrorand see a sore that won’t go away, and we are fear-ful that we are going to be diagnosed. And we allhave AIDS in our hearts,” he said. ”All of us havelost people we love.”
This protest resulted in 176 arrests.In August 1990, Dr. Louis Lasagna, then chair-
man of a presidential advisory panel on drug ap-proval, estimated that thousands of lives were losteach year due to delays in approval and market-ing of drugs for cancer and AIDS. Partly in re-sponse to these criticisms, the National Institute ofHealth and FDA issued in 1992, new rules to expe-dite approval of drugs for life-threatening diseases,and expanded pre-approval access to drugs for pa-tients with limited treatment options. The first ofthese new rules was the ”IND exemption” or ”treat-ment IND” rule, which allowed expanded access toa drug undergoing phase II or III trials (or in ex-traordinary cases even earlier) if it potentially rep-resented a safer or better alternative to treatmentscurrently available for terminal or serious illness.A second new rule, the ”Parallel Track Policy”, al-lowed a drug company to set up a mechanism for
access to a new potentially lifesaving drug by pa-tients who for various reasons would be unable toparticipate in ongoing clinical trials. The ”paralleltrack” designation could be made at the time of INDsubmission. The accelerated approval rules werefurther expanded and codified in 1992.
All of the initial drugs approved for the treat-ment of HIV/AIDS in the early years were ap-proved through accelerated approval mechanisms.For example, a ”treatment IND” was issued for thefirst HIV drug, AZT, in 1985, and approval wasgranted just two years later in 1987. Three of thefirst five drugs targeting HIV were approved in theUnited States before they were approved in anyother country.
John Thomas attended as a consumer repre-sentative in every advisory committee meeting ofHIV/AIDS new drugs. He made presentations andin tears asked the committee to approve the newdrug because of its potential usage. In 1997, I sawhim the last time at the advisory committee meet-ing. He made a presentation in tears and cried“Ipleaded at every AIDs AC meeting for the approvalof new treatments for HIV. But all I saw in the lastten years, the approval process made all sponsorsmuch richer but none helped my friends or me.”
John dies of AIDS in 1999.
Yi Tsong, Ph.D.Division DirectorCDER/OTS/OB/DBVIU.S. Food and Drug Administra-tion
© www.icsa.org 77
Yi’s FDA Story: Where Statistics Met Regulation ICSA Bulletin July 2019 Vol.31/2
1989Yi Tsong
*This article reflects the review of the author and shouldnot be construed to represent FDA’s views or policies.
The Accutane Case:Each federal agency may recruit professional ex-
perts to serve as a special government employee(SGE) who may be consulted on special issues, pro-vides service on some special projects or serves onan advisory committee. When FDA has any is-sues regarding the drug efficacy, safety or manufac-turing or quality control, members of the advisorycommittee may be invited to participate in a publicmeeting to listen and discuss the issues. The discus-sion and recommendation of the committee will beused to help the agency to determine what action totake. My first experience of an advisory committeemeeting was about the safety of a drug called Accu-tane (Tretinoin) for which I was involved with post-marketing safety assessment using the AERS (Ad-verse Event Reporting System) database.
Accutane is a medication used for the treatmentof acne and acute promyelocytic leukemia. Foracne it is applied to the skin as a cream or oint-ment. For leukemia it is taken by mouth for up tothree months. Its common side effects when usedby mouth include shortness of breath, headache,numbness, depression, skin dryness, itchiness, hairloss, vomiting, muscle pains, and vision changes.Other severe side effects include high white bloodcell counts and blood clots. When used as a creamside effects include skin redness, peeling, and sunsensitivity. Use during pregnancy is known toharm the baby. Accutane was patented in 1957 andapproved for medical use in 1962. The compoundwas first studied in the 1960s at Roche Laborato-ries in Switzerland by Werner Bollag as a treatmentfor skin cancer. Experiments completed in 1971showed that the compound was likely to be inef-fective for cancer and, surprisingly, that it could beuseful to treat acne. However, they also showedthat the compound was likely to cause birth defects.
So in light of the events around thalidomide (ofwhich I will report in in a later issue of this jour-nal), Roche abandoned the product. In 1975, GaryPeck and Frank Yoder independently rediscoveredthe drug’s use as a treatment of cystic acne (the mostsevere form of acne) while studying it as a treatmentfor lamellar ichthyosis, and published that work.Roche resumed work on the drug. In clinical trials,subjects were carefully screened to avoid includ-
ing women who were or might become pregnant.Roche’s New Drug Application for isotretinoin forthe treatment of acne included data showing thatthe drug caused birth defects in rabbits. The FDAapproved the application in 1982.
Scientists involved in the clinical trials pub-lished articles warning of birth defects at the sametime the drug was launched in the US, but nonethe-less Accutane was taken up quickly and widely,both among dermatologists and general practition-ers. Cases of birth defects showed up in the firstyear, leading the FDA to begin publishing case re-ports and to Roche sending warning letters to doc-tors and placing warning stickers on drug bottles,and including stronger warnings on the label. Law-suits against Roche started to be filed. In 1983the FDA’s advisory committee was convened andrecommended stronger measures, which the FDAtook and were at that time unprecedented: warn-ing blood banks not to accept blood from peopletaking the drug and adding a warning to the la-bel advising women to start taking contraceptivesa month before starting the drug. However, use ofthe drug continued to grow, as did the number ofbabies born with birth defects. In 1985 the label wasupdated to include a boxed warning (see below).
In early 1988 the FDA called for a second advi-
78 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 Yi’s FDA Story: Where Statistics Met Regulation
sory committee. An internal FDA memo was leakedto the New York Times a few days before the meet-ing stating that around 1,000 babies had been bornwith birth defects due Accutane, that up to around1,000 miscarriages had been caused, and that be-tween 5,000 and 7,000 women had had abortionsdue to Accutane. This led to a storm of media at-tention. In the committee meeting, dermatologistsand Roche each argued to keep the drug on the mar-ket but to increase education efforts; pediatricians,FDA safety reviewers and the Centers for DiseaseControl (CDC) argued to withdraw the drug fromthe market. The committee recommended to re-strict physicians who could prescribe the drug andto require a second opinion before it could be pre-scribed. The FDA, believing it did not have au-thority under the law to restrict who had the rightto prescribe the drug, kept the drug on the mar-ket but took further unprecedented measures: it re-quired to Roche to make warnings yet more visibleand graphic, provide doctors with informed con-sent forms to be used when prescribing the drug,and to conduct follow up studies to test whetherthe measures were reducing exposure of pregnantwomen to the drug. Roche implemented those mea-sures and offered to pay for contraception counsel-ing and pregnancy testing for women prescribedthe drug - the program was called the ”PregnancyPrevention Program”.
A CDC report published in 2000 showed prob-lems with the Pregnancy Prevention Program andshowed that the increase in prescriptions was fromoff-label use, and prompted Roche to revamp itsprogram, renaming it the ”Targeted Pregnancy Pre-vention Program” and adding label changes likerequirements for two pregnancy tests, two kindsof contraception, and for doctors to provide phar-macists with prescriptions directly; providing ad-ditional educational materials, and providing freepregnancy tests. The FDA had another advisorymeeting in late 2000 that again debated how to pre-vent pregnant women from being exposed to thedrug; dermatologists testified about the remark-able efficacy of the drug, the psychological impactof acne, and demanded autonomy to prescribe thedrug; others argued that the drug be withdrawn ormuch stricter measures be taken.
In 2001 the FDA announced a new regulatoryscheme called SMART (System to Manage AccutaneRelated Teratogenicity) that required Roche to pro-vide defined training materials to doctors, and fordoctors to sign and return a letter to Roche acknowl-edging that they had reviewed the training materi-als, for Roche to then send stickers to doctors which
they would have to place on prescriptions they givepeople after they have confirmed a negative preg-nancy test; prescriptions could only be written for30 days and could not be renewed, thus requiring anew pregnancy test for each prescription.
As reported in Wikipedia that in February 2002,Roche’s patents for Accutane expired, and there arenow many other companies selling cheaper genericversions of the drug. On June 29, 2009, Roche Phar-maceuticals, the original creator and distributor ofAccutane, officially discontinued both the manu-facture and distribution of their Accutane brandin the United States due to what the company de-scribed as business reasons related to low marketshare (below 5%), coupled with the high cost of de-fending personal-injury lawsuits brought by somepeople who took the drug. Generic versions ofAccutane remained available in the United Statesthrough various manufacturers however. RocheUSA continues to defend Accutane and claims tohave treated over 13 million people since its intro-duction in 1982.
It was also reported that several trials over in-flammatory bowel disease claims have been heldin the United States thus far, with many of themresulting in multimillion-dollar judgments againstthe makers of Accutane or the its generics. F.Hoffmann-La Roche Ltd. continues to manufac-ture and distribute Accutane under the brand nameRoAccutane outside of the United States
In 1976, World Health Organization reportedmore than 600 cases of Ebola found in Congo andSudan. It raised major concerns in health organi-zations worldwide. In 1989, there was an outbreakof Ebola-related cases found in Northern Virginianear where I live. Even though it is not FDA workrelated, I found it is interesting to also report as fol-low.
HOT ZONE:1943 Isaac Newton Square, Reston, VA was
about 3 miles from the apartment we lived in 1987and less than 7 miles from my current home inHerndon VA. A company called Hazelton Researchoperated a quarantine center there for monkeys thatwere destined for laboratories. In October 1989,when an unusually high number of their mon-keys began to die, their veterinarian decided tosend some samples to Fort Detrick (USAMRIID)for study. At the time, it was believed that thevirus was Simian hemorrhagic fever virus, a viralhemorrhagic fever harmless to humans but almostalways fatal to other primates. Early during thetesting process in biosafety level 3, when one ofthe flasks appeared to be contaminated with harm-
© www.icsa.org 79
Yi’s FDA Story: Where Statistics Met Regulation ICSA Bulletin July 2019 Vol.31/2
less pseudomonas bacterium, two USAMRIID sci-entists exposed themselves to the virus by waftingthe flask. When they eventually tested the sam-ples with known Level 4 agents, only EBOV reactedwith the unknown samples. They decided not totell anyone about their exposure, but they did se-cretly tested their blood every day. After one ofthe monkey house staff members became ill withnausea and violent vomiting, USAMRIID was givenpermission to send in a team to euthanize all themonkeys at the facility and collect tissue samples.They later determined that, while the virus is terri-fyingly lethal to monkeys, humans can be infectedwith it without any health effects at all. This virusis now known as Reston virus (RESTV), a relativeof Ebola. The virus found was a mutated form ofthe original Ebola virus, and was initially mistakenfor Simian Hemorrhagic Fever (SHV). The originalReston facility located at 1946 Isaac Newton Squarewas torn down sometime between 1995 and 1998.
Richard Preston investigated the event andwrote a book called“The Hot Zone”. Due to thedetailed and graphic descriptions of the effects ofexotic tropical diseases, as well as the revelationthat an ebola virus was found a few miles awayfrom Washington D.C., The Hot Zone was hailedby many as a chilling and accurate story of lethalviruses and their encounters with humans. BecausePreston’s writing style is that of a ”science fact”thriller, some critics accused Preston of dramatiz-ing and exaggerating the effects of an Ebola virusinfection and embellishing facts with his own imag-ination. Since its publication over a decade ago,however, The Hot Zone is generally regarded as anonfiction work and acknowledged for its masterfuldramatization. In his blurb, horror writer StephenKing called the book, ”one of the most horrifyingthings I’ve ever read.”
In 1993, 20th Century Fox and producer LyndaObst purchased the film rights from the book au-thor. In response to being outbid, Warner Bros.immediately began working on a similarly themedfilm entitled“Outbreak”. This competing film ledto the collapse of“Crisis in the Hot Zone”. With dif-ficulty to get confirmation of the leading actors andactresses, Fox dropped the project in 1994. Accord-ing to Wikipedia, in 2014, Lynda Obst and DirectorRidley Scott again planned to adapt the book into aTV series that started to film in September of 2018.
Yi Tsong, Ph.D.Division DirectorCDER/OTS/OB/DBVIU.S. Food and Drug Administra-tion
80 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 Hints from Hans
Impact and CitationsHans Rudolf Künsch
Research can be a frustrating and lonely enterprisewhen all your attempts to prove a conjecture leadnowhere, or when you receive reports by refereeswho didn’t understand the new idea of your pa-per, or when nobody among your colleagues showsany interest in what you are trying to do. So whydo you not give up? One reason is extrinsic: Suc-cessful research is an essential condition to obtain aposition, a promotion or a salary rise. But there isalso the pure intrinsic motivation, ”the pleasure offinding things out” as the title of a book by RichardFeynman says (Perseus, 1999). A third reason is pre-sumably that we would like to make an impact onscience and – more ambitiously – on society. Thisthird motivation interacts with the two others: Suc-cess of research is usually judged by its impact, andthe pleasure of finding things out is much higher ifwhat you have found is also relevant for others.
The number of citations of a researcher is an in-dicator for her or his impact that is better than thenumber of publications. The main reason for us-ing citations in a scientific work is to distinguish be-tween your own work and that of others, therebyrespecting intellectual property. If you cite some-one else’s work, you acknowledge that this work hashad an impact on your research, because it has beenuseful or even essential for you to arrive at your ownresult or conclusion. However, the number of ci-tations as an impact measure has its shortcomingstoo. For instance, the cited work is often not the ori-gin of the idea or result that was used, but it is moreeasily accessible or understandable than the origi-nal source. This means that someone with highlyoriginal ideas that are difficult to understand maynot get as many citations as he or she deserves. Thisis similar to ”Stigler’s Law of eponomy” (due RobertK. Merton and maybe others) which states that noscientific discovery is named after the person whodiscovered it. Often a citation also serves to con-vince the reader of the importance of the topic con-sidered. In that case, the cited author is typicallysomeone well known with a high reputation, butnot necessarily the one who has made the most es-sential contribution. In other cases the cited workcontains a somewhat different approach to the sameproblem and the reason for citing it is that it al-lows the author to show the superiority of the newmethod. Finally one should not forget that some au-thors find ways to manipulate citation statistics to
their advantage, e.g. by citing earlier work of theirown or of their network even though it has little rel-evance for the paper on hand, or by requiring cita-tions of their own work when refereeing someoneelse’s manuscript.
In order to really understand what a citationmeans one has to look at the paper where this cita-tion has been made. The quality of research of a sci-entist or a whole department should therefore notbe evaluated on the basis of citation statistics alone.Reducing citation statistics to a single index is evenworse. A serious evaluation of someone’s researchrequires that one looks at some of his or her papers.I think it is therefore a good idea to list your five bestor most important papers (which need not be yourmost cited ones) on your cv.
For the same reason, you should not base yourintrinisic motivation for research too much on yourown citation statistics. Progress in research is notonly achieved by singular contributions of a fewgeniuses, but also by a complex web of interac-tions between many different contributions. Thishas been noticed already in the 18th century byGeorg Christoph Lichtenberg (1742-1799), a Ger-man physicist and satirist. Today he is best knownfor his aphorisms, concise and witty philosophicalstatements. He wrote once ”If I hadn’t written thisbook, then in 1000 years from today between 6 and7 pm people in many German towns would speakabout entirely different things than those they willactually be speaking about.” Surely, he didn’t ex-pect that people in a 1000 years in Germany wouldstill be reading and discussing this one book, butrather that everything that is published continuesto have some unpredictable indirect effect in the fu-ture, even if it is completely forgotten in a shorttime – a bit like the famous butterfly in Brazil of EdLorenz that sets off a tornado in the US by flappinghis wings.
Moreover, you can also make an impact by otherthings than your publications. The guidance andfeedback you give to your students, the questionsyou ask after a seminar talk, the suggestions yougive in a referee report all can have a positive ef-fect on the scientific progress. Obviously, by thesame means you can also have a negative impactby missing to see the value of an important, butbadly described idea. To give an example of a pos-itive impact I received in my own career, my mostcited paper that introduced the block bootstrap wastriggered by a question from Colin Mallows at Bell
© www.icsa.org 81
Terence’s Stuff ICSA Bulletin July 2019 Vol.31/2
Labs. I had told him about my earlier work ondefining an influence function for time series data.He then remarked that the influence function hasa close relation to the jackknife and that I thereforeshould look at the jackknife for time series.
I hope that you experience in your career notonly the pleasure of finding things out, but also thepleasure of sharing your ideas with others. It willhave a positive impact even though it is not alwaysmeasurable.
A more detailed discussion about the use andmisuse of citation statistics can be found in the pa-
per by R. Adler, J. Ewing and P. Taylor, StatisticalScience 24 (2009), 1-14.
Hans Rudolf Künsch, Ph.D.Professor Emeritus of MathematicsSeminar für StatistikETH ZürichSwitzerland
Give Industry a ChanceTerry Speed
Editorial: This is a reprint from a column article pub-lished in the IMS Bulletin (Volume 43, Issue 2: March2014; http://bulletin.imstat.org/wp-content/uploads/Bulletin43_2.pdf) with IMS’permission.
What do people promoting data science and bigdata want that we statisticians do not seem to have?Why do so few PhD candidates build their thesesaround a specific application? Why do so many ofour PhD grads want to be professors? Why don’tmore PhD students in statistics do internships withindustry over their summers? Nothing is simplewith questions like these, so my initial answers tothese questions are bound to be simplistic, proba-bly wrong-headed, and definitely tendentious, butlet me give them anyway.
One thing I notice about discussions of data sci-ence and big data is that they are invariably in thecontext of specific application areas. Global change,brain signals, earthquake signals, supernovae, so-cial unrest, traffic accidents and smart thermostatsare all named on the website of the Berkeley Insti-tute of Data Sciences, part of a three campus $38Minitiative launched by the White House in Novem-ber last year, to be housed in the University Libraryat Berkeley. One thing I notice about discussions ofmathematical statistics is that they are rarely in thecontext of a specific application area.
Recently I talked to participants in a Mathemat-ics in Industry Study Group (MISG) and to studentsat an Industry Doctoral Training Centre (IDTC),both in Australia; to postdocs at a US academic insti-
tution who are struggling to maintain—not to men-tion further—their careers there; and to a scientistfrom a small, successful US company keen to em-ploy statisticians or statistics interns. I found somecommon themes in our discussions, including theperennial“applied”-vs-applied, and academia- vs-industry divisions. There was also a general feelingthat these issues do not get as fully discussed as theyshould in academia, where most of us reside.
When writing about factors motivating the firstAustralian Mathematics in Industry Study Group,30 years ago, one of the organizers remarked that hehad attended an applied mathematics conference atwhich just two or three of the 120 delegates werefrom industry, and only about 20 out of 56 talkshad specific applications in mind, i.e. were reallyapplied. In part, the MISG was started to changethis, and it has been very successful. I wonder whatproportion of the statistics talks at a typical IMSconference are applied, in the preceding sense, andwhether we need an initiative to change this? Myfeeling is: possibly.
The Industry Doctoral Training Centre seeks tobring together an industrial problem and an indus-try sponsor with a PhD student and an academicadvisor. This seemed a wonderful program: in-dustry gets problems solved, while the students notonly get something with direct real-world value towork on for their PhD, they develop communica-tion, teamwork and leadership skills, as well as im-mersing themselves in the subject matter of theirproblem. I heard that finding industrial problemsand students wanting to work on them was not ashard as finding suitable academic advisors.
Chatting with people, some of whom were on
82 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 XL-Files
their second or third postdoc, about the difficultyof getting research grants or tenure-track academicjobs, I couldn’t resist asking them how seriouslythey had considered other careers. How wouldthey like a job, I asked them, which didn’t re-quire applying for grants, which permitted them towork reasonable hours, paid well, and was enjoy-able, challenging and fulfilling? Sounds good, wasthe response. There are plenty of such jobs outsideacademia, I told them. They freely conceded theirvalue system had probably been skewed by theirlong sojourn in academia, and by their respect andadmiration for their advisors, and that their beliefthat academic jobs were the best by far was not nec-essarily based on complete information. It’s my im-pression, I told them, that professors are very goodat helping students develop into people like them-selves, often many more than could reasonably havea career like them, even if that was desirable, butless good at pointing to or promoting alternative ca-reers. That, I said was up to them to explore. In abetter world, their advisors would be well informedand be able to discuss a broader range of futureswith them.
What should be done? Stat department chairs
could establish or strengthen existing links withindustry, strongly encourage all grad students totake summer internships, ensure that they all knowabout careers outside academia, arrange adjunctappointments for interested and suitably qualifiedpeople from industry, and encourage their facultyto become involved with industry as well, for ex-ample via sabbaticals or summer internships. Thenwe might hope to see more grad students workingon statistical problems from industry for their PhDresearch, something which would serve the triplepurpose of solving an applied problem of real in-terest and perhaps importance, familiarizing them-selves with a field other than statistics, and broad-ening their career opportunities.
Terry Speed, Ph.D.Professor and Lab HeadBioinformatics DivisionWalter & Eliza Hall Institute ofMedical ResearchParkville, VictoriaAustralia
A Fundamental Link between Statisticsand HumorXiao-Li Meng
Editorial: This is a reprint from a column article pub-lished in the IMS Bulletin (Volume 42, Issue 2: March2013; http://bulletin.imstat.org/wp-content/uploads/Bulletin42_2.pdf) with IMS’permission.
My new year’s reading started with a holidaygift: On the Money, a collection of over 400 car-toons in The New Yorker from 1925–2009. No, aftermonths of learning about fundraising, money wasleast alluring on a day when my alarm clock took arest. But I could use a few laughs, even at my ownexpense or with irony. Indeed, the gift was froman alumna, and I wondered if it was meant to bea friendly reminder: relax, don’t take money (andyour job) too seriously.
However I did end up taking the book very se-riously, reading it word by word. Huh? Reading
cartoons? Of course not. But there is an introduc-tion by Malcolm Gladwell, whose name might notbe as familiar to statisticians as the title of one of hisbest-sellers: Outliers (and perhaps also The TippingPoint and Blink). Gladwell’s philosophical intro-duction also started with irony. As a writer for TheNew Yorker, he considered the book very strange,“because we are a magazine for people for whom
money is a secondary concern. …So what on earthdoes The New Yorker think about when we thinkabout money?”Regardless of your opinions abouthis characterization, his short answer was, “wemake jokes about it.”
Gladwell’s long answer began with an anecdoteof how stunned he was at a corporate retreat, whena CFO used a business-like PowerPoint presentationto tell his life story. It ended with the key point:“People who want the world to conform to the prin-
ciples of business are Realists. People who thinkthe other way around—this is true whether they
© www.icsa.org 83
XL-Files ICSA Bulletin July 2019 Vol.31/2
spend their days parsing sonnets or actuarial tables—are Romantics, and the Romantic position…is thecomic position.”
Gladwell’s main supporting example is a sta-tistical one. Bernard Madoff’s fund had a 96%“winning”percentage, with“annual gains that fell
like clockwork between 10–12 percent.”Why didn’the mimic the volatility of a hedge fund, with amore enticing long-run gain and a much less suspi-cious winning pattern? Gladwell’s answer is that itis “because Madoff understood what consistencymeans in personal terms: it means trustworthi-ness, mastery, competence, safety.”It was preciselythis consistency that convinced Harry Markopolos,Madoff’s bête noire, that Madoff was really a Made-up. As Markopolos argued in his 17-page memo tothe SEC,“No major league baseball hitter bats .960,no NFL team has ever gone 96 wins and only fourlosses over a 100-game span, and you can bet ev-erything you own that no money manager is up 96percent of the months, either.”In other words, whenwe take a Realist’s position, Madoff’s made-up con-sistency points us to exactly the opposite conclusionit had aimed to achieve, that is, he is completely un-trustworthy.
By no means is this example meant to glorify theRealists’ position. As Gladwell emphasized: “Thevictims of Bernie Madoff would have done well tothink of Madoff in business terms, not personalterms. Then again, the traders at AIG, who havecost taxpayers many, many multiples of what Mad-off cost the world, would have done well to importa healthy dose of personal virtue into their profes-sional practices.”
Despite the painful context, Gladwell’s under-lying message inspired me to think about a fun-damental link between statistics and humor. Thediscipline of statistics is essentially about separat-ing commonalities (e.g., patterns, signals) from in-dividualities (e.g., variability, noise). In contrast,the best kind of humor is often the result of judi-ciously mixing commonalities and individualities
to create comic effect.A story told by Rick Cleary (visiting Harvard
from Bentley) at our 2008 holiday party, illustratesthis point well. At the end of the last lecture of an in-troductory course for which he was a teaching assis-tant, the professor (the late George Casella—whoseobituary appears here) encouraged the class to askany remaining questions on anything that had beencovered. A student who had never asked any ques-tion before raised her hand. “Professor Casella, Ireally enjoyed your class, but there is one thing thathas puzzled me for the entire semester. Why arestandard deviations always six?”
You will be laughing now, or in a few seconds, ifyou are a real statistician. Otherwise you would belaughing at how nerdy statisticians must be if theycan find humor in the number six. What makes thisstory greatly humorous to statisticians is the mixingof a well-understood commonality (standard devia-tion is commonly denoted by σ) and an unexpectedindividuality (the student’s mistaking σ for 6). Itwould not be humorous at all if six were replacedby one because George, for whatever reason, de-cided to use the letter l for standard deviation in hiscourse.
What could be a more joyful way to celebratethe International Year of Statistics than by tellingthe world that Statistics is the most enjoyable pro-fession on earth, because along with every depress-ing study or erroneous argument there is an entic-ing recipe for entertaining ourselves?
Xiao-Li Meng, Ph.D.Whipple V. N. Jones Professorof StatisticsDepartment of StatisticsHarvard University
84 © www.icsa.org
ICSA Bulletin July 2019 Vol.31/2 Upcoming Events
Upcoming EventsPlease find below a list of upcoming ICSA meetingsand co-sponsored meetings. This list also appearson the ICSA website. If you have any questions,please contact Dr. Gang Li, the ICSA Executive Di-rector ([email protected]).
ICSA Sponsored Meetings:
The 11th ICSA International Conference
December 20 - December 22, 2019Hangzhou, Zhejiang, ChinaOrganizing committee Chair: Hongzhe Li([email protected])
2020 ICSA Applied Statistics Symposium
Huston, TX, USAMay 17 - May20, 2020Professor Hulin Wu at University of Texas HealthScience Center at Houston ([email protected])will be the Chair of the organizing committee.
ICSA Co-sponsored Meetings:
Conference on Current Trends in SurveyStatistics
August 13 - August 16, 2019National University of Singapore, SingaporeFor more information about the conference, pleasevisit https://ims.nus.edu.sg/orgsites/2019data/.
The 6th International Symposium on Bio-pharmaceutical StatisticsAugust 26 - August 30, 2019Kyoto, JapanFor more information about the conference,please visit https://www2.aeplan.co.jp/isbs2019/index.html.
The 3rd International Conference on Sta-tistical Distributions and ApplicationsOctober 10 - October 12, 2019Grand Rapids, MI, USAFor more information about the conference, pleasevisit http://people.cst.cmich.edu/lee1c/icosda2019.
The 11th International Multiple Compar-isons Procedures (MCP) ConferenceDecember 12 - December 15, 2019TaipeiFor more information about the conference, pleasevisit http://www.mcp-conference.org/hp/2019/.
Chapter Meeting:The 4rd ICSA –Canada Chapter Sym-posium, Advances and Innovations inStatistics and Data ScienceAugust 9- August 11, 2019Queen’s University in Kingston, Ontario, CanadaPlease visit http://www.icsa-canada-chapter.org/symposium2019/home.html for moreupdated information.
© www.icsa.org 85
2019 ICSA Applied Statistics Symposium Student Paper Awardees
2019 ICSA Applied Statistics Symposium Student Volunteers
Photo courtesy of Space Center Houston, Sam Houston Race Park and Dario Failla
2020 ICSA Applied Statistics Symposium May 17-20, The Westin Galleria Houston, Houston, Texas, USA
The 2020 ICSA Applied Statistics Symposium will be held from Sunday, May 17th to Wednesday, May 20th , 2020, at The Westin Galleria Houston, Houston, Texas, USA. This will be the 29th annual symposium for the International Chinese Statistical Association (ICSA). The theme of this conference is Advancing Statistics for Data Intelligence. For more information, please contact [email protected] and visit the conference website https://symposium2020.icsa.org .
Speakers
Key Dates
• Early Bird Registration Deadline: March 15, 2020 • Invited Session Proposal Submission Deadline: November 1, 2019 • Short Course Proposal Submission Deadline: December 15, 2019 • Student Paper Award Application Deadline: February 15, 2020 • Invited Session Abstract Submission Deadline: March 15, 2020 • Contributed Session Abstract Submission Deadline: April 15, 2020 • CV/Resume for Career Services Submission Deadline: May 1, 2020
Houston Houston is the fourth-most populous city in the United States. Notable facilities include the Museum District, the Houston Zoo, and NASA's Lyndon B. Johnson Space Center. Houston is the seat of the internationally renowned Texas Medical Center (TMC), which contains the world's largest concentration of research and healthcare institutions including University of Texas Health Science Center at Houston, MD Anderson Cancer Center, Baylor College of Medicine, Memorial Hermann Hospital, The Methodist Hospital, and Texas Children's Hospital. Several universities of higher education such as The University of St. Thomas, University of Houston, and Rice University are located within the city.
Keynote Speaker Xihong Lin, PhD
Professor, Harvard University
Keynote Speaker
Michael I. Jordan, PhD Professor, University of
California, Berkeley
Keynote Speaker Josh Chen, PhD, Head,
Global Biostatistical Sciences, Sanofi Pasteur
Space Center at Houston Sam Houston Race Park
Banquet Speaker Hong Ogle, MS Bank of America
Houston Market President
Galveston Cruise