chemotherapy in advanced ovarian overview of randomised ...conclusions-in the past, randomised...

Chemotherapy in advanced ovarian cancer: an overview ofrandomised clinical trials

Advanced Ovarian Cancer Trialists Group

AbstractObjectives-To consider the role of platinum and

the relative merits of single agent and combinationchemotherapy in the treatment of advanced ovariancancer.Design-Formal quantitative overview using

updated individual patient data from all availablerandomised trials (published and unpublished).Subjects-8139 patients (6408 deaths) included in

45 different trials.Results-No firm conclusions could be reached.

Nevertheless, the results suggest that in terms ofsurvival immediate platinum based treatment wasbetter than non-platinum regimens (overall relativerisk 0-93; 95% confidence interval 0-83 to 1-05);platinum in combination was better than single agentplatinum when used in the same dose (overallrelative risk 0-85; 0*72 to 1-00); and cisplatin andcarboplatin were equally effective (overall relativerisk 1-05; 0-94 to 1-18).Conclusions-In the past, randomised clinical

trials of chemotherapy in advanced ovarian cancerhave been much too small to detect the degree ofbenefit which this overview suggests is realistic forcurrently available chemotherapeutic regimens.Hence a new trial comparing cisplatin, doxorubicin,and cyclophosphamide (CAP) with carboplatin hasbeen launched and plans to accrue 2000 patients.

IntroductionOvarian carcinoma is the seventh most common

cancer ofwomen in the world. Some 140 000 new casesare diagnosed every year and the disease is responsiblefor the greatest number of deaths from gynaecologicalmalignancy in Europe and North America.' Despiteover 50 randomised clinical trials having examinedthe relative efficacy of different chemotherapeuticregimens in advanced disease (FIGO (InternationalFederation of Gynaecology and Obstetrics) stages IIIand IV), individually these trials have been too small toshow clear benefit of one type of chemotherapy overanother. Nevertheless, many of these trials have had animportant influence on clinical practice, and conse-quently the type and intensity of chemotherapy usedroutinely for patients with advanced disease havefluctuated greatly.

Ovarian cancer was one of the first solid malignanttumours to be treated by chemotherapy, and the singlealkylating agents that were first used over 30 years agowere considered optimal treatment until the mid-1970s. The past 15 years, however, have seen manychanges in disease management. In 1978 a smallrandomised trial in advanced disease found that Hexa-CAF, a combination of cytotoxic drugs (hexamethyl-melamine, cyclophosphamide, methotrexate, andfluorouracil), achieved higher response rates thanthe single alkylating agent melphalan and suggestedthe possibility of a corresponding improvement insurvival.2 At about the same time phase II studiessuggested that cisplatin was the most promising newdrug then available.' These results rapidly led to thestandard use of cisplatin in combination with other

cytotoxic drugs, usually doxorubicin and cyclophos-phamide with or without hexamethylmelamine.45When, however, doubt was cast on the effectiveness

of doxorubicin by both randomised phase III trials6-8and phase II studies in patients not responding tocisplatin9 several major centres adopted cisplatin pluscyclophosphamide as standard. Furthermore, whenother trials failed to find significant survival differencesbetween single agent cisplatin and cisplatin in combi-nation with other drugs'0 some centres reverted tousing single agent platinum as routine first line treat-ment. Recently a large number of trials have comparedcisplatin with its less nephrotoxic and neurotoxicanalogue carboplatin, and although follow up timeswere often short, many institutions have now adoptedcarboplatin as standard.

Currently it is unclear what constitutes optimalchemotherapy for advanced disease and treatmentstrategies vary both nationally and internationally.What is clear is that to date no individual clinical trialhas been large enough to detect survival differences ofthe magnitude that could reasonably be expected withavailable treatment.'2 Consequently the inconclusiveresults of over 50 such trials reported could be con-sistent with moderate treatment benefits.'3 The BritishMedical Research Council Gynaecological CancerWorking Party realised the need to synthesise theinformation from these trials to evaluate currently usedchemotherapeutic regimens. Given the problems asso-ciated with a qualitative review of published work,'3this group initiated an overview which used formalcluantitative methods to combine the results from allAvailable randomised trials examining the role ofplatinum and of combination chemotherapy in thetreatment ofadvanced ovarian cancer. At the outset theMRC overview secretariat contacted the investigatorsresponsible for each trial, inviting their collaboration.In so doing it established the Advanced OvarianCancer Trialists Group, under whose auspices theoverview was conducted.

Methods and dataThe relative merits of single agent and combination

chemotherapy and the role of platinum in diseasemanagement were sought, and five comparisonsbetween different forms of chemotherapy were identi-fied as being of interest. These were: (I) single non-platinum agent versus non-platinum combination; (II)single non-platinum agent versus platinum combina-tion; (III) addition of platinum to a regimen; (IV)single agent platinum versus platinum combination;(V) cisplatin versus carboplatin.

Trials were eligible for inclusion in the overviewif they examined first line treatment for advancedepithelial ovarian carcinoma and made one or more ofthe comparisons listed above. Each had also to beunconfounded and believed to have been randomisedin a manner that precluded prior knowledge of the nexttreatment assignment. To avoid publication bias'4 itwas essential that both published and unpublishedstudies were included, and various methods wereemployed to identify relevant trials. A bibliographic

BMJ VOLUME 303 12 OCTOBER 1991

Advanced Ovarian CancerTrialists GroupMembers of the trialistsgroup and the organisationsand groups that collaboratedin the overview are listed atthe end of this report.

Correspondence to: Dr L AStewart, MRC Cancer TrialsOffice, 1 BrooklandsAvenue, CambridgeCB2 2BB.

BMJ_ 1991;303:884-93

884

on 15 August 2021 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

101

0 8-

06-

0I)0*4-

0 2-L= Combination

Single agent

0 1 2 3 4

Single non-platinum Numbers at riskagent 1379 649 315 185 152

Non-platinum 1767 893 427 258 182combination

5Years

126

138 l

individual trials. Virtually all of this information wasrecovered, so that 99 5% of all patients randomised inthe studies included in the overview were available foranalysis. After collection, processing, and checking ofthe data individual trialists were sent a printout of theirdata set as stored in the overview database togetherwith calculated survival curves so that they could checkfor errors.

All analyses were carried out on an intention to treatbasis, patients being analysed according to theirallocated treatment irrespective of whether theyactually received that treatment. For each comparisonsurvival analyses were stratified across all trials togenerate log rank statistics. The individual observedand expected numbers of deaths as calculated in theseactuarial survival analyses were pooled to provide anoverall relative risk for the relevant comparison. Thisvalue gives the relative risk of death associated withtwo types of treatment. For example, when comparingtreatment A with treatment B a relative risk of 0-8represents a 20% decrease in the risk of death when

6 i 6 lbX1 using treatment A whereas a value of 1-2 represents a20% increase in the relative risk of death when usingtreatment A. A relative risk of unity represents nodifference in the risk of death associated with the two

104 90 76 66 50 treatments. Appendix C explains the graphical dis-116 94 82 70 61 plays used in the figures to present the relative risks for

the various trials.I-platinum combination) All p values are two sided, and the x2 values were

calculated on one degree of freedom unless otherwisespecified.

review by means of MEDLINE, CancerLit, andpublished texts was carried out. This was supple-mented by examining the trial registers produced bythe National Cancer Institute (PDQ ClinProt) and theUnited Kingdom Co-ordinating Committee on CancerResearch.'I The proceedings of relevant clinical meet-ings were also consulted. In addition, questionnaireswere sent to the principal authors of published trialsand to international members of the Royal College ofObstetricians and Gynaecologists asking them tosupplement a provisional list of trials. Pharmaceuticalcompanies concerned in ovarian cancer treatment werealso approached in this way.

Fifty three eligible randomised trials were identi-fied, two other potentially eligible studies having to beexcluded on the ground that they did not seem to beappropriately randomised. At the time of data collec-tion for the overview roughly 30% of these eligibletrials had not been published fully. Information was

available from the 45 studies listed in appendixA6 8 10 11 16-50 (unpublished references (A)-(F)) andunavailable from the eight eligible trials listed inappendix B.25' Information was sought for eachindividual patient randomised in these studies, and the45 available trials included data on 8139 patients (6408deaths) originating from different countries. Thesepatients accounted for 95% of all known patientsentered into randomised clinical trials of chemo-therapy in advanced ovarian carcinoma that met theoverview criteria for eligibility.

In all but two cases the individual patient informa-tion that was supplied for studies was updated for theoverview. Thus the overview used information from anextended period with, for example, a median followup of 10 years in trials comparing non-platinumsingle agents with non-platinum combinations.Incoming data were checked for any obvious flaws or

inconsistencies such as missing values, dates out ofsequence, and apparent differences between the dataset and publication. Problems were rectified by corres-

pondence with the principal investigator. In order toavoid the potential bias of exclusion after randomisa-tion it was necessary to recover data from patients whohad been randomised but excluded from the analysis of

ResultsTable I summarises the numbers of patients and

deaths in comparisons I-V.

TABLE I-Numbers oftrials, deaths, and patients per comparison

Comparison

IIIIIIVV

Total*

No of No ofavailable unavailable

trials trials

16 611 28 06 0

11 0

No ofdeaths

2817113611347121771

No ofpatients

314613291408925

2061

45 8 6408 8139

*Total numbers of trials, deaths, and patients are not simple sum of thosegiven for comparisons I-V. Some trials were included in two comparisonsbut counted only once in totals.

COMPARISON I: SINGLE NON-PLATINUM AGENT VERSUS

NON-PLATINUM COMBINATION

For comparison I data were available from 16 of 22eligible trials, comprising 1379 patients randomised toreceive a single non-platinum agent and 1767 to receivenon-platinum combinations6 16-28 (unpublished refer-ence (A)). The reason for the imbalance in numberswas that some trials used two or more combinationarms. A total of 2817 deaths were observed and themedian duration of follow up was 10 years. Figures 1

and 2 show the results. There was no evidence of anyoverall difference between the two types of treatment(x2=0-6S; p=042), the overall relative risk being 0-98(95% confidence interval 0 91 to 1-05). This equival-ence is apparent in figure 2, which shows a seeminglyrandom scatter of relative risk values, 10 trials favour-ing combination chemotherapy, seven favouring thesingle agent, and two favouring neither. With theexception of a single trial which indicated a significantimprovement in survival with combination chemo-therapy all the trials yielded confidence intervalsstraddling unity, indicating their inconclusive out-come.


FIG 1-Survival curves for comparison I (single non-platinum agent v non

885



ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

No of events I No enteredRelative risk

Reference

Aabo et al 16

Bolis et all6Brodovsky et al7Bruckner et al'8Chylak et al 9

Edmonson et al 20

Miller et al 21MRC Working party on Ovarian Cancer 22MRC Gynaecological Cancer Working JTrial a

Party (unpublished ref (A)) LTrial bOmura et al (trial a)23

Trial a

Park et al 24 Trial bTrial c

Sturgeon et al 25Trope et al 26Turbow et al27

Wharton etal28Trial aTrial b

Combination Single agent

87/9732 / 37188 / 199233 / 24533 /3452 / 55116 / 125148 / 1736/840 /49199 / 229182 / 20454/6310 / 2140 / 4076 / 8124 / 2428/3341/50

73 / 8233 /37199/ 210

78 / 8632 /3552 /56118 /129145 /1718 / 1043/4995 / 11065 / 7127/419 / 1839 / 4383 / 8720 / 2488 / 9521/25

O-E Variance

-3.39-0.166.604.523.13-2.250.62

-4.72-1.633.281.57-6 767.10-0.436.05

-19.952.10

-33.41-2 57

39.5216.2596.5459.5316.2225.9858.4973.193.48

20.6264.5846.2118.964.7219.6339.2710.9722.0813..

Total -10.30 649.

Test for heterogeneity:. 2183 = 20.46; p= 0.308

0

43 -

69

I I I I l Il I I0-0 0 5 1.0 1-5 2 0

Combination better Single agent better

FIG 2 -Relative risks for comparison I. (O- E=Observed minus expected. Negative value indicates that treatment group fared better thancontrols. Trials making more than one comparison of interest (that is, with multiple treatment arms) are labelled a, b, etc. Open squares representrelative risks in trials. Sizes ofsymbols are directly proportional to amount of information in trials. Bars are 99% confidence intervals. Occludedlozenge represents pooled total relative risk and 95% confidence interval. See appendix Cforfull explanation ofgraphical displays ofrelative risk)

No ofpatients in

No of trials trials usingusing dose dose

1 801 116 9401 1001 891 109

1 0O

0-8-

0-6-

0-4-

0-2

COMPARISON II: SINGLE NON-PLATINUM AGENT VERSUSPLATINUM COMBINATION

Comparison II contained 13 eligible trials2529-365859(unpublished references (A), (B)), of which data wereavailable from 11. In these trials 659 patients wererandomised to receive a single non-platinum drug and670 to receive a platinum combination, although manyof the patients allocated to the single agent receivedplatinum on relapse. The median follow up period was6 5 years and the total number of observed deaths1136. The dose of cisplatin used in the combinationarm of these trials ranged from 30 to 100 mg/M2,50 mg/m2 being used in six trials (table II). The survivalcurves (fig 3) showed an initial separation between thetwo treatments, although the curves had converged by

Combination

Single agent

0 1 2 3 4Years

Single non-platinum Numbers at riskagent 659 421 208 152 113

Platinum combination 670 441 249 167 121FIG 3 -Survival curves for comparison II (single non-platinum agent v plati

year 6. There was no overall significant differencebetween the two treatments (x2=1 09; p=0 30), theoverall relative risk being 0 93 (95% confidenceinterval 0 83 to 105) (fig 4). There was no obviousrelation between the dose of cisplatin used and theestimate of the treatment difference (fig 4). Neverthe-less, the pattern seemed to favour the combinations,most of the relative risk values lying to the left of unity(fig 4).

COMPARISON III: ADDITION OF PLATINUM TO A REGIMEN

Data were available from all eight trials eligible forcomparison III3133343738 (unpublished references (A)-(C)), which examined the effect of adding platinum toeither a single drug or a drug combination. As five trialscompared a single alkylating agent with the samealkylating agent plus cisplatin, these trials were eligiblefor inclusion in comparisons II and III. Data fromthese five trials were therefore common to both this andcomparison II.A total of 712 patients were randomised to the

non-platinum arms and 696 to receive the platinumcombinations. The median follow up period was sixyears, and a total of 1134 deaths were observed. Thedose of cisplatin used in these studies showed lessvariation than in comparison II, ranging from 40 to 80mg/M2, five trials using 50 mg/M2. The survival curves(fig 5) were similar to those in comparison II (fig 3), anearly trend favouring the cisplatin combination up toabout year 6. The overall difference between the twocurves was, however, not significant (X2 =2-8; p=0 1).The pooled relative risk of 0 91 (95% confidenceinterval 0 81 to 1-02) also favoured the platinumcombination (fig 6). There was no obvious correlationbetween the dose of cisplatin used and the estimate ofthe treatment difference.

COMPARISON IV: SINGLE AGENT PLATINUM VERSUS

5 6 7 8 PLATINUM COMBINATION

Data from all six trials eligible for comparison IV

89 61 43 26 available'01' 3940 (unpublished (D),

(E)), five of these trials using cisplatin and one

94 63 41 31 (unpublished reference (E)) using carboplatin. Theinum combination). imbalance in the numbers of patients randomised-


TABLE II-Doses ofcisplatinused in combination incomparison II

Cisplatindose(Mg/M2)

3040507580100

0

No of events / No entered

886

Relative risk



ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

TABLE iII-Doses ofplatinumused in comparison V

No ofCarbo- No of patients

Cisplatin platin trials in trialsdose dose using using(mg/m2) (mg/m2) dose dose

Single agent

100 400 3 392

Co?nbinaton20 350 1 33550 200 1 16550 250 1 5160 150 1 104

75 300 1 44780 350 1 173100 300 2 394

rU


Reference

Cisplatin <50 mg mV:Leonard et al32Wilbur et al 34

Combination Single agent O-E Variance Relative risk

IE.l40 /42 35 / 38 0.73 18.694/4 7/7 0.96 2.38

Subtotal 1.69 21.07

Cisplatin 50 mg/M2:Bell et al29 14/18COSA'5 163 /183Decker et al 3' 19 /21Masding et al 33 48/81Sturgeon et al 25 37 /40

VogI et al36 116 /128

Cisplatin >50 mg/M2: 50/57

Crowther (unpublished ref (1B))

MRC Gynaecological Cancer Working 39/ 45Party (unpublished ref (A))

Williams et al 35

18 /20161 / 18721 /2152 / 7639/43109 / 122

-4.61-1.05-8.31-6.47-3.02-3.37

-7.9481.009.59

24.9518.9956.13

1-0I

I

1-5 2-0

Subtotal 2.73 63.09

Total -20.31 282.76

0-0Test for heterogeneity: X2(,o) = 12.83; p= 0.233

0-5Combination better Single agent better

FIG 4-Relative risksfor comparison II. (Open lozenges represent pooled sub-total relative risks and 95% confidence intervals. Other details as inlegend tofig 2 and appendix C)

360 to single agent platinum, 565 to platinuamcombinations-was due to the three arm GICOG study(Gruppo Interegionale Cooperativo OncologicoGinecologia)'° having utilised two different platinumcombinations. A large proportion of patients incomparison IV (60%) were from that trial. A total of712 deaths were observed, and the median follow up

period was 6 5 years. One trial40 had a slightly differentobjective from the others, in that it compared high dosecisplatin (100 mg/m2) given as a single agent with lowdose cisplatin (20 mg/m2) given in combination. In allother trials in comparison IV the dose of platinum was

about equal in both arms and the dose of cisplatin did

not exceed 60 mg/m2. The analysis in comparison IVwas therefore performed both including and excludingthat trial.

All trials-The survival curves (fig 7) suggested a

difference in favour of the combination chemotherapyafter two years, which was maintained until about year

8. The overall difference between the curves was notsignificant (X2=2-53; p=01 1). Likewise, the overallrelative risk of 0-89 had a confidence interval straddl-ing unity (95% confidence interval 0-76 to 1 -04) (fig 8).

Excluding high dose-low dose study-After excludingthe high dose-low dose study the difference between thetwo curves achieved borderline significance (X2=4-82;p=003), and the overall relative risk of 0 85 (95%confidence interval 0-72 to 1 00) (fig 8) suggested a 15%reduction in the risk of death in favour of the platinumcombination.

CIO

016-

0-4-

0-2-

0 1

Numbers at riskNo platinum 712 460

Platinum 696 470FIG 5 -Survival curves for comparison

V Platinum

No platinum, -

2 3 4Years

COMPARISON V: CISPLATIN VERSUS CARBOPLATIN

Comparison V included trials comparing carbo-platin with cisplatin either as single agents or incombination with other drugs. Data were availablefrom all 11 eligible trials4"0 (unpublished reference(F)), in which 1023 patients were randomised toreceive cisplatin and 1038 to receive carboplatin. Thetotal number of observed deaths was 1771, and a

median follow up of only three years reflected thecomparative recency of these trials. Table III gives thedoses of platinum used. The survival curves (fig 9)showed no significant difference in overall survivalbetween the two treatments (x2=0 91; p=0-34), andthis was reflected in an overall relative risk of 1-05 (95%confidence interval 0 94 to 1-18) (fig 10). There was no

apparent relation between the estimate of treatmentdifference and whether single agents or combinationsof cisplatin and carboplatin were compared. Noindividual study indicated significant benefit for either

5 6 7 8 treatment (fig 10).

222 149 95 75 48 25258 170 121 85 47 25

III (non-platinum regimen v same regimen plus plantinum)

1013

DiscussionIn sharp contrast to the curative potential of chemo-

therapy in some cancers-for example, testicular


Subtotal -24.73 19859

39/52 5.27 22.13

43/49 0.18 20.72

42/44 -2.72 20.24

. . . . . . . . . . . . . . . . . . . .

n

0

887



ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/


Reference

Cisplatin <50 mg/M2:Wilbur et al 34

Cisplatin 50 mg / M2:COSA30Decker et al 3tMasding et al 3

De Oliveira et al 37

Omura etal3

Cisplatin >50 mg / M2:MRC Gynaecological Cancer

Working Party (trial b)(unpublished ref (A))

Turbow (unpublished ref (C))

Platinum No platinum O-E Variance

4/4 7/ 7 0.96 2.38

163/183

19/21

48/81

51/72201 / 244

161 / 187

21 /21

52 / 76

52 / 77

208 / 251

1.05

-8.31

-6.47

-0.41

-11.96

Relative risk

81.00

9.59

24.95

25.75

102.21

Subtotal -26.10 243.49

40 /51 43 / 49 0.18 20.72

30/40 34/44 -2.41 1598

Subtotal -2.23 36.70

Total -27.37 282.58

Test for heterogeneity: X27, = 8.40; p= 0.299 0*5 1-0 1-5 2-0Platinum better No platinum better

FIG 6-Relative risks for comparison III. (Details as in legends to figs 2 and 4 and appendix C)

1

CIO

0.

0*

Single agentplatinumPlatinum

combination

teratoma-plainly chemotia substantial improvementadvanced ovarian carcinonhave been gained, howevenot dramatic given the incbe extremely important to Isuch moderate improvencomparisons of interest tcollation and analysis of ution on over 8000 patiecountries.The lack ofrandomised t

control group did not pern

Combination

Single agent

0 1 2 3 4

Numbers at rsk Years360 246 125 73 47

565 397 230 156 116FIG 7-Survival curves for comparison IV (single agent platinum v pla

ierapy has not yet achieved chemotherapy in its own right. In addition, the widein survival in patients with variety of drug treatments tested did not permitna. Moderate benefits may analyses of individual regimens but led to trials beingr, which-though possibly grouped according to the type of drugs that they com-idence of the disease-may pared. The comparison of single non-platinum agentspublic health. To detect any with non-platinum combinations (comparison I)nents in survival in the included trials that were reported over roughly 10his overview presents the years, mainly during the 1970s. Consequently, notpdated individual informa- only were the treatment regimens that were groupednts from 45 trials in 11 together rather heterogeneous but the dose and

scheduling of drugs also varied a good deal. Some oftrials including an untreated the treatments that were being tested in those trialsnit an analysis of the role of might be considered ineffective today. Given these

limitations, this overview offers no clear evidence thatnon-platinum drug combinations are superior to non-platinum single agents.The comparison of single non-platinum agents with

platinum combinations (comparison II) might beexpected to have shown the largest treatment effect asit was comparing what is often thought of as the leastintensive form of treatment with the most intensive.There was no significant difference between the twotypes of treatment. Of these trials, one compared thesequential versus combined use of chlorambucil andcisplatin,30 and only one of the remainder (unpublishedreference (B)) did not allow the use of cisplatin onrelapse for those patients allocated to receive a singlenon-platinum agent. Many of the patients thereforereceived some form of platinum based treatment. Thusthe comparison may have been examining immediateversus delayed platinum treatment. The dose ofplatinum used in the combination arm varied, butmore than half of the studies used 50 mg/m2.The dose of cisplatin used in trials examining the

addition of platinum to a regimen (comparison III) wasmore homogeneous, and any additional drugs werecommon to both arms. This was therefore a somewhatcleaner comparison. Results were similar to those in5 6 7 8comparison II, an early trend favouring the platinumbased regimen being lost over time. The similarity was

35 24 16 9 not surprising as several trials were common to boththese comparisons and there were the same problems

9569 38 1 4associated with patients allocated non-platinum treat-

95 69 38 14 ment receiving cisplatin on relapse. Figures 4 and 6

rtinum combination) illustrate a pattern favouring the platinum combina-




ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

No of events / No enteredRelative risk

FIG 8-Relative risks for comparison IV. (Details as in legends to figs 2 and 4 and appendix C)

tions, which, although not significant, suggests a

survival benefit associated with using immediateplatinum treatment as opposed to not giving or

delaying platinum until relapse.

SURVIVAL PATTERNS

The survival curves for the comparison of singleagent platinum with platinum combinations (compari-son IV) seemed to favour the combination arm aftertwo years, in contrast with the earlier treatment advan-tage suggested in comparisons II and III. A possibleexplanation might be that in comparison IV bothtreatment groups received some form of immediateplatinum treatment. Thus the comparison did notsuffer from the problem of "control" patients receiving

1 -0

0-8-

_ 06-

Cl)

0-4-

0-2-

0 1

Numbers at riskCisplatin 1023 693

Carboplatin 1038 717

Cisplatin

Carboplatin

Years

383363

4

8172

195187

FIG 9-Survival curves for comparison V (cisplatin v carboplatin)

platinum on relapse from which, arguably, they mightderive a late benefit. The results of the analyses implythat platinum in combitation is more effective thansingle agent platinum in the doses used, particularlyafter two years. Notably, however, in most studies thedose of platinum used as a single agent was lower thanis currently used. It will therefore be important todetermine whether the apparent difference betweentreatments was due solely to the addition ofother drugsor was, in fact a question ofdose as implied in the studyby Wiltshaw et al."0 If so, patients receiving drugcombinations fared better simply because theyreceived a higher dose of drug irrespective of type.

Similar results and a similar problem of dose were

found in a separate overview60 which examined the roleof doxorubicin in ovarian carcinoma. That overview,which reviewed studies of cyclophosphamide, doxo-rubicin, and cisplatin (CAP) versus cyclophosphamideand cisplatin (CP), found a significant survival advan-tage for patients treated with CAP over those treatedwith CP but also concluded that the observed differ-ence might have been due simply to patients in thethree drug combination arm having higher total dosesof drugs. The only way to resolve this is by comparingcurrently favoured doses of single agent platinumwith platinum combinations in a large prospectiverandomised clinical trial.The results of the comparison of cisplatin and

carboplatin (comparison V) offer no good evidence thatcisplatin is either superior or inferior to carboplatin interms of survival when given either as a single agent or

in combination. Some of the trials included in thiscomparison, however, were at a comparatively earlystage of follow up, so that although the informationpresented on the first four years was reliable, furtherfollow up was required. It should also be noted that thetrials included in comparison V used doses of carbo-platin based on surface area rather than on renalclearance as currently recommended.6

It is clear from the confidence intervals associatedwith the overall relative risks that even combining theresults from many studies does not provide enoughpatients to make any firm statements about the rIsultsof the comparisons except perhaps for the firsfbne.Comparisons of platinum with non-platinum regimens(comparisons II and III) each contained fewer than


Reference Combination Single agent O-E Variance

Excluding low / high dose trial:

Cohen etal39 17/18 15/18 1.12 7.99 0

GICOG0 298 / 384 156 /178 -21.72 98.60

Gilby etal Triala 6/17 7/13 -2.57 3.16 a(unpublished ref(D)) ,Trial b 0 /3 1 /2 -

Kaye (unpublished ref (E)) 52 /76 55 /85 2.27 26.68

Tomirotti etaal 12/23 15/21 -2.99 6.72

Subtotal -23.89 143.15

High/low dosetrial: 40/44 38/43 5.15 19.44 __

Wiltshaw eta140

Total -18.74 162.59

0-0 05 1-0 1-5 20

Test for heterogeneity: 2(4, = 7.76; p= 0.170 Combination better Single agent better

I

No of events / No entered Relative risk

889



ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

1500 patients, and the comparison of platinum as asingle agent and in combination had fewer than 1000.Consequently, although all of these comparisons weresuggestive oftreatment benefit associated with first lineplatinum based combination chemotherapy, no con-clusive evidence emerges. Trials of about double thesize of these would be required to establish reliablywhether these trends represent actual treatment bene-fits. The comparison of cisplatin and carboplatinapproaches such numbers of patients and providesreliable information on fairly short term survival (up tofour years), although, as mentioned above, longerfollow up is required. This uncertainty emphasises thatin the past clinical trials have been an order -ofmagnitude too small to detect the size of differenceswhich this overview suggests may be realistic forcurrently available chemotherapy regimens.

IMPLICATIONS

Arguably an overview using a cross section ofpatienttypes from many different trials is more likely to reflectthe real world than an individual trial and therefore toestimate more accurately the type of treatment effectsthat are generally achievable. Even so, just as an indi-vidual trial cannot provide a prescription for treatingany individual patient, neither can an overview.Nevertheless, a well conducted overview provides theleast biased and most accurate summary of existinginformation from clinical trials, from which the clini-cian can make his or her own decisions about diseasemanagement.

Despite not being able to provide any firm conclu-sions about the most effective forms of treatment foradvanced ovarian cancer, this overview raises threeimportant hypotheses-namely, (a) platinum com-binations are generally better than non-platinumregimens as first line treatment (comparisons II andIII); (b) platinum combinations are generally betterthan single agent platinum when platinum is used inthe same dose (comparison IV (this is supported by theresults from the CAP/CP overview,' as if the combina-tion of cyclophosphamide, doxorubicin, and cisplatinis superior to cyclophosphamide plus cisplatin it islikely to be superior to standard dose cisplatin alone));

(c) cisplatin and carboplatin are equally effective(comparison V).At the first meeting of the Advanced Ovarian Cancer

Trialists Group, held in June 1990, when the results ofthese analyses were presented in a preliminary form,consideration of the above hypotheses led to theproposal that it would be appropriate to compare thecyclophosphamide, doxorubicin, cisplatin regimenwith optimal dose carboplatin in a large internationaltrial. Such a comparison would have important impli-cations for the management of advanced ovariancancer. As discussed above, even a moderate improve-ment in survival could prolong the lives ofthousands ofwomen worldwide. A null result would be equallyimportant as the lower toxicity of carboplatin whengiven as a single agent would almost certainly improvea patient's quality of life. An international trial, icon-2(International collaborative ovarian neoplasm study),has now been set up to compare cyclophosphamide,doxorubicin, and cisplatin versus optimal dose carbo-platin and aims at accruing 2000 patients worldwide. Aseries of independent parallel trials, each with closelysimilar protocols, are planned, the data from whichwill be pooled in a prospective overview. The newstudy will include centres that participated in thisoverview, but the trial is open to any clinician whowishes to enter patients either as an individual or aspart of a group. At present icon-2 has been launched inthe United Kingdom and Italy and is recruitingpatients.

The advanced ovarian cancer overview was initiated by theBritish MRC, which also funded the overview administration.We are indebted to the following institutions that funded theAdvanced Ovarian Cancer Trialists Group meeting: BristolMyers Squibb Company; British MRC; BUPA; CancerResearch Campaign; Ciba-Geigy; Eli Lilly and Co Ltd;Epidemiology and Public Health Research Unit, Universityof Surrey; Fitton Trust; Harold Smith Charitable Trust;Imperial Cancer Research Fund; Lederle; Schering-PloughLtd; Sir Samuel Scott ofYews Trust; and Sterling Oncology.We also thank the staff of Bristol Myers PharmaceuticalResearch and Development Division, Wallingford, USA, forhelp in identifying relevant studies and for providing datafrom trials that they had sponsored. We are, indeed, grateful

FIG 10 -Relative risks forcomparison V. (Details as inlegends to figs 2 and 4 andappendix C)




ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

to all the data centres around the world whose efforts inproviding individual patient data made this overview possible.

Organisations and groups that collaborated in the overviewwere the British MRC, Eastern Cooperative Oncology Group,European Organisation for Research and Treatment ofCancer, German Ovarian Cancer Study Group, GruppoIntergionale Cooperativo Oncologico Ginecologia, GruppoOncologico Nord Ovest, Gynecologic Oncology Group,Gynaecological Group Clinical Oncological Society ofAustralia, Mario Negri Institute, National Cancer Institute ofCanada, Southwest Oncology Group, and the NorthernCalifornia Oncology Group.Members of the Advanced Ovarian Cancer Trialists Group

were K Aabo, ,P Adnitt, M Adams, D S Alberts, V Barley,D R Bell, U Bianchi, J Blessing, M Buyse, G Bolis,H S Brodovsky, H Bruckner, R Canetta, V Chylak, C JCohen, N Colombo, P F Conte, D Crowther, C F De Oliveira,J H Edmonson, E Gilbey, D Guthrie (secretariat and writingcommittee), S B Kaye, A H Laing, F Landoni, R C Leonard,C Lewis, P Y Liu, C Mangioni, S Marsoni (writing com-mittee), H Meerpohl, U Muller, G A Omura (secretariat),M K B Parmar (secretariat and writing committee), J Pater(writing committee), S Pecorelli, M Presti, W Sauerbrei, R VSmalley, H J Solomon, L A Stewart (secretariat and writingcommittee), J F G Sturgeon, M H N Tattersall (writingcommittee), J T Wharton (writing committee), W W tenBokkel Huinink, M Tomirotti, W Torri, C Trope, M MTurbow, J B Vermorken, M J Webb, D W Wilbur, C JWilliams (secretariat and writing committee), E Wiltshaw, RWood (secretariat), and B Y Yeap.

1 Parkin DM, Laara E, Muir CS. Estimates of the worldwide frequency ofsixteen major cancers in 1980. IntJ Cancer 1988;41:184-97.

2 Young RC, Chabner BA, Hubbard SP, et al. Advanced ovarian carcinoma: aprospective clinical trial of melphalan (L-PAM) versus combination chemo-therapy. NEngl7Med 1978;299:1261-6.

3 Wiltshaw E, Kroner T. Phase II study of cisdichlorodiammineplatinuin (II)(NSC 119875). Cancer Treat Rep 1976;60:55-60.

4 Williams CJ, Stevensen K, Buchanan R, et al. A pilot study of cisdichloro-diammineplatinum (II) in combination with Adriamycin and cyclophospha-mide in previously untreated patients and as a single agent in previouslyuntreated patients. Cancer Treat Rep 1979;63:1745-53.

Williams CJ, Mead GM, Arnold A. Chemotherapy of advanced ovariancarcinoma: initial experiences of a platinum based combination. CancerTreatRep 1982;63:311-7.

6 Bolis G, Bortolozzi G, Carinelli G, et al. Low-dose cyclophosphamide versusAdriamycin plus cyclophosphamide in advanced ovarian cancer. CancerChemother Pharmacol 1980;4:129-32.

7 Bertelsen K, Jakobsen J, Andersen JE, et al. A randomized study ofcyclophosphamide and cis-platinum with or without doxorubicin inadvanced ovariani carcinoma. Gynecol Oncol 1978;62:1375-7.

8 Omura GA, Bundy BN, Berek JS, Curry S, Delgado G, Mortel R. Random-ized trial of cyclophosphamide pluts cisplatin with or withour doxorubicin inovarian carcinoma: a Gynecologic Oncology Group study. j Clin Oncol1989;7:457-65.

9 Hubbard S, Barker P, Young RC. Adriamycin therapy for advanced ovariancancer recurrent after chemotherapy. Cancer Treat Rep 1978;62:1375-7.

10 Gruppo Interegionale Cooperativo Oncologico Ginecologia. Randomisedcomparison of cisplatin with cyclophosphamide/cisplatin and withcyclophosphamide/doxorubicin/cisplatin in advanced ovarian cancer.Lancet 1987;ii:353-9.

11 Tomirotti M, Perrone S, Gie P, et al. Cisplatin (P) versus cyclophosphamide,Adriamycin and cisplatin (CAP) for stage III-IV epithelial ovariancarcinoma: a prospective randomised trial. Tuonor 1988;74:573-7.

12 Marsoni S, Torri W, Taiana A, et al. Critical review of quality anddevelopment of randomised clinical trials and their influence on thetreatment of advanced ovarian cancer. Annals ofOncology 1990;1:343-50.

13 Medical Research Council Gynaecological Cancer Working Party. An over-view in the treatment of advanced ovarian cancer. Br J Cancer 1990;61:495-6.

14 Simes R]. Publication bias: the case for an international registry of clinicaltrials. J Clin Oncol 1986;4:1529-41.

15 United Kingdom Co-ordinating Committee on Cancer Research. UK cancertrials register. London: UKCCCR, 1986.

16 Aabo K, Hald I, Horbov S, et al. A randomised study of single agent vscombination chemotherapy in FIGO stages IIB, III and IV ovarianadenocarcinoma. EurJ Cancer Clin Oncol 1985;21 :475-81.

17 Brodovsky HS, Bauer M, Horton J, Elson PJ. Comparisoii of melphalan withcyclophosphamide, methotrexate and 5-fluorouracil in patients with ovariancancer. Cancer 1984;53:844-52.

18 Bruckner H, Pagano M, Falkson G, et al. Controlled prospective trial ofcombination chemotherapy with cyclophosphamide, Adriamycin and5-fluorouracil for the treatment of advanced ovarian cancer: a preliminaryreport. Cancer Treat Rep 1979;63:297-9.

19 Chylak V, Kolaric K, Krusic K. Controlled clinical trial of chemotherapy inadvanced ovarian cancer-cyclophosphamide mono-chemotherapy versus acombination of Adriamycin, cyclophosphamide, 5-fluorouracil and metho-trexate. Lijec Vjesen 1986;109:230-4.

20 Edmonson JH, Flemming TR, Decker DG, et al. Different chemotherapeuticsensitivities and host factors affecting prognosis in advanced ovariancarcinoma versus minimal residual disease. Cancer Treat Rep 1979;61:355-7.

21 Miller A, Klaassen DJ, Boyes DA, et al. Combination v sequential therapywith melphalan, 5-fluorouracil and methotrexate for advanced ovariancancer. Can Med AssocJ 1980;123:363-7 1.

22 Medical Research Council Working Party on Ovarian Cancer. MedicalResearch Council study on chemotherapy in advanced ovarian cancer.BrJ Obstet Gynaecol 1981;88: 1174-85.

23 Omura GA, Morrow CP, Blessing JA, et al. A randomised comparison of

melphalan versus melphalan plus hexamethylmelamine versus Adriamycinplus cyclophosphamide in ovarian carcinoma. Cancer 1983;51:783-9.

24 Park RC, Blom J, Disaia PJ, Lagasse LD, Blessing JA. Treatment of womenwith disseminated or recurrent advanced ovarian cancer with melphalanalone in combination with 5-fluorouracil and dactinomycin or with thecombination of Cytoxan, 5-fluorouracil and dactinomycin. Cancer 1980;45:2529-42.

25 Sturgeon JFG, Fine S, Gospodarowicz MK, et al. A randomised trial ofmelphalan alone vs combination chemotherapy in advanced ovarian cancer.Proceedings oftheAnterican Society ofClintcal Oncology 1982;1: 108.

26 Trope C. Melphalan with and without doxorubicin in advanced ovariancancer. Obstet Gynecol 1987;70:582-6.

27 Turbow MM, Jones H, Yu VK, Greenberg B, Hannigan J, Torti FM.Chemotherapy of ovarian carcinoma: a comparison of melphalan vsAdriamycin-cyclophosphamide. Proceedings of the American Association ofCancer Research and American Society ofClinical Oncology 1980;21: 1%.

28 Wharton JT, Edwards CL, Rutledge FN. Long-term survival after chemo-therapy for advanced epithelial ovarian carcinoma. Am J Obstet Gynecol1984;148:997-1005.

29 Bell DR, Woods RL, Levi JA, Fox RM, Tattersall MHN. Advanced ovariancancer: a prospective randomised trial of chlorambucil versus combinedcyclophosphamide and cis-diamminedichloroplatinum. Aust N Z J Med1982;12:245-9.

30 Gynaecological Group, Clinical Oncological Society of Australia and theSydney Branch, Ludwig Institute for Cancer Research. Chemotherapy ofadvanced ovarian adenocarcinoma: a randomised comparison of combina-tion versus sequential therapy using chlorambucil and cisplatin. GynecolOncol 1986;23:1-13.

31 Decker DG, Thomas MD, Fleming R, et al. Cyclophosphamide plus cis-platinum in combination: treatment program for stage III or IV ovariancarcinoma. ObstetGynecol 1982;60:481-7.

32 Leonard RC, Smart GE, Livingston JRB, et al. Randomised trial comparingprednimustine with combination chemotherapy in advanced ovariancarcinoma. Cancer ChemotherPharmacol 1989;23:105-10.

33 Masding J, Sarkar T, White JF, et al. Intravenous treosullan versuisintravenous treosulfan plus cisplatinum in advanced ovarian carcinoma.Br.7 Obstet Gvnaecol 1990;97:342-5 1.

34 Wilbur DW, Rentschler RE, Wagner RJ, Keeney ED, King A, Hilliard DA.Randomized trial of the addition of cts-platin (DDP) and/or BCG tocyclophosphamide (CTX) chemotherapy for ovarian carcinoma.J Surg Oncol 1987;34:165-9.

35 Williams CJ, Mead GM, Macbeth FR, et al. Cisplatin combinatiott chemo-therapy versus chlorambucil in advanced ovarian carcinoma: mature resultsof a randomized trial. J Clin Oncol 1985;3: 1455-62.

36 Vogl S, Kaplan B, Pagano M. Diatnminedichloroplatinum-based combinationchemotherapy is superior to melphalan for advaniced ovarian cancer whenage >50 and tumor diatneter >2 cm. Proceedings of the American Society ofClinical Oncology 1982;1: 119.

37 De Oliveira CF, Lacave AJ, Villani C, et al. Randomised comparison ofcyclophosphamide, doxorubicin and cisplatin for the treatment of advancedovarian cancer. EurJ7 Gynaecol Oncol 1990;11:323-30.

38 Omura G, Blessing JA, Ehrlich CE, et al. A randomized trial of cyclophospha-mide and doxorubicin with or without cisplatin in advanced ovariancarcinoma. A Gynecologic Oncology Group study. Cancer 1986;57:1725-30.

39 Cohen CJ, Goldberg JD, Holland JF, et al. Improved therapy with cisplatillregimens for patients with ovarian carcinoma (FIGO stages III and IV) asmeasured by surgical end-staging (second-look) operation. Am J ObstetGynecol 1983;145:955-65.

40 Wiltshaw E, Evans B, Rustin G, Gilbev E, Baker J, Barker G. A prospectiverandomised trial comparing high-dose cisplatin with low-dose cisplatin andchlorambucil in advanced ovarian carcinoma.j Clin Oncol 1986;4:722-9.

41 Adams M, Kerby IJ, Rocker I, Evans A, Johansen K, Franks CR. Acomparison of the toxicity and efficacy of cisplatin and carboplatin inadvanced ovarian cancer. Acta Oncol 1989;28:57-60.

42 Mangioni C, Bolis G, Pecorelli S, et al. Randomized trial in advanced ovariancancer comparing cisplatin and carboplatin. J Natl Cancer Inst 1989;81:1464-71.

43 Wiltshaw E, Evans B, Harland S. Phase III randomised trial cisplatin versusJM8 (carboplatin) in 112 ovarian cancer patienits, stages III alnd IV.Proceedings of the American Society ofClinical Oncology 1985;4:121.

44 Alberts D, Green S, Hannigan E. Improved efficacy of carhoplatinlcyclophosphamide vs cisplatin/cyclophosphamide: preliminary report of aphase III, randomised trial in stages III-IV, suboptimal ovarian cancer.Proceedings of the American Society ofClinical Oncology 1989;8:588.

45 Anderson H, Wagstaff J, Crowther D, Evans A, Johansen K, Franks CR.Comparative toxicity of cisplatin, carboplatin (CBDCA) and iproplatin(CHIP) in combination with cyclophosphanide in patients with advancedepithelial ovarian cancer. EurJ Cancer Clin Oncol 1988;24:1471-9.

46 Conte PF, Bruzzone M, Carnino F, et al. Carboplatin, doxorubicin andcyclophosphamide versus cisplatin, doxorubicin and cyclophosphamide: arandomized trial in stage III-IV epithelial ovarian carcinoma. J7 Clin Oncol(in press).

47 Edmonson JH, McCormack GM, Wieand HS, et al. Cyclophosphamide-cisplatin vs cyclophosphamide-carboplatin in stage III-IV ovarianicarcinoma: a comparison of equally myelosuppressive regimens. J7 NaitlCancer Inst 1989;81:1500-4.

48 Kato T, Nishimura H, Yamabe T, et al. Phase III study of carboplatin forovarian cancer.Japanese.ournal of Cancer Chemother 1988;15:2297-304.

49 Pater J. Cyclophosphamide (C)/cisplatin (CDDP) versus cyclophosphamide/carboplatin (CBDCA) in macroscopic residual ovarian cancer. Initial resultsof a National Cancer Institute of Canada Clinical Trials Group trial.Proceedings of the American Society ofClinical Oncology 1990;9:121.

50 ten Bokkel Huinink WW, van der Burg ME, van Oosterom AT, et al.Carboplatin in combination chemotherapy for ovarian cancer. Cancer TreatRev 1988;l5(suppl B):9-15.

51 Adams M, Johansen KA, James KW, Rocker I. A controlled clinical trial inadvanced o)varian cancer. Clin Radiol 1982;33: 161-3.

52 Barlow JJ, Lele SB, Emrich UJ. Long-term survival rates with variouschemotherapeutic regimens in stage III and IV ovarian adenocarcinoma.AmJfObstetGynecol 198S;}52:310-4.

53 Carmo-Pereira J, Oliveira Costa F, Henriques E, Almneida Ricardo J.Advanced ovarian carcinoma: a prospective and randomised clinical trial ofcyclophosphamide vs combination cytotoxic chemotherapy (Hexa-CAF).Cancer 1981;48:1947-51.

54 Delgado G, Smith FP, McLaughlin EK, Tuholski N. Single agent vscombination chemotherapy for ovarian cancer. Am J Clin Oncol 1985;8:33-7.

BMJ VOLUME 303 12 OCTOBER 1991 891



ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

55 De Palo G, De I,ena M, Bonadonna G. Adriamycin vs Adriamycin plusmelphalan in advanced ovarian carcinoma. Cancer Treat Rep 1977;61:355-7.

56 Gronroos M, Nieminen U, Kauppila A, Kauppila 0, Saksela E, Vayrynen M.A prospective randomised national trial for treatment of ovarian cancer: therole of chemotherapy and external radiation. Eurj Obstet Gynecol ReprodBtol 1984;17:33-42.

Appendix ADetails ofrandomised clinical trials included in advanced ovarian cancer overview

Comparisons I and II

No ofSingle patients in

Reference drug Combination overview

Comparlson I: single non-platinum agent v non-platinum combinationAabo et al" CTX ADR, CTX, FU 179

BUBolis et atl CTX ADR, CTX 74Brodovskv et alt L-PAM CTX, FU, MTX 409Bruckner et al" L-PAM MTX, TSPA 331

ADR, CTX, FUChylak et al" CTX ADR, CTX, FU, MTX 69Edmonson et al7° CTX ADR, CTX 111Miller et al" L-PAM FU, L-PAM, MTX 254MRC Working Party on Ovarian Cancer2" CTX CTX, HMM, MTX 344MRC Gynaecological Cancer Working CTX ADR, CTX 116

Party (unpublished ref (A))Omura et al" L-PAM ADR, CTX 339

HMM, L-PAMPark et alt4 L-PAM FU, L-PAM 418

ACT-D, FU, L-PAMACT-D, CTX, FU

Sturgeon ct al L-PAM CTX, FU, HMM, MTX 83Trope2 L-PAM ADR, L-PAM 168Turbow et alt> L-PAM ADR, CTX 48Wharton et al-t L-PAM CTX. HMM 75

ADR, CTX, HMMWharton et al" L-PAM CTX, HMM 128

ADRHMM

Comparison II: single non-platinum agent v platinum combinationBell et alt" CLB CTX, CACP 38COSA" CLB CLB, CACP 370Crowther (unpublished ref (B)) CTX BLE, CTX, CACP 109Decker et al" CTX CTX, CACP 42Leonard et al" PRED FU, HMM, PRED, CACP 80Masding et al" TREO TREO, CACP 157MviRC Gynaecological Cancer Working CTX CTX, CACP 100Party (unpublished ref (A))

StuLrgeon et al" L-PAM ADR, CTX, CACP 83Wilbur et at4" CTX CTX, CACP 11Williams et alt CLB ADR, CTX, CACP 89Vogl et alt" L-PAM ADR, CTX, HMM, CACP 250

Comparisons II-V

No ofpatients in

Reference Arm 1 Arm 2 overview

Comparison III: addition ofplatinum to regimenCOSA"' CLB CLB, CACP 370Decker et at CTX CTX, CACP 42De Oliveira et alo ADR, CTX ADR, CTX, CACP 149Masding et all' TREO TREO, CACP 157MRC Gynaecological Cancer Working CTX CTX, CACP 100

Party (unpublished ref (A))Omura et alt ADR, CTX ADR, CTX, CACP 495Turbow (unpublished ref (C)) ADR, CTX ADR, C, ,, CACP 84Wilbur et alt CTX CTX, CACP 11

Comparison IV: single-agent platinum v platinum combinationColsen et al" CACP ADR, CACP 36GICOG"' CACP ADR, CACP 562

ADR, CTX, CACPGilby et al (unpublished ref (D)) CACP IFOS,CACP 30

CBDSA IFOS, CBDSA 5Kaye (unpublished ref (E)) CBDSA CLB, CBDSA 161Tomirotti et alt CACP ADR, CTX, CACP 44Wiltshaw et aP° CACP CLB, CACP 87

Comparison V: cisplatin v carboplatinAdams et al" CBDSA CACP 88Mangioni et at"' CBDSA CACP 173Wiltshaw etal" CBDSA CACP 131Alberts et a`4" CTX, CBDSA CTX, CACP 338Andersoti et at4 CBDSA, CTX CACP, CTX 56Conte et at4' ADR, CTX, CBDSA ADR, CTX, CACP 165Edmonson et al4 CTX, CBDSA CTX, CACP 104Kato et al' ADR,CTX,CBDSA ADR,CTX,CACP 51Meerpohl et al (unpublished ref (F)) CTX, CBDSA CTX, CACP 173Pater" CTX, CBDSA CTX, CACP 447ten Bokkel Huinink et alt" ADR, CTX, HMM, ADR, CTX, HMM, CACP 335

CBDSA

ACT-D=Actinomycin D. ADR=Doxorubicin. BLE=Bleomycin. BU=Busulphan. CACP=Cisplatin. CBDSA=Carboplatin. CLB=Chlorambucil. CTX=Cyclophosphamide. FU= Fluorouracil. HMM= Hexamethylmelarmine.IFOS =Ifosfamide. L-PAM=Melphalan. MTX=Methotrexate. PRED=Prednimustine. TSPA=Thiotepa.TREO=Treosulfan.

57 Senn HJ, Lei D, Castano-Almendral A, et al. Chemo-(hormon)-therapiefortgeschrittener ovarialkarzinome der FIGO-stadien III und IV.Prospektive SAKK-studie 20/71. Schweiz Med Wochenschr 1980;110:1202-8.

58 Carmo-Pereira J, Oliveira Costa F, Henriques E. Cis-platinum, Adriamycinand hexamethylmelamine vs cyclophosphamide in advanced ovariancarcinoma. Cancer Chemother Pharmacol 1983;10: 100-3.

59 Harvey HA, Lipton A, Simmonds M, et al. A randomised trial of Alkeranversus cvclophosphamide, hexamethvlnielamine, Adriamycin and cis-platinum combination chemotherapy in advanced ovarian carcinoma.ClinRes 1982;30:418.

60 Ovarian Cancer Meta-Analysis Project. CP versus CAP chemotherapy ofovarian carcinoma: a meta-analysis. J Clin Oncol (in press).

61 Calvert AH, Newell DR, Gumbrell L, et al. Carboplatin dosage: prospectiveevaluation of a simple formula based on renal function. 7 Clin 0nol1989;5: 1748-56.

62 Early Breast Cancer Trialists Collaborative Group. Treatment of early breastcancer. Vol 1. Worldwide evidence 1990. Oxford: Oxford University Press,1990.

Unpublished trials included in overview;A Medical Research Council Gynaecological Cancer Working Party. Inadequacy

of trials of chemotherapy in advanced ovarian carcinoma: a randomised trialof three regimens. (In preparation.)

'1B) Crowther D. A randomised trial of chemothterapy in advanced residual (stageIIB-IV) ovarian cancer. Manchester Ovarian Cancer Clinical Study Groupprotocol, 1986. (Unpublished.)

C: Turbow MM. Chemotherapy of advanced ovarian cancer: Adriamycin-cyclophosphamide versus piatinum-Adriamycin-cvclophosphamide.Northern California Oncology Group protocol 5091, 1980. (Unpublished.)

(D) Gilby E, Pollard W, Barley V, et al. Ovarian cancer trial, 1986. South westoncology study. (Unpublished.)

(E' Kaye S. Randomised study of carboplatin and chlorambucil/carboplatin inadvanced ovarian cancer. (Unpublished.)

F) Meerpoihl HG, Kuhnle H, Sauerbrei W, Achterrath W, Pfeiderer A.Cyclophosphamide/cisplatin (CTXJPT) vs CTX/carboplatin (carbopt) inadvanced ovarian carcinoma: a randomised multicenter study. 15th Inter-national cancer congress, Hamburg, 1990 [meeting abstract]. Hamburg:UICC. 1990.

(Accepted 4 iuly 1991)

Appendix BTrials not included in advanced ovarian cancer overview

No ofSingle patients

Reference drug Combination randomised

Companrson I: single non-platinum agent v non-platinum combinationAdams et al"* L-PAM ADR, CTX, FU 40Barlow etIal"t L-PAM ACT-D, CTX, FU 108Carmo-Pereira CTX CTX, FU, HMM, MTX 57

et al"tDelgado et al"t L-PAM CTX, FU, HAIM 27De Palo etal1'5 ADR ADR, L-PAM 29Gronroos et alt TREO CTX, TREO 108

ADR, FU, TREOSenn et alI7| CTX CTX, FU 89Young etal2§ L-PAM CTX, FU, HMM, MTX 80

Comnparlson II: single non-platinum agent v platinum combinationCarmo-Pereira CTX CACP, ADR, HMM 59

et al"tHarvey etal"5 L-PAM CACP, ADR, CTX, HMM 40

*Trial unavailable because data lost.tTrial unavailable because of inability or unwillingness of those responsibleto collaborate itl os cC AM.tTrial ineligible because did not seem appropriately randomised.5Trial currently unavailable but promised at later date.(Trial unavailable because data destroyed.ACT-D=Actinomycin D. ADR=Doxorubicin. BLE=Bleomycin.BU=Busulphan. CACP==Cisplatin. CBDSA=Carboplatin.CLB=Chlorambucil. CTX=Cyclophosphamide. FU=Fluorouracil.HMM=Hexamethylmelamine. IFOS=Ifosfamide. L-PAM =lelphalan.MTX= Methotrexate. PRED- Prednimustine. TSPA= Thiotepa.TREO=Treosulfan.

Appendix CGRAPHICAL ANALYSES OF RELATIVE RISKS IN AVAILABLE TRIALS

In the graphical displays of relative risk in the publishedand unpublished trials (figs 2, 4, 6, 8, 10) the numbers ofpatients and deaths are given for each arm of the trialstogether with the observed minus the expected numbers ofdeaths for the treatment arm. A negative observed minusexpected value indicates that the treatment group fared betterthan the controls whereas a positive observed minus expectedvalue indicates the opposite. The relative risk for each studyis plotted as an open square, whose size is directly propor-tional to the amount of information in that trial. Horizontallines extend outwards from this to display the 99% confidenceinterval. The pooled relative risk is shown as an occludedlozenge whose extremities denote the 95% overall confidenceinterval. Wider confidence intervals-that is, 99% as opposedto the more conventional 95%-are imposed on the individualtrials because the problems of their multiplicity increase the

892 BMJ VOLUME 303 12 OCTOBER 1991



ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

chance of observing a false positive result.62 Pooled relativerisks for subtotals are shown as open lozenges. The verticalline through unity indicates the point where there is nodifference between treatments. Trials indicating an advantagefor treatment A lie to the left of this line and those showingadvantage to treatment B lie to the right. Individual trialsindicating a statistically significant result at the p= 001 level,lie wholly to one side of the line, such that their confidenceintervals will not straddle it.When trials had multiple treatment arms and made more

than one comparison of interest the patients from the relevantarms were included in each appropriate comparison, providedthat there had been a direct randomisation between thetreatment categories used. Thus two trials25 (unpublished

reference (A)) are included in more than one comparison. Intrials where more than one arm was of the same treatmentcategory the patients in these arms were grouped together foranalysis. For example, in the trial comparing cisplatin versuscisplatin plus cyclophosphamide versus cisplatin, cyclophos-phamide, and doxorubicin'° the patients from the twocombination arms were amalgamated and compared with thepatients assigned single agent treatment. Two trials stoppedrandomising to certain arms early,2324 and a further twostudies employed two separate randomisation schemes23(unpublished reference (A)). Each of these trials was sub-divided into the appropriate number of data sets, and foranalysis each was treated as an independent trial and labelleda, b, etc.

Scientific SteeringCommittee of the SimonBroome Register GroupMembers of the scientificsteering committee andparticipating physicians andclinics are listed at the end ofthis report.

Correspondence to:Dr Margaret Thorogood,Department of PublicHealth and Primary Care,Gibson LaboratoryBuilding, RadcliffeInfirmary, OxfordOX2 6HE.

BMJ 1991;303:893-6

Risk of fatal coronary heart disease in familial hypercholesterolaemia

Scientific Steering Committee on behalf of the Simon Broome Register Group

AbstractObjectives-(a) To determine the excess mortality

from ali causes and from coronary heart disease inpatients with familial hypercholesterolaemia; (b) toexamine how useftl various criteria for selectivemeasurement of cholesterol concentration in cardio-vascular screening programmes are in identifyingthese patients.Design-Prospective cohort study.Setting-Eleven hospital outpatient lipid clinics in

the United Kingdom.Patients-282 men and 244 women aged 20-74

with heterozygous familial hypercholesterolaemia.Main outcome measure-Standardised mortality

ratio, all adults in England and Wales being taken asstandard (standardised mortality ratio= 100 forstandard population).Results-The cohort was foliowed up for 2234

person years during 1980-9. Fifteen of the 24 deathswere due to coronary heart disease, giving a stand-ardised mortality ratio of 386 (95% confidence inter-val 210 to 639). The excess mortality from this causewas highest at age 20-39 (standardised mortality ratio9686; 3670 to 21 800) and decreased significantly withage. The standardised mortality ratio for all causeswas 183 (117 to 273) and also was highest at age 20-39(standardised mortality ratio 902; 329 to 1950).There was no significant difference between menand women. Criteria for measurement of cholesterolconcentration in cardiovascular screening program-mes (family history, presence of myocardial infarc-tion, angina, stroke, corneal arcus, xanthelasma,obesity, hypertension, diabetes, or any of these)were present in 78% of patients.

Conclusions-Familial hypercholesterolaemia isassociated with a substantial excess mortality fromcoronary heart disease in young adults but may notbe associated with a substantial excess mortality inolder patients. Criteria for selective measurement ofcholesterol concentration in cardiovascular screen-ing programmes identify about three quarters ofpatients with the clinically overt condition.

IntroductionFamilial hypercholesterolaemia is an autosomal

dominant disorder of lipoprotein metabolism charac-terised by mutations of the low density lipoproteinreceptor resulting in an accumulation of low densitylipoprotein cholesterol in the plasma.' Heterozygousfamilial hypercholesterolaemia has been estimated toaffect about one in 500 of the British population. ' Mostaffected subjects remain undiagnosed.

It is generally accepted that patients with familialhypercholesterolaemia are at greater risk of coronaryheart disease than those with polygenic hypercholes-terolaemia. The first report of a substantially increasedrisk in heterozygous familial hypercholesterolaemiadescribed a 51% chance of fatal or non-fatal coronaryheart disease by the age of 50 in men and a correspond-ing risk of 12% in women.2 That and most subsequentstudies were retrospective analyses, which are subjectto inherent biases. Only one study seems to havefollowed up a large cohort ofpatients (588 heterozygousfamilial hypercholesterolaemic patients3), and no lifetable survival analysis has been reported. Publisheddata on morbidity and mortality relate to a time whenmost patients did not receive effective lipid loweringtreatment. A reassessment of the mortality associatedwith familial hypercholesterolaemia is thereforeappropriate as the widespread use of lipid loweringdrugs during the past decade, especially bile acidsequestrants and fibric acid derivatives, may havereduced cardiovascular morbidity and mortality.We have recruited a cohort of patients with definite

or possible familial hypercholesterolaemia, which willallow epidemiological, clinical, genetic, and metabolicstudies to be performed in a well characterised popula-tion. We report the characteristics of 526 patients withdefinite familial hypercholesterolaemia and theirmortality during the first 10 years of follow up. Wehave also used the information collected at registrationto examine how useful the various criteria for selectivemeasurement of cholesterol concentration suggestedfor use in cardiovascular screening programmes4 are inidentifying such patients.

Patients and methodsRecruitment of patients to the Simon Broome

Register of Familial Hyperlipidaemia began in 1980.Patients were registered by the participating lipidclinics, to which they had been referred by eithergeneral practitioners or hospital specialists. Threecategories of patients were admitted to the register:those defined as having definite familial hypercholes-terolaemia, whose families contained at least onemember in whom tendon xanthomas were present;those with possible familial hypercholesterolaemia,whose families had no member affected with tendonxanthomas; and a much smaller group with severehypertriglyceridaemia. We report on patients whowere classified as having definite familial hyper-cholesterolaemia. This was defined as a total cholesterolconcentration above 7-5 mmol/I or, when available, alow density lipoprotein cholesterol concentration

BMJ VOLUME 303 12 OCTOBER 1991 893



ww

.bmj.com

/B


j.303.6807.884 on 12 October 1991. D

ownloaded from

http://www.bmj.com/

chemotherapy in advanced ovarian overview of randomised ...conclusions-in the past, randomised...

Documents