Nonproteinuric VersusProteinuricPhenotypes inDiabeticKidney Disease: A PropensityScore–Matched Analysis of aNationwide, Biopsy-Based CohortStudyDiabetes Care 2019;42:891–902 | https://doi.org/10.2337/dc18-1320

OBJECTIVE

Clinicopathological characteristics, renal prognosis, and mortality in patients withtype 2 diabetes and reduced renal function without overt proteinuria are scarce.

RESEARCH DESIGN AND METHODS

We retrospectively assessed 526 patients with type 2 diabetes and reduced renalfunction (estimated glomerular filtration rate [eGFR] <60 mL/min/1.73 m2),who underwent clinical renal biopsy and had follow-up data, from Japan’s nation-wide multicenter renal biopsy registry. For comparative analyses, we derived one-to-two cohorts of those without proteinuria versus those with proteinuria usingpropensity score–matching methods addressing the imbalances of age, sex,diabetes duration, and baseline eGFR. The primary end point was progressionof chronic kidney disease (CKD) defined as new-onset end-stage renal disease,decreaseofeGFRby‡50%,ordoublingof serumcreatinine. Thesecondaryendpointwas all-cause mortality.

RESULTS

Eighty-two patients with nonproteinuria (urine albumin-to-creatinine ratio [UACR]<300 mg/g) had lower systolic blood pressure and less severe pathological lesionscompared with 164 propensity score–matched patients with proteinuria (UACR‡300 mg/g). After a median follow-up of 1.9 years (interquartile range 0.9–5.0 years) from the date of renal biopsy, the 5-year CKD progression-free survivalwas 86.6% (95% CI 72.5–93.8) for the nonproteinuric group and 30.3% (95% CI22.4–38.6) for the proteinuric group (log-rank test P < 0.001). The lower renal riskwas consistent across all subgroup analyses. The all-causemortality was also lowerin the nonproteinuric group (log-rank test P 5 0.005).

CONCLUSIONS

Patients with nonproteinuric diabetic kidney disease had better-controlled bloodpressure and fewer typical morphological changes and were at lower risk of CKDprogression and all-cause mortality.

1Department of Nephrology and LaboratoryMedicine, Faculty of Medicine, Institute of Med-ical, Pharmaceutical and Health Sciences, GraduateSchool of Medical Sciences, Kanazawa University,Ishikawa, Japan2NephrologyCenter, ToranomonHospital, Tokyo,Japan3Nephrology Center, Toranomon Hospital Kaji-gaya, Kanagawa, Japan4Okinaka Memorial Institute for Medical Re-search, Tokyo, Japan5Division of Nephrology, Kanazawa UniversityHospital, Kanazawa, Japan6Department of Pathology, Toranomon Hospital,Tokyo, Japan7Department of Pathology, Yokohama City Univer-sity Graduate School ofMedicine, Kanagawa, Japan8Department of Nephrology, Fujita Health Uni-versity School of Medicine, Aichi, Japan9Department of Pathology, Clinical ResearchCenter, National Hospital Organization Chiba-East National Hospital, Chiba, Japan10Health Administration Center, Niigata Univer-sity, Niigata, Japan11ClinicalPharmacologyandTherapeutics, TohokuUniversity Graduate School of PharmaceuticalSciences, Miyagi, Japan12Department of Pathology, Faculty of Medicine,Fukuoka University, Fukuoka, Japan13Department of Pathology, Dokkyo MedicalUniversity Saitama Medical Center, Saitama,Japan14DivisionofNephrologyandKidneyCenter, KobeUniversity Graduate School of Medicine, Hyogo,Japan15Department ofNephrology, KanazawaMedicalUniversity School of Medicine, Ishikawa, Japan16Department of Nephrology, Nagasaki Univer-sity Hospital, Nagasaki, Japan

Masayuki Yamanouchi,1,2,3,4

Kengo Furuichi,5 Junichi Hoshino,2,4

Tadashi Toyama,5 Akinori Hara,5

Miho Shimizu,5 Keiichi Kinowaki,6

Takeshi Fujii,6 Kenichi Ohashi,6,7

Yukio Yuzawa,8 Hiroshi Kitamura,9

Yoshiki Suzuki,10 Hiroshi Sato,11

Noriko Uesugi,12 Satoshi Hisano,12

Yoshihiko Ueda,13 Shinichi Nishi,14

Hitoshi Yokoyama,15 Tomoya Nishino,16

Kenichi Samejima,17 Kentaro Kohagura,18

Yugo Shibagaki,19 Koki Mise,20

Hirofumi Makino,20 Seiichi Matsuo,21

Yoshifumi Ubara,3,4 Takashi Wada,1,5

Research Group of Diabetic Nephropathy,

theMinistry of Health, Labour andWelfare,

and the Japan Agency forMedical Research

and Development

Diabetes Care Volume 42, May 2019 891

EPIDEM

IOLO

GY/H

EALTH

SERVICES

RESEA

RCH

Mainly based on the analysis of the datafrompatientswith type 1 diabetes, in theclinical course of diabetic kidney diseaseit has long been considered that an in-crease of albuminuria, from normoalb-uminuria (urine albumin-to-creatinineratio ratio [UACR],30 mg/g) to micro-albuminuria (UACR 30–299 mg/g) tomacroalbuminuria (UACR $300 mg/g),precedes the progression of renaldecline (defined as estimated glomer-ular filtration rate [eGFR] ,60 mL/min/1.73 m2) (1–3). Morphological changesknown as nodular glomerular sclerosis(Kimmelstiel-Wilson nodule) have alsobeen observed in patients with diabetesand loss of renal function (4,5). There-fore, patients with diabetes and reducedrenal function are deemed to have overtproteinuria with nodular glomerularsclerosis. Recently, however, cumulativeevidence from several cross-sectionalstudies revealed that a proportionof patients with type 2 diabetes de-velop progression of renal decline with-out proteinuria (macroalbuminuria) oreven without microalbuminuria, sug-gesting the existence of a nonprotein-uric phenotype of diabetic kidneydisease defined as eGFR,60 mL/min/1.73 m2 and UACR ,300 mg/g (6–11).Despite increasing attention, few clini-cal trials and longitudinal studies intype 2 diabetes include individualswithout proteinuria or individuals withbiopsy-proven diabetic kidney dis-ease, and therefore their clinicopath-ological characteristics, renal prognosis,and all-cause mortality are very lim-ited.Similar to the U.S. and most countries

in Europe, Japan has been suffering fromthe expanding trend in the continuedincrease of the prevalence of diabetickidney disease that leads to end-stagerenal disease (ESRD) and high mortality(12–15). Commissioned by the Ministryof Health, Labour and Welfare and the

Japan Agency for Medical Research andDevelopment with a goal of better un-derstanding and halting the pandemic ofdiabetic kidney disease, we established anationwide biopsy-based cohort of di-abetic kidney disease with followed-updata, including ESRD and death ascer-tainment. Using this nationwide cohortand propensity score–matching meth-ods, we aimed to investigate clinicopath-ological characteristics, renal prognosis,and mortality in patients with the non-proteinuric phenotype of diabetic kidneydisease compared with patients with theclassical proteinuric phenotype of dia-betic kidney disease.

RESEARCH DESIGN AND METHODS

Study Design and PopulationThis is a retrospective study of patientswho underwent clinical renal biopsyperformed from1 January 1985 to 31De-cember 2016 and had a pathologicaldiagnosis of diabetic kidney disease atthe following 18 hospitals in Japan: Tora-nomon Hospital (Tokyo, Japan), Torano-mon Hospital Kajigaya (Kanagawa,Japan), Kanazawa University Hospital(Ishikawa, Japan), Fujita Health Univer-sity Hospital (Aichi, Japan), National Hos-pital Organization Chiba-East NationalHospital (Chiba, Japan), Niigata Univer-sity Hospital (Niigata, Japan), TohokuUniversity Hospital (Miyagi, Japan), Uni-versity of Tsukuba Hospital (Ibaraki,Japan), Fukuoka University Hospital(Fukuoka, Japan), Dokkyo Medical Uni-versity SaitamaMedical Center (Saitama,Japan), Kobe University Hospital (Hyogo,Japan), Kanazawa Medical UniversityHospital (Ishikawa, Japan), NagasakiUniversity Hospital (Nagasaki, Japan),Nara Medical University (Nara, Japan),University of the Ryukyus Hospital(Okinawa, Japan), Okayama UniversityHospital (Okayama, Japan), St. MariannaUniversity School of Medicine Hos-pital (Kanagawa, Japan), and Nagoya

University Hospital (Nagoya, Japan). Theindications for biopsy were 1) renal im-pairment and 2) urinary abnormalities,such as albuminuria, proteinuria, hema-turia, or casts. The majority of patientswere under the care of the above-mentioned main hospitals or satelliteclinics, and they were followed until31 December 2017. Patients were ex-cluded if they had a diagnosis of con-comitant renal disease with diabetickidney disease, if they underwentprotocol renal transplant biopsy, or ifthey were followed for ,3 months. Thisstudy was approved by the institutionalreview boards of Toranomon Hospi-tal, Toranomon Hospital Kajigaya, andKanazawa University Hospital.

Clinical Characteristics, LaboratoryData, and Pathological ClassificationClinical characteristics at the time ofbiopsy were ascertained from the med-ical records, including the age, sex, BMI,existence of diabetic retinopathy, smok-ing status, use of renin-angiotensin-aldosterone system (RAAS) blockade, useof glucose-lowering agents, use of sta-tins, use of erythropoietin-stimulatingagents, systolic blood pressure, and di-astolic blood pressure. With the reviewof legal annual health checkup, the on-set of diabetes was defined as the datewhen patients first met one of the fol-lowing criteria: a fasting plasma glucose$126 mg/dL (7.0 mmol/L), a randomplasma glucose$200 mg/dL (11.1 mmol/L),an HbA1c $6.5% (48 mmol/mol), or theuse of glucose-lowering agents. Type 2diabetes was defined as having diabetesonset after age 30 years and not takinginsulin at the baseline visit in our hos-pitals. Diabetes duration was definedas the time from the date of diabetesonset to the date of renal biopsy. Smok-ing status was determined at the timeof biopsy. The existence of diabeticretinopathy was defined as having

17Department of Nephrology, Nara Medical Uni-versity, Nara, Japan18Department of Cardiovascular Medicine, Ne-phrology and Neurology, University of the Ryu-kyus School of Medicine, Okinawa, Japan19Division of Nephrology, Department of InternalMedicine, St. Marianna University School ofMedicine, Kanagawa, Japan20Department of Nephrology, Rheumatology,Endocrinology and Metabolism, Okayama Uni-versity Graduate School of Medicine, Dentistryand Pharmaceutical Sciences, Okayama, Japan

21Division of Nephrology, Department of InternalMedicine, Nagoya University Graduate School ofMedicine, Nagoya, Japan

Corresponding authors: Masayuki Yamanouchi,m.yamanouchi@toranomon.gr.jp, and TakashiWada, twada@m-kanazawa.jp

Received 19 June 2018 and accepted 4 February2019

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc18-1320/-/DC1.

© 2019 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More infor-mation is available at http://www.diabetesjournals.org/content/license.

892 Nonproteinuric Diabetic Kidney Disease Diabetes Care Volume 42, May 2019

microaneurysms, retinaldotandblothem-orrhage, or neovascularization in the retina.Data obtained closest to the date of biopsywere used for this study.Laboratory data at the time of biopsy

were also obtained from the medicalrecords, including hemoglobin, HbA1c, to-tal cholesterol, triglycerides, LDL choles-terol, HDL cholesterol, uric acid, serumcreatinine, eGFR (calculated by theMDRDstudy equation for Japanese [16]), andUACR. Nonproteinuric diabetic kidneydisease was defined as having an eGFR,60 mL/min/1.73 m2 with a UACR,300 mg/g at the time of renal biopsy.Renal biopsy specimens were pro-

cessed for light microscopy, immunoflu-orescence, and electron microscopy. Allbiopsies were evaluated by three path-ologists. If all three pathologists did notreach the same conclusion, discussionwas held until consensus was reached.Patients with at least a 5-year durationof diabetes and proven glomerular base-ment membrane thickening on electronmicroscopy (glomerular basementmem-brane .430 nm in men or .395 nm inwomen [13]) were classified as havingdiabetic kidney disease, and other path-ological findings were evaluated accord-ing to the following three classifications:the classification developed by Fiorettoet al. (17), the Pathologic Classification ofDiabetic Nephropathy developed by Ter-vaert et al. (18) on behalf of the RenalPathology Society (RPS), and the Japa-nese classification of diabetic nephrop-athy (19).For the Fioretto classification, diabetic

kidney disease is classified as follows:category I, normal or near-normal struc-turewith verymildmesangial expansion,tubulointerstitial or arteriolar changes,or arteriolar hyalinosis in any combina-tion; category II, typical diabetic kidneydisease featured by established diabeticlesions with balanced severity of glomer-ular, tubulointerstitial, and arteriolarchanges; and category III, atypical pat-ternsof renal injury featuredbyabsent oronly mild glomerular diabetic changeswith disproportionately severe intersti-tial or arteriolar lesions. For the Tervaert(RPS) classification, diabetic kidney dis-ease was classified as follows: class I,glomerular basement membrane thick-ening and only mild, nonspecific changeson light microscopy; class II, mild (IIa) orsevere (IIb)mesangial expansionwithouteither nodular lesions or global sclerosis

in .50% of the glomeruli; class III, nod-ular lesions without global sclerosisin .50% of the glomeruli; and classIV, global sclerosis in .50% of theglomeruli. Other pathological findingsevaluated were interstitial lesions (inter-stitial fibrosis and tubular atrophy [IFTA][grades 0–3] and interstitial inflamma-tion [grades 0–2]) and vascular lesions(arteriolar hyalinosis [grades 0–2] andarteriosclerosis [grades 0–2]). Patientswere excluded if histological and sero-logical data confirmed concomitant renaldisease or if the biopsy was inadequatefor diagnosis.

For the Japanese classification of di-abetic nephropathy, diabetic nephropa-thy was evaluated on the basis of nineglomerular lesions (diffuse lesion [grades0–3]; nodular lesion [grades 0–1]; sub-endothelial spacewidening [grades 0–3];exudative lesion [grades 0–1]; mesan-giolysis/microaneurysm [grades 0–1];perihilar neovascularization [grades 0–1]; global glomerulosclerosis, collapsingglomerulopathy, and ischemic nephrop-athy [grades 0–1]; segmental sclerosis[grades 0–1]; and glomerulomegaly[grades 0–1]), two interstitial lesions(IFTA [grades 0–3] and interstitial cellinfiltration [grades 0–3]), and two vas-cular lesions (arteriolar hyalinosis [grades0–3] and intimal thickening [grades0–3]). We also evaluated the percentglomerular sclerosis as defined by thenumber of total global and segmentalsclerotic glomeruli per the total numberof glomeruli.

Outcome Measures and Follow-UpThe primary outcome of this study waschronic kidney disease (CKD) progres-sion, which was defined as new-onsetESRD, decrease of eGFR by $50%, ordoubling of the serum creatinine level.ESRD was defined as initiation of hemo-dialysis/peritoneal dialysis, renal trans-plantation, or death as a result of uremia,and occurrence of ESRD was ascertainedby reviewing the database of the Japa-nese Society for Dialysis Therapy (JSDT).Since 1968, the JSDT has kept a completeannual renal data registry (JSDT RenalData registry) that covers patients ondialysis (20). If ESRD did not developduring follow-up, we evaluated whethertherewasadecreaseof eGFRby$50%ordoubling of serum creatinine. The sec-ondary outcome of this study was all-cause mortality. The event of death was

ascertained from the medical records.Patients who did not reach the outcomeof interest or who were lost to follow-upwere censored at their last follow-upvisit.

Statistical AnalysesThe study cohort was not from a clinicaltrial but from a real-world clinical prac-tice; therefore, the sample size was de-termined by the number of patients whounderwent a renal biopsy in real clinicalsettings. To address the imbalance of theeffects of age, sex, known duration ofdiabetes, and baseline renal function onthe outcomes of interest and to fairlycompare clinical, laboratory, patholog-ical profiles, and outcomes betweennonproteinuric and proteinuric diabetickidney disease, we matched the non-proteinuric diabetic kidney diseasegroup with a proteinuric diabetic kidneydisease group using propensity scoreswith a one-to-two nearest neighbor cal-iper width of 0.01 (maximum allowabledifference in propensity scores). We cal-culated the propensity score for patientswith nonproteinuric diabetic kidney dis-ease and patients with proteinuric di-abetic kidney disease using a logisticregression model to estimate the prob-ability of the disease assignment onthe basis of baseline variables such asage, sex, known duration of diabetes,and eGFR. In both the baseline andthe propensity score–matched cohorts,differences in clinical, laboratory, andpathological profiles between nonprotein-uric and proteinuric diabetic kidney dis-ease were analyzed by Student t testor the Wilcoxon test for continuousvariables, whereas the x2 test or Fisherexact test was used for categoricalvariables.

Variables independently associatedwith nonproteinuric diabetic kidney dis-ease were identified by multivariablelogistic regression in the overall cohort.Variables that had missing values .20%were not used in the analysis.

In both the entire cohort and thepropensity score–matched cohort, weconstructed Kaplan-Meier curves fortime-to-event end points (CKD progres-sion and all-cause mortality), taking time0 as the date of renal biopsy. We used alog-rank test to determine a difference insurvival rate.

The incidence of CKD progression wasthen calculated as the number of patients

care.diabetesjournals.org Yamanouchi and Associates 893

who developed new-onset ESRD, de-crease of eGFR by $50%, or doublingof the serum creatinine level during thefollow-up period divided by the totalfollow-up in person-years. The risk ofCKD progression for patients with non-proteinuric diabetic kidney disease com-pared with those with proteinuric diabetickidney disease was estimated as ad-justed hazard ratios (HRs) with 95%CIs by using the multivariable Cox pro-portional hazards model adjusted forage, sex, known duration of diabetes,and baseline eGFR. We conducted aseparate interim analysis with updatedvariables at the 1-year and 3-year studypoints, comparing the risk of CKD pro-gression between nonproteinuric andproteinuric groups that were also up-dated.Results are expressed as the mean

with SD or the median with interquartilerange (IQR) for continuous data and aspercentages for categorical data. Statis-tical tests were considered significant atP, 0.05 (two-sided). All statistical anal-yses were conducted using Stata 14.1software (StataCorp, CollegeStation, TX).

RESULTS

Between 1 January 1985 and 31 Decem-ber2016, 895patients underwent clinicalrenal biopsy and had a pathological di-agnosis of diabetic kidney disease in ourcohort; 13 patients were excluded be-cause of concomitant renal diseasewith diabetic kidney disease, protocolrenal transplant biopsy, or ,3 monthsof follow-up. We identified 526 who hadan eGFR ,60 mL/min/1.73 m2 at thetime of biopsy. Among them, 88 hadnonproteinuric diabetic kidney disease(UACR ,300 mg/g), and 438 had pro-teinuric diabetic kidney disease (UACR$300mg/g) at baseline. After propensityscore matching, the nonproteinuric di-abetic kidney disease group comprised82 patients and the proteinuric diabetickidney disease group comprised 164 pa-tients (Supplementary Fig. 1).Table 1 shows the clinical and path-

ological characteristics of the studygroups before and after propensity scorematching. In propensity score–matchedcohorts, the blood pressure in patientswith nonproteinuric diabetic kidney dis-ease was better controlled comparedwith patients with proteinuric diabetickidney disease, although patients with

nonproteinuric diabetic kidney diseasewere less prescribed RAAS blockade.Patients with nonproteinuric diabetickidney disease had lower total choles-terol levels and higher hemoglobinlevels. For pathological characteristics,there was a difference in classificationassignment for diabetic kidney diseasebetween the nonproteinuric diabetickidney disease group and proteinuricdiabetic kidney disease group. For exam-ple, with the Fioretto classification, cat-egory I, which is normal or near-normalrenal structure, was the most prevalentgroup (62%) in the nonproteinuric di-abetic kidney disease group, whereascategory II, which is typical diabetic kid-neydiseasewith balanced interstitial andarteriolar changes, was the most preva-lent group (66%) in the proteinuric di-abetic kidney disease group. Similar tothe Fioretto classification, in the Ter-vaert classification, advanced diabetickidney disease (class III or more) wasobserved in only 27% of patients withnonproteinuric diabetic kidney disease,whereas it was seen in 62% of patientswith proteinuric diabetic kidney disease.Compared with the proteinuric diabetickidney disease group, the nonproteinuricdiabetic kidney disease group had lesssevere interstitial and vascular lesions.

Table 2 shows the association of base-line characteristics with nonproteinuricdiabetic kidney disease in the overallcohort. In a multivariable logistic re-gression model, older age, lower systolicblood pressure, higher hemoglobin level,and higher HbA1c were significantly as-sociated with a higher odds of nonpro-teinuric diabetic kidney disease.

Table 3 shows predictors for CKD pro-gression in patients with nonproteinuricdiabetic kidney disease. Coxproportionalhazards analysis revealed low hemoglobinlevel, use of erythropoietin-stimulatingagents, and severity of IFTA as the pre-dictors of CKD progression in univariableanalysis. The multivariable analysis re-vealed that only severe IFTA was asso-ciated with CKD progression.

Figure 1 shows the Kaplan-Meiercurves for time-to-event end points inboth the entire cohort and the matchedcohort, taking time 0 as the date of renalbiopsy. After a median follow-up of 1.8years (IQR 0.9–3.7) from the date ofrenal biopsy, 297 (56%) of the 526 pa-tients had renal events. The 5-year CKDprogression-free survival was 33.2% (95%

CI 28.4–38.2%) for all patients, 86.9%(95% CI 73.1–93.9%) for the nonpro-teinuric diabetic kidney disease group,and 24.5% (95% CI 19.8–29.5%) for theproteinuric diabetic kidneydiseasegroup(log-rank test P , 0.001) (Fig. 1A). Thesame trend was seen in the propensityscore–matched cohort: After a medianfollow-up of 1.9 years (IQR 0.9–5.0) fromthe date of renal biopsy, 124 (50%) of the246 matched patients had renal events.The 5-year CKD progression-free survivalwas 46.4% (95% CI 38.7–53.6%) for allpatients, 86.6% (95% CI 72.5–93.8%) forthe nonproteinuric diabetic kidney dis-ease group, and 30.3% (95% CI 22.4–38.6%) for the proteinuric diabetickidney disease group (log-rank test P, 0.001) (Fig. 1B). Similarly, for thesecondary outcome (all-causemortality),after amedian follow-upof 2.7 years (IQR1.1–5.7) from the date of renal biopsy,55 (10%) of the 526 patients had deathevents. The 5-year death-free survivalwas 89.7% (95% CI 85.6–92.7%) for allpatients, 98.4% (95% CI 89.1–99.8%) forthe nonproteinuric diabetic kidney dis-ease group, and 87.5% (95% CI 82.5–91.2%) for the proteinuric diabetic kid-ney disease group (log-rank test P ,0.001) (Fig. 1C). The same trend wasseen in the propensity matched cohort:Afteramedian follow-upof3.1years (IQR1.3–7.0) from the date of renal biopsy,35 (14%) of the 246 matched patientshad death events. The 5-year death-freesurvival was 88.2% (95% CI 82.0–92.3%)for all patients, 98.3% (95% CI 88.7–99.8%) for the nonproteinuric diabetickidney disease group, and 82.6% (95% CI73.6–88.8%) for the proteinuric diabetickidney disease group (log-rank test P =0.005) (Fig. 1D).

Table 4 displays the comparison ofCKD progression risk in the nonproteinuricdiabetic kidney disease group and the pro-teinuricdiabetickidneydiseasegroupin thepropensity score–matched cohort. Theoverall CKD progression incidence wassignificantly lower in the nonproteinuricdiabetic kidney disease group (30 [95%CI18–50] per 1,000 person-years) than inthe proteinuric diabetic kidney diseasegroup (231 [95% CI 191–278] per 1,000person-years; crude HR 0.15 [95% CI0.08–0.26]). After adjustment for age,sex, known duration of diabetes, andbaseline eGFR, the risk of CKD progres-sion remained lower in the nonprotein-uric diabetic kidney disease cohort than

894 Nonproteinuric Diabetic Kidney Disease Diabetes Care Volume 42, May 2019

Table 1—Baseline clinical and pathological characteristics before and after propensity score matching

Entire cohort Propensity score–matched cohort

CharacteristicNonproteinuric

(n = 88)Proteinuric(n = 438) P value

Nonproteinuric(n = 82)

Proteinuric(n = 164) P value

Clinical characteristics at biopsyAge (years) 63 (57, 68) 61 (52, 68) 0.081 63 (56, 67) 64 (56, 70) 0.52Male 61 74 0.016 66 68 0.68BMI (kg/m2) 23 (21, 25) 24 (22, 26) 0.033 23 (21, 25) 24 (22, 26) 0.098Diabetes duration (years) 12 (8, 18) 13 (9, 20) 0.36 12 (8, 18) 13 (8, 21) 0.45Diabetic retinopathy 62 74 0.11 62 69 0.44Smoking 63 63 1.00 63 61 0.90RAAS blockade 48 67 0.017 48 69 0.015Glucose-lowering agents 93 86 0.28 93 90 0.57Statins 31 28 0.73 31 20 0.21Erythropoietin-stimulating agents 17 12 0.51 17 13 0.65sBP (mmHg) 132 (120, 146) 148 (137, 164) ,0.001 130 (120, 145) 146 (134, 162) ,0.001dBP (mmHg) 76 (68, 80) 80 (71, 90) 0.003 75 (68, 80) 80 (70, 90) 0.009Hemoglobin (g/dL) 12 (11, 14) 11 (10, 13) ,0.001 12 (11, 14) 11 (10, 13) 0.002HbA1c (mmol/mol) 55.2 (47.5, 74.9) 50.8 (41.0, 65.0) 0.004 55.2 (47.5, 74.9) 51.9 (42.1, 67.2) 0.033HbA1c (%) 7.2 (6.5, 9.0) 6.8 (5.9, 8.1) 0.004 7.2 (6.5, 9.0) 6.9 (6.0, 8.3) 0.033Total cholesterol (mmol/L) 4.8 (4.0, 5.8) 5.4 (4.4, 6.5) ,0.001 5.0 (3.9, 5.8) 5.4 (4.6, 6.4) 0.002Triglycerides (mmol/L) 1.5 (1.1, 2.2) 1.8 (1.2, 2.4) 0.11 1.5 (1.1, 2.2) 1.7 (1.2, 2.4) 0.21LDL-C (mmol/L) 2.8 (2.1, 3.4) 3.1 (2.5, 4.2) 0.081 2.8 (2.1, 3.4) 3.3 (2.6, 4.1) 0.033HDL-C (mmol/L) 1.0 (0.8, 1.3) 1.1 (0.9, 1.3) 0.25 1.0 (0.8, 1.3) 1.1 (0.8, 1.5) 0.28Uric acid (mg/dL) 6.8 (5.9, 7.5) 6.8 (5.9, 7.0) 0.33 6.8 (5.9, 7.5) 6.5 (5.7, 7.8) 0.90eGFR (mL/min/1.73 m2) 45 (33, 54) 36 (24, 46) ,0.001 45 (33, 54) 44 (29, 50) 0.12UACR (mg/g) 110 (40, 210) 2,200 (1,100, 4,000) ,0.001 100 (30, 180) 2,100 (1,140, 3,570)Albuminuria status* (n)Normoalbuminuria 19 0 19 0Microalbuminuria 69 0 63 0Macroalbuminuria 0 438 0 164

Hematuria 10 12 0.54 10 7 0.57

Pathological characteristics at biopsyFioretto classification† ,0.001 ,0.001Category I 60 17 62 17Category II 23 70 20 66Category III 17 13 18 17

Tervaert (RPS) classification‡ ,0.001 ,0.001Class I 30 5 31 4Class IIa 32 13 22 14Class IIb 9 21 10 20Class III 27 51 25 52Class IV 2 10 2 10

Japanese classificationGlomerular lesions%GS 20 (9, 54) 39 (22, 67) 0.007 16 (6, 37) 33 (17, 44) ,0.001Diffuse lesions ,0.001 ,0.001

0 16 3 16 11 41 14 43 172 19 25 17 293 24 58 24 53

GBM doubling ,0.001 0.0010 65 23 66 231 18 41 17 412 7 21 8 233 10 15 9 13

Exudative lesion 24 60 ,0.001 24 61 ,0.001Nodular lesion 25 53 ,0.001 22 54 ,0.001Mesangiolysis 19 49 ,0.001 19 49 ,0.001Polar vasculosis 56 76 0.005 54 73 0.014Glomerulomegaly 25 40 0.033 26 37 0.13

Interstitial lesionsIFTA ,0.001 ,0.001

0 10 2 11 21 51 19 53 24

Continued on p. 896

care.diabetesjournals.org Yamanouchi and Associates 895

in the proteinuric diabetic kidney diseasecohort (adjusted HR 0.13 [95% CI 0.08–0.24]). The risk of CKD progression wasconsistently lower in the nonproteinuricdiabetic kidney disease group than in theproteinuric diabetic kidneydiseasegroupwhen stratified by potential confounderssuch as age, sex, obesity, retinopathy,smoking status, use of RAAS blockade,hypertension, dyslipidemia, poor gly-cemic control, lower eGFR, and patho-logical findings.

CONCLUSIONS

This propensity score–matched cohortstudy of patients with nonproteinuricdiabetic kidney disease and proteinuricdiabetic kidney disease demonstratesthat patients with nonproteinuric dia-betic kidney disease have lower systolicblood pressure and less frequent typi-cal pathological lesions and are at alower risk of CKD progression and all-cause mortality. From a clinical point ofview, analysis of data mainly from pa-tients with type 1 diabetes, it has beenconsidered that the increase of albumin-uria to macroalbuminuria or proteinuriareflects glomerular abnormalities andprecedes and accompanies the decline

of renal function (1–3). Recently, how-ever, a number of cross-sectional studiesdispelled this widely held belief, report-ing that a proportion of patients withtype 2 diabetes and CKD present with anormal range of albuminuria or onlymicroalbuminuria, even in their latestages of CKD (defined as eGFR ,60mL/min/1.73 m2) (6–11). However, theclinical pictures for patients with non-proteinuria are inconsistent in these pre-vious reports, although several features,such as female sex, hypertension, smok-ing, hyperglycemia, no evidence of mi-croangiopathy (represented as diabeticretinopathy), and the use of RAAS block-ade, were reported as risk factors fornonproteinuric diabetic kidney disease.These inconsistent findings may arisefrom the fact that patients with type 2diabetes and CKD were clinically diag-nosed as having a diabetic kidney dis-ease. Therefore, it is possible that thesepatients have glomerular diseases otherthan diabetic kidney disease or concom-itant renal disease with diabetic kidneydisease. We therefore investigated a co-hort with biopsy-proven diabetic kidneydisease as the only glomerular diseasediagnosis rather than inaccurate clinicaldiagnoses of diabetic kidney disease. We

found that older age, lower systolic bloodpressure, higher hemoglobin, and higherHbA1c are associatedwithnonproteinuricdiabetic kidneydisease. Furthermore,weconducted propensity score matching toaddress the imbalance of backgroundfactors such as age, sex, diabetes dura-tion, and baseline eGFR that affect re-nal prognosis and mortality. We foundthat patients with nonproteinuric dia-betic kidney disease had better-controlledblood pressure, although they were lessfrequently prescribed RAAS blockade.These results are plausible because anumber of studies reported that highersystolic blood pressure is associatedwithan increase in albuminuria (21–23). Theeffect of RAAS blockade on albuminuriaand blood pressure should bementionedsince previous studies report that RAASblockade decreases the amount of albu-minuria, and these effects are inde-pendent from changes in systemicblood pressure (21–23). Thus, we didsubgroupanalyses to test for the effect ofRAAS blockade and having hypertensionon renal outcome but found that thelower renal risk in patients with non-proteinuric diabetic kidney diseasewas consistent across these subgroupanalyses.

Table 1—Continued

Entire cohort Propensity score–matched cohort

CharacteristicNonproteinuric

(n = 88)Proteinuric(n = 438) P value

Nonproteinuric(n = 82)

Proteinuric(n = 164) P value

2 25 36 23 373 14 43 13 37

Inflammation 0.005 0.0210 14 4 15 41 61 60 62 642 19 24 18 223 6 12 5 10

Vascular lesionsArteriolar hyalinosis ,0.001 0.0020 14 3 15 41 24 14 23 162 29 42 29 483 33 41 33 32

Arteriosclerosis 0.011 0.0020 15 7 16 51 35 47 35 512 47 45 47 443 3 1 2 0

Data are median (25th, 75th percentile) or percentage unless otherwise indicated. dBP, diastolic blood pressure; GBM, glomerular basementmembrane; %GS, percent glomerular sclerosis defined as the number of global or segmental sclerosis glomeruli per total glomeruli; HDL-C, HDLcholesterol; LDL-C, LDL cholesterol; sBP, systolic blood pressure. *Albuminuria status: normoalbuminuria, UACR,30 mg/g; microalbuminuria, UACR30–299 mg/g; macroalbuminuria, UACR $300 mg/g. †Fioretto classification: category I, normal or near-normal renal structure; category II, typicaldiabetic kidney disease; category III, atypical patterns of renal injury. ‡Tervaert (RPS) classification: class I, mild or nonspecific lightmicroscopy changesand electronmicroscopy–proven GBM thickening; class IIa, mildmesangial expansion; class IIb, severemesangial expansion; class III, nodular sclerosis(Kimmelstiel-Wilson lesion); class IV, advanced diabetic glomerular sclerosis.

896 Nonproteinuric Diabetic Kidney Disease Diabetes Care Volume 42, May 2019

From a pathological perspective, theearly studies of morphological changeswith diabetes have been done mostly inpatients with type 1 diabetes (4,24–26).These studies showed specific lesionswith diabetes that include diffuse le-sions characterized by a thickened glo-merular basement membrane andmesangial expansion, nodular lesionscharacterized by nodular glomerular

sclerosis known as Kimmelstiel-Wilsonnodule, and hyalinosis lesions character-ized by an exudative/insudative lesionand fibrin cap. Especially, nodular glo-merular sclerosis is believed to be ahallmark of classical diabetic nephropa-thy and observed in patients with long-standing diabetes and reduced renalfunction. Since the recognition of mor-phological changes with diabetes, later

studies in type 1 diabetes have shownthat glomerular structure componentsare associated with measured GFR byiothalamate or iohexol or 24-h creatinineclearance (27,28). In these studies, thestructural-functional relationship be-came stronger in patients with reducedrenal function, although most of thosewith reduce renal function had protein-uria. Conversely, available biopsy-based

Table 2—Baseline characteristics associated with nonproteinuric diabetic kidney disease in the overall cohort (N = 526)

Univariable Multivariable

Variable OR (95% CI) P value OR (95% CI) P value

Clinical parameterAge (years) 1.02 (1.00–1.04) 0.06 1.05 (1.01–1.08) 0.004Male (yes/no) 0.56 (0.35–0.90) 0.02BMI (kg/m2) 0.91 (0.85–0.98) 0.01Diabetes duration (years) 0.98 (0.94–1.02) 0.31Diabetic retinopathy (yes/no) 0.56 (0.28–1.15) 0.12Smoking (yes/no) 1.00 (0.42–2.40) 0.99RAAS blockade (yes/no) 0.45 (0.23–0.88) 0.02Glucose-lowering agents (yes/no) 2.22 (0.51–9.72) 0.29Statins (yes/no) 1.16 (0.50–2.67) 0.73Erythropoietin-stimulating agents (yes/no) 1.47 (0.46–4.69) 0.52sBP (mmHg) 0.95 (0.94–0.97) ,0.001 0.95 (0.93–0.97) ,0.001dBP (mmHg) 0.97 (0.95–0.99) 0.002Hemoglobin (g/dL) 1.30 (1.16–1.46) ,0.001 1.22 (1.05–1.41) 0.010HbA1c (%) 1.22 (1.08–1.39) 0.002 1.32 (1.11–1.56) 0.002Total cholesterol (mmol/L) 0.99 (0.99–1.00) ,0.001Triglycerides (mmol/L) 1.00 (0.99–1.00) 0.26LDL-C (mmol/L) 0.99 (0.98–1.00) 0.078HDL-C (mmol/L) 0.97 (0.93–1.01) 0.16Uric acid (mg/dL) 0.89 (0.73–1.08) 0.23eGFR (mL/min/1.73 m2) 1.04 (1.02–1.06) ,0.001Hematuria (yes/no) 0.55 (0.27–1.16) 0.16

Histological parametersFioretto classification*Category I 1.00 (Reference)Category II 0.09 (0.05–0.16) ,0.001Category III 0.35 (0.18–0.68) 0.002

Tervaert classification (class $III)† 0.26 (0.15–0.43) ,0.001Japanese classificationGlomerular lesions%GS 0.97 (0.95–0.98) ,0.001Diffuse lesions ($3) 0.26 (0.14–0.50) ,0.001GBM doubling ($3) 1.75 (0.64–4.81) 0.28Exudative lesion (yes/no) 0.21 (0.11–0.39) ,0.001Nodular lesion (yes/no) 0.30 (0.16–0.55) ,0.001Mesangiolysis (yes/no) 0.25 (0.13–0.49) ,0.001Polar vasculosis (yes/no) 0.40 (0.23–0.71) 0.002Glomerulomegaly (yes/no) 0.52 (0.28–0.96) 0.035

Interstitial lesionsIFTA score ($3) 0.23 (0.12–0.43) ,0.001Interstitial inflammation score ($3) 0.49 (0.19–1.26) 0.14

Vascular lesionsArteriolar hyalinosis score ($3) 0.85 (0.51–1.41) 0.53Arteriosclerosis score ($3) 3.32 (0.78–14.2) 0.11

Odds ratios (ORs) for CKD progression were determined for various clinical and histological characteristics by the logistic model. dBP, diastolic bloodpressure; GBM, glomerular basement membrane; %GS, percent glomerular sclerosis defined as the number of global or segmental sclerosis glomeruliper total glomeruli; HDL-C, HDL cholesterol; LDL-C, LDL cholesterol; sBP, systolic blood pressure. *Fioretto classification: category I, normal or near-normal renal structure; category II, typicaldiabetic kidneydisease; category III, atypicalpatternsof renal injury.†Tervaert (RPS) classification: class I,mildor nonspecific light microscopy changes and electron microscopy–proven GBM thickening; class IIa, mild mesangial expansion; class IIb, severemesangial expansion; class III, nodular sclerosis (Kimmelstiel-Wilson lesion); class IV, advanced diabetic glomerular sclerosis.

care.diabetesjournals.org Yamanouchi and Associates 897

studies in type 2 diabetes, especiallywithout proteinuria, have been limitedsince diabetic kidney disease has beenclinically diagnosed in patients withtype 2 diabetes. Only a few studieshave analyzed the pathological lesionsin patients with type 2 diabetes withnonproteinuric diabetic kidney disease(29–31). The findings of these cross-sectional studies, however, are inconsistent.

One study reported that patients withnonproteinuric diabetic kidney diseasehave advanced diabetic glomerularchanges but well-preserved interstitialand arterial lesions. Other studies re-ported that patients with nonproteinuricdiabetic kidney disease have less typicaldiabetic kidney disease but dispropor-tionately damaged interstitial and arte-rial lesions. This inconsistency may arise

from the timing of the biopsy; for exam-ple, the pathology in those with eGFR55 mL/min/1.73 m2 may be differentfrom that in those with the same back-grounds but with eGFR 20 mL/min/1.73 m2, or age and sex may affect theirpathology. A number of studies reportedthat aging leads to various anatomicaland physiological changes of the kid-ney (32,33). Also, female sex through

Table 3—Predictors for CKD progression in patients with nonproteinuric diabetic kidney disease

Univariable Multivariable

Variable HR (95% CI) P value HR (95% CI) P value

Clinical parameterAge (years) 1.03 (0.97–1.09) 0.32Male (yes/no) 1.05 (0.36–3.02) 0.94BMI (kg/m2) 0.94 (0.77–1.15) 0.54Diabetes duration (years) 1.02 (0.93–1.12) 0.63Diabetic retinopathy (yes/no) 0.90 (0.21–3.80) 0.88Smoking (yes/no) 2.23 (0.23–21.7) 0.49RAAS blockade (yes/no) 0.74 (0.16–3.42) 0.70Glucose-lowering agents (yes/no) 0.60 (0.07–5.42) 0.65Statins (yes/no) 1.45 (0.13–16.2) 0.76Erythropoietin-stimulating agents (yes/no) 14.2 (1.44–141) 0.02sBP (mmHg) 1.00 (0.98–1.03) 0.85dBP (mmHg) 0.99 (0.95–1.04) 0.80Hemoglobin (g/dL) 0.63 (0.44–0.88) 0.007HbA1c (%) 0.87 (0.63–1.19) 0.38Total cholesterol (mmol/L) 0.98 (0.97–1.00) 0.008Triglycerides (mmol/L) 0.99 (0.97–1.01) 0.20LDL-C (mmol/L) 0.93 (0.88–0.99) 0.026HDL-C (mmol/L) 0.94 (0.84–1.05) 0.25Uric acid (mg/dL) 1.09 (0.62–1.94) 0.76eGFR (mL/min/1.73 m2) 0.96 (0.93–1.00) 0.039UACR (mg/g) 6.67 (0.42–10.6) 0.10

Histological parameterFioretto classification*Category I 1.00 (Reference)Category II 4.10 (1.26–13.4) 0.019Category III 2.33 (0.68–8.00) 0.18

Tervaert classification (class $III)† 2.4 (0.86–6.60) 0.095Japanese classificationGlomerular lesions

%GS 1.02 (1.00–1.05) 0.11Diffuse lesion ($3) 1.33 (0.34–5.22) 0.68GBM doubling ($3) 1.14 (0.19–6.95) 0.89Exudative lesion (yes/no) 0.75 (0.10–6.00) 0.79Nodular lesion (yes/no) 1.75 (0.51–6.05) 0.38Mesangiolysis (yes/no) 1.49 (0.32–6.90) 0.61Polar vasculosis (yes/no) 1.05 (0.31–3.62) 0.94Glomerulomegaly (yes/no) 1.56 (0.41–5.97) 0.51

Interstitial lesionsIFTA score ($3) 4.05 (1.34–12.2) 0.013 14.8 (3.0–73.8) 0.001Interstitial inflammation score ($3) 0.98 (0.22–4.37) 0.98

Vascular lesionsArteriolar hyalinosis score ($3) 0.85 (0.29–2.47) 0.77Arteriosclerosis score ($3) NA NA

HRs forCKDprogressionweredetermined for various clinical andhistological characteristicsby theCoxproportionalhazardsmodel. dBP,diastolicbloodpressure; GBM, glomerular basement membrane; %GS, percent glomerular sclerosis defined as the number of global or segmental sclerosis glomeruliper total glomeruli; HDL-C, HDL cholesterol; LDL-C, LDL cholesterol; NA, not applicable; sBP, systolic blood pressure. *Fioretto classification: category I,normal or near-normal renal structure; category II, typical diabetic kidney disease; category III, atypical patterns of renal injury. †Tervaert (RPS)classification: class I,mildornonspecific lightmicroscopychangesandelectronmicroscopy–provenGBMthickening; class IIa,mildmesangial expansion;class IIb, severe mesangial expansion; class III, nodular sclerosis (Kimmelstiel-Wilson lesion); class IV, advanced diabetic glomerular sclerosis.

898 Nonproteinuric Diabetic Kidney Disease Diabetes Care Volume 42, May 2019

hormonal changes is known as a riskfactor of glomerular sclerosis and inter-stitial fibrosis. Again, we investigateda propensity score–matched cohort ofbiopsy-proven diabetic kidney diseaseto address these issues.In this study, we used three different

classification systems of diabetic ne-phropathy to try to capture the differ-ences of pathological features betweennonproteinuric and proteinuric diabetickidney disease because each classifica-tion has its strengths and weaknesses.The Fioretto classification was developedas a basic distinction between typical andatypical diabetic nephropathy (17). Thisclassification evaluates not only glomer-ular lesions but also interstitial and ar-teriolar lesions and classifies these intothree categories based on the balance ofthe severity of glomerular, tubulointer-stitial, and arteriolar changes. This clas-sification enables us to see the overviewof pathological lesions. On the otherhand, the Tervaert (RPS) classificationfocuses purely on glomerular lesions,assigning classes I–V, which correspondto the severity of glomerular lesions (18).Themajor strength of this classification isthat it was proposed by the ResearchCommittee of the RPS as a platform for

clinical use. However, the shortcoming ofthis classification is that it puts lessemphasis on interstitial and arteriolarlesions and may not be able to cap-ture a broad spectrum of disease sever-ity. To overcome the shortcomings ofthe Tervaert (RPS) classification, theJapanese classification features detailedexplanations that are not evaluated inthe Tervaert classification and treatsglomerular, interstitial, and arteriolarlesions equally (19). However, it is notyet ready for clinical application. In anycase, with these classifications, we foundthat patients with nonproteinuric dia-betic kidney disease had less typicalpathological changes of diabetic kidneydisease or less severe diabetic changes.However, these results should be inter-preted with caution because the patho-logical classifications we used in ourstudy were not quantitative measuresbut all qualitative measures. Previousquantitative studies in both type 1 andtype 2 diabetes showed the structural-functional relationships independent ofalbuminuria, meaning that those withreduced renal function had more severestructural changes, even without albumin-uria (34–36). Renal samples in our studywere not evaluated with quantitative

measures of glomerular structure, suchas mesangial fractional volume, meanglomerular volume, glomerular filtra-tion surface density, and so on, all ofwhich are strongly associated with GFRdecline. Although the majority of thosewith nonproteinuria had only mild di-abetic lesions with the qualitativemeasures in our study, it is possiblethat with quantitative measures, theymay have had evident glomerular struc-tural changes that could explain theirreduced renal function.

In our study, we observed a decreasedrisk of CKD progression in patients withnonproteinuric diabetic kidney diseasecompared with those with proteinuricdiabetic kidney disease. We also ob-served, however, that a proportion ofpatients with nonproteinuric diabetickidney disease had CKD progression. Inthe absence of proteinuria, whatwill be arisk factor of CKD progression in thesepatients? Only in patients with nonpro-teinuric diabetic kidney disease, we per-formed univariable and multivariableCox proportional hazards analyses toinvestigate candidate predictors for pro-gression of CKD. Low hemoglobin level,use of erythropoietin-stimulating agents,and severity of IFTA were the only

Table 4—Comparison of CKD progression risk in the nonproteinuric diabetic kidney disease cohort and the proteinuric diabetickidney disease cohort

Nonproteinuric (n = 82) Proteinuric (n = 164)

Characteristic Events PY IR (95% CI) Events PY IR (95% CI) Adjusted HR (95% CI) P value

Overall 15 498 30 (18–50) 110 477 231 (191–278) 0.13 (0.08–0.24) ,0.001

Elderly 4 131 31 (11–81) 48 207 232 (175–308) 0.11 (0.04–0.31) ,0.001

Male 11 349 32 (17–57) 79 318 248 (199–309) 0.13 (0.07–0.26) ,0.001

Obesity 5 183 27 (11–66) 41 195 210 (155–285) 0.10 (0.04–0.26) ,0.001

Diabetic retinopathy 5 113 44 (18–107) 38 190 200 (146–275) 0.23 (0.09–0.61) ,0.001

Smoking 4 80 50 (19–134) 36 172 210 (151–291) 0.15 (0.05–0.45) ,0.001

RAAS blockade 4 77 52 (19–138) 54 227 238 (183–311) 0.13 (0.04–0.42) ,0.001

Hypertension 8 229 35 (17–70) 88 336 262 (213–323) 0.14 (0.06–0.28) ,0.001

Dyslipidemia 10 449 22 (12–41) 95 397 239 (196–293) 0.10 (0.05–0.20) ,0.001

Poor glycemic control 7 249 28 (13–59) 44 208 212 (158–284) 0.12 (0.05–0.29) ,0.001

eGFR ,45 mL/min/1.73 m2 6 336 18 (8–40) 48 306 157 (118–208) 0.13 (0.05–0.33) ,0.001

Fioretto classification*Category I 7 344 20 (9–43) 16 168 95 (58–156) 0.17 (0.06–0.46) ,0.001Category II 4 70 57 (22–153) 75 244 308 (245–386) 0.17 (0.06–0.47) ,0.001Category III 4 84 47 (18–126) 19 65 291 (186–457) 0.10 (0.03–0.36) ,0.001

Tervaert classification (class $III)† 6 128 47 (21–104) 66 197 335 (264–427) 0.14 (0.05–0.35) ,0.001

Events indicate number of patients who developed CKD progression. Adjusted HR was adjusted for age, sex, known duration of diabetes,and eGFR. Obesity was defined as BMI $25 kg/m2. Hypertension was defined as a systolic blood pressure $140 mmHg or diastolic bloodpressure $90 mmHg. Dyslipidemia was defined as total cholesterol level $240 mg/dL, LDL cholesterol level $140 mg/dL, triglyceridelevel$150 mg/dL, or HDL cholesterol level#40 mg/dL. Poor glycemic control was defined by an HbA1c .7.5%. IR, incidence rate; PY, person-years.*Fioretto classification: category I, normal or near-normal renal structure; category II, typical diabetic kidney disease; category III, atypicalpatterns of renal injury. †Tervaert (RPS) classification: class I, mild or nonspecific light microscopy changes and electron microscopy–provenglomerular basement membrane thickening; class IIa, mild mesangial expansion; class IIb, severe mesangial expansion; class III, nodularsclerosis (Kimmelstiel-Wilson lesion); class IV, advanced diabetic glomerular sclerosis.

care.diabetesjournals.org Yamanouchi and Associates 899

predictors of CKD progression in univari-able analysis. The multivariable analysisrevealed that only severe IFTA is associ-ated with CKD progression (Table 3).Although the number of patients withnonproteinuric diabetic kidney diseasewas limited in this analysis, these resultsimply that interstitial fibrosis, whichcauses anemia and higher chances ofbeing prescribed erythropoietin agents,is a key player in renal decline. Theseresults are similar to some previous re-ports that demonstrated that interstitialinjury plays a major role in the decline ofeGFR (37–39). In addition to interstitialfibrosis, quantitative glomerular struc-tural measurements may improve pre-dictive value. Again, the pathologicalclassifications we used in our analysisare qualitative classifications, whichare easy to implement but may notlink directly to renal function in the ab-sence of proteinuria. Quantitative struc-tural studies have shown more precisestructural-functional relationships and

predictive value independent of albu-minuria in patients with type 2 diabetes(35,36). We believe that future quanti-tative structural classifications of dia-betic kidney disease may be able tobetter predict CKD progression in pa-tients with type 2 diabetes and non-proteinuria.

Although our study suggests that thosewith nonproteinuria are at low risk forCKD progression, it is possible that pa-tients with nonproteinuria will becomeproteinuric or vice versa. We observedthat ,10% of patients with nonprotein-uria became proteinuric later and fewwith proteinuria became nonproteinuric.Most of the patients with nonproteinuriawhobecameproteinuric later had aUACRnear the upper limit of 300 mg/g atbaseline. To be sure that the findingsare consistent in updated variables, weconducted an interim analysis with up-dated variables at the 1-year and3-year study time points, comparing therisk of CKD progression between the

nonproteinuric and proteinuric groupsthat were also updated. In the fully ad-justed model with age, sex, eGFR, systolicblood pressure, HbA1c, and use of RAASblockade, the pattern of CKD progressionwas similar to that at baseline.

The strengths of our study are thenationwide study population in Japanover three decades, the use of biopsy-proven rather than inaccurate clinicaldiagnosis of diabetic kidney disease,the use of a longitudinal design ratherthan a cross-sectional design, the highlyreliable linkage-based ESRD ascertain-ment, and the use of propensity score–matching methods addressing theimbalance of confounders between non-proteinuric and proteinuric diabetickidney disease. All these design andmethodological elements enabled a ro-bust analysis of the comparison of clini-copathological backgrounds and the riskof CKD progression and mortality inpatients with nonproteinuric and protei-nuric diabetic kidney disease.

Figure 1—Renal event-free survival for the 526 patients in the entire cohort and the 164 patients in the propensity score–matched cohort. A: Kaplan-Meier curves of CKD progression-free survival in the entire cohort. B: Kaplan-Meier curves of CKD progression-free survival in the propensity score–matched cohort. C: Kaplan-Meier curves of death event–free survival in the entire cohort. D: Kaplan-Meier curves of death event–free survival in thepropensity score–matched cohort. CKD progression was defined as new-onset ESRD, decreased eGFR $50%, or doubling of serum creatinine.Nonproteinurics were defined as patients with an eGFR,60 mL/min/1.73 m2 without proteinuria (UACR,300 mg/g); proteinurics were defined aspatients with an eGFR ,60 mL/min/1.73 m2 and proteinuria (UACR $300 mg/g).

900 Nonproteinuric Diabetic Kidney Disease Diabetes Care Volume 42, May 2019

Several limitations of this study, how-ever, should be mentioned. First, as withobservational studies, especially withbiopsy-based cohort studies, there is apossibility of confounding by indication,since the selection of the study groupcould have been biased by nephrologistswho were interested in diabetic kidneydisease or the patients might have un-dergone biopsy because they were sus-pected to have other renal diseases. Justfor reference,;30% of our patients witha clinical diagnosis of diabetic kidneydisease underwent biopsy in our insti-tutions. Among 526patientswith biopsy-proven diabetic kidney disease, theprebiopsy clinical diagnoses were dia-betic kidney disease as the only causein 399 (76%), diabetic kidney disease plushematuria in 54 (10%), diabetic kidneydisease with rapid eGFR decline or ab-normal casts in 42 (8%), and diabetickidney diseasewith nephrotic-range pro-teinuria in 31 (6%). On the contrary,however, we believe that the accuratediagnosis of diabetic kidney diseaseprovides a clearer picture of clinico-pathological features of diabetic kidneydisease. Second, biopsy results for di-abetic kidney disease may vary by localsite biopsy policies (40). Among 18 hos-pitals, 5 pursued a restricted policy, and13 pursued an unrestricted policy. Notsurprisingly, all 88 patients with non-proteinuria were from the hospitalsthat adopt an unrestricted biopsy policy.However, therewere no statistical differ-ences of biopsy findings among adoptedbiopsy policies (x2 test P = 0.093). Third,the data of the previous renal eventsbefore the time of renal biopsy were notavailable to be accounted for in ouranalysis, which might affect and explainthe risk of nonproteinuric diabetic kidneydisease at the time of biopsy. However,all patients with nonproteinuric diabetickidney disease denied at least a history ofacute kidney injury or taking medications,such as nonsteroidal anti-inflammatorydrugs, that would affect their renal func-tion. Furthermore, to exclude anotherpotential explanation for having nonpro-teinuric diabetic kidney disease, most ofthese patients were confirmed to nothave renal stenosis by MRI or ultraso-nography. Fourth, the study popula-tion was limited to Japanese patientswith biopsy-confirmed diabetic kidneydisease; hence, our findings may not bewidely generalizable. We were unable

to compare our results directly with othercohorts; however, we found a study fromthe Chronic Renal Insufficiency Cohort(CRIC) that reported the CKD progressionrate for those with different baselinelevels of urine albumin (41). We did atrial calculation of the CKD progressionrate for those with nonproteinuria(UACR 30–299 mg/g) and proteinuria(UACR $300 mg/g): CKD progressionrates were 39 and213per 1,000person-years, respectively. Theprogression ratesin the CRIC study were very similar tothose in our study (CKD progression rateswere 30 and 194 per 1,000 person-yearsfor patients with nonproteinuria andproteinuria, respectively), despite thefact that patients in our cohort all hadbiopsy-proven rather than a clinical di-agnosis of diabetic kidney disease. Fifth,we do not have the data on UACR beforetaking RAAS blockade. It is possible thatthe nonproteinuric group mostly com-prised patients who responded well toRAASblockade, resulting in a lower risk ofCKD progression. However, we believethat this is unlikely to be a major factorsince the lower renal risk was the sameamong thosewhowere on RAAS blockadeas among those who were not on RAASblockade in the subgroup analyses (in-cidence rate 52 [95% CI 19–136] vs.48[95%CI16–151]per1,000person-years,respectively). Sixth, the renal functionmeasurement we used in this study wasnot measured GFRs obtained using iotha-lamate but estimates using serum creat-inine, which might affect our results.Previous structural-functional studieswere mostly done with measured GFR.Finally, unmeasured confounders werenot fully adjusted for in our study. Weidentified older age, lower systolicblood pressure, higher hemoglobin level,and higher HbA1c level with higher oddsof nonproteinuric diabetic kidney dis-ease assignment. The finding thathigher HbA1c is associated with nonpro-teinuria, in other words, with milderlesions, goes against the vast majority ofthe published literature. HbA1c is mostwidely accepted and used formonitoringlong-term glycemic control in patientswith diabetes; however, we acknowl-edge that HbA1c in patients with CKDcan be influenced by many unmeasuredfactors. One of these factors is the doseof recombinant human erythropoietinagents (42). We believe it is possiblethat with more severe interstitial fibrosis,

thosewithproteinuriawereusingahigherdose of recombinant human erythropoi-etin-stimulating agents that caused a lowHbA1c level. Conversely, it is possible thatwith milder interstitial fibrosis, thosewith nonproteinuria were using lesserythropoietin-stimulating agents, whichhad little effect on their HbA1c levelsand made their HbA1c levels look higherthan those in patients with proteinuria.

In conclusion, in propensity score–matched cohorts of biopsy-proven non-proteinuric diabetic kidney disease andproteinuric diabetic kidney disease, pa-tients with nonproteinuric diabetic kid-ney disease had lower blood pressurewith less frequent typical pathologicallesions and were at lower risk of CKDprogression and all-cause mortality. Fur-ther studies are warranted to confirmthese findings in other cohorts.

Funding. This study was supported in part by aMinistry of Health, Labour andWelfare Grant-in-Aid for Diabetic Nephropathy and Nephroscle-rosis Research (JP17ek0310003).The funding source had no role in the study

design or execution, data analysis, manuscriptwriting, or manuscript submission.Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.AuthorContributions.M.Y., K.F., J.H., T.T., andT.W. designed the study protocol, researcheddata, contributed to the discussion, wrote themanuscript, and reviewed and edited the man-uscript.A.H.,M.S.,K.Ki., T.F.,K.O., Y.Y.,H.K., Y.Su.,H.S., N.U., S.H., Y.Ue., S.N., H.Y., T.N., K.S., K.Ko.,Y.Sh., K.M., H.M., S.M., and Y.Ub. researcheddata, contributedto thediscussion,andreviewedand edited themanuscript. T.W. is the guarantorof this work and, as such, had full access to all thedata in the study and takes responsibility for theintegrity of the data and the accuracy of the dataanalysis.

References1. Viberti GC, Jarrett RJ, Keen H. Microalbumin-uria as prediction of nephropathy in diabetics.Lancet 1982;2:6112. Mogensen CE, Christensen CK, Vittinghus E.The stages in diabetic renal disease: with em-phasis on the stageof incipientdiabetic nephrop-athy. Diabetes 1983;32(Suppl. 2):64–783. Mogensen CE, Christensen CK. Predicting di-abetic nephropathy in insulin-dependent pa-tients. N Engl J Med 1984;311:89–934. Kimmelstiel P, Wilson C. Intercapillary lesionsin the glomeruli of the kidney. Am J Pathol 1936;12:83–98.75. Olson JL, Laszik ZG. Diabetic nephropathy. InHeptinstall’s Pathology of the Kidney. 7th ed.Jenette JC, Olson JL, Schwartz MM, Silva FG,Eds. Philadelphia, Wolters Kluwer, 2015, p. 897–9496. Kramer HJ, Nguyen QD, Curhan G, Hsu CY.Renal insufficiency in the absence of albuminuria

care.diabetesjournals.org Yamanouchi and Associates 901

and retinopathy among adults with type 2 di-abetes mellitus. JAMA 2003;289:3273–32777. MacIsaac RJ, Tsalamandris C, Panagiotopoulos S,Smith TJ, McNeil KJ, Jerums G. Nonalbuminuricrenal insufficiency in type 2 diabetes. DiabetesCare 2004;27:195–2008. Yokoyama H, Sone H, Oishi M, Kawai K,Fukumoto Y, Kobayashi M; Japan Diabetes Clin-ical Data Management Study Group. Prevalenceof albuminuria and renal insufficiency and asso-ciated clinical factors in type 2 diabetes: theJapan Diabetes Clinical Data Management study(JDDM15). Nephrol Dial Transplant 2009;24:1212–12199. Thomas MC, Macisaac RJ, Jerums G, et al.Nonalbuminuric renal impairment in type 2 di-abetic patients and in the general population(National Evaluation of the Frequency of Renalimpairment cO-existing with NIDDM [NEFRON]11). Diabetes Care 2009;32:1497–150210. Mottl AK, Kwon KS, Mauer M, Mayer-DavisEJ, Hogan SL, Kshirsagar AV. Normoalbuminuricdiabetic kidney disease in the U.S. population. JDiabetes Complications 2013;27:123–12711. Bash LD, Selvin E, Steffes M, Coresh J, AstorBC. Poor glycemic control in diabetes and the riskof incident chronic kidney disease even in theabsence of albuminuria and retinopathy: Ath-erosclerosis Risk in Communities (ARIC) Study.Arch Intern Med 2008;168:2440–244712. Jha V, Garcia-Garcia G, Iseki K, et al. Chronickidney disease: global dimension and perspec-tives. Lancet 2013;382:260–27213. United States Renal Data System. 2016USRDS Annual Data Report: Epidemiology ofKidney Disease in the United States. Bethesda,MD, National Institutes of Health, National In-stitute of Diabetes and Digestive and KidneyDiseases, 201614. ERA-EDTA Registry. ERA-EDTA Registry An-nual Report 2015. Amsterdam, the Netherlands,Academic Medical Center, Department of Med-ical Informatics, 201715. Japanese Society for Dialysis Therapy RenalData Registry. Annual report 2014, JSDT RenalData Registry (JRDR). Ren Replace Ther 3:1816. Imai E, Horio M, Nitta K, et al. Estimation ofglomerular filtration rate by the MDRD studyequation modified for Japanese patients withchronickidneydisease.ClinExpNephrol2007;11:41–5017. FiorettoP,MauerM,Brocco E, et al. Patternsof renal injury in NIDDM patients with micro-albuminuria. Diabetologia 1996;39:1569–157618. Tervaert TW, Mooyaart AL, Amann K,et al.; Renal Pathology Society. Pathologic

Classification of Diabetic Nephropathy. J AmSoc Nephrol 2010;21:556–56319. Furuichi K, Shimizu M, Yuzawa Y, et al.;Research Group of Diabetic Nephropathy, Min-istry of Health, Labour and Welfare of Japan,and Japan Agency for Medical Research andDevelopment. Clinicopathological analysis ofbiopsy-proven diabetic nephropathy based onthe Japanese classification of diabetic nephrop-athy. Clin Exp Nephrol 2017;22:570–58220. Hanafusa N, Nakai S, Iseki K, Tsubakihara Y.Japanese society for dialysis therapy renal dataregistryda window through which we can viewthe details of Japanese dialysis population. Kid-ney Int Suppl (2011) 2015;5:15–2221. Kasiske BL, Kalil RS,Ma JZ, LiaoM, KeaneWF.Effect of antihypertensive therapy on the kidneyin patients with diabetes: a meta-regressionanalysis. Ann Intern Med 1993;118:129–13822. NielsenS, SchmitzA, PoulsenPL,HansenKW,Mogensen CE. Albuminuria and 24-h ambulatoryblood pressure in normoalbuminuric and micro-albuminuric NIDDM patients. A longitudinalstudy. Diabetes Care 1995;18:1434–144123. Glassock RJ, Rule AD. The implications ofanatomical and functional changes of the agingkidney: with an emphasis on the glomeruli.Kidney Int 2012;82:270–27724. Osterby R. Morphometric studies of theperipheral glomerular basement membrane inearly juvenile diabetes. I. Development of initialbasement membrane thickening. Diabetologia1972;8:84–9225. Steffes MW, Bilous RW, Sutherland DE,Mauer SM. Cell and matrix components ofthe glomerular mesangium in type I diabetes.Diabetes 1992;41:679–68426. Harris RD, Steffes MW, Bilous RW,Sutherland DE, Mauer SM. Global glomerularsclerosis and glomerular arteriolar hyalinosis ininsulin dependent diabetes. Kidney Int 1991;40:107–11427. Mauer SM, Steffes MW, Ellis EN, SutherlandDE, Brown DM, Goetz FC. Structural-functionalrelationships in diabetic nephropathy. J ClinInvest 1984;74:1143–115528. Caramori ML, Kim Y, Huang C, et al. Cellularbasis of diabetic nephropathy: 1. Study designand renal structural-functional relationships inpatients with long-standing type 1 diabetes[published correction appears in Diabetes 2002;51:1294]. Diabetes 2002;51:506–51329. Budhiraja P, Thajudeen B, Popovtzer M.Absence of albuminuria in type 2 diabeticswith classical diabetic nephropathy: clinical path-ological study. J Biomed Sci Eng 2013;6:20–25

30. Shimizu M, Furuichi K, Yokoyama H, et al.Kidney lesions in diabetic patients with normoal-buminuric renal insufficiency. Clin Exp Nephrol2014;18:305–31231. Ekinci EI, Jerums G, Skene A, et al. Renalstructure in normoalbuminuric and albumin-uric patients with type 2 diabetes and impairedrenal function. Diabetes Care 2013;36:3620–362632. O’Sullivan ED, Hughes J, Ferenbach DA. Renalaging: causes and consequences. J Am Soc Neph-rol 2017;28:407–42033. Hommos MS, Glassock RJ, Rule AD. Struc-tural and functional changes in human kidneyswith healthy aging. J Am Soc Nephrol 2017;28:2838–284434. Mauer M, Caramori ML, Fioretto P, NajafianB. Glomerular structural-functional relationshipmodels of diabetic nephropathy are robust intype 1 diabetic patients. Nephrol Dial Transplant2015;30:918–92335. Weil EJ, Lemley KV, Mason CC, et al. Podo-cyte detachment and reduced glomerular cap-illary endothelial fenestration promote kidneydisease in type 2 diabetic nephropathy. KidneyInt 2012;82:1010–101736. FufaaGD,Weil EJ, LemleyKV, et al. Structuralpredictors of loss of renal function in AmericanIndians with type 2 diabetes. Clin J Am SocNephrol 2016;11:254–26137. Liu Y. Renal fibrosis: new insights into thepathogenesis and therapeutics. Kidney Int 2006;69:213–21738. GandhiM,Olson JL,Meyer TW. Contributionof tubular injury to loss of remnant kidneyfunction. Kidney Int 1998;54:1157–116539. Chevalier RL. The proximal tubule is theprimary target of injury and progression ofkidney disease: role of the glomerulotubularjunction. Am J Physiol Renal Physiol 2016;311:F145–F16140. Mazzucco G, Bertani T, Fortunato M, et al.Different patterns of renal damage in type 2diabetesmellitus: a multicentric study on 393 bi-opsies. Am J Kidney Dis 2002;39:713–72041. Koye DN, Magliano DJ, Reid CM, Jepson C,Feldman HI, Herman WH. Risk of progression ofnonalbuminuric CKD to end-stage kidney diseasein people with diabetes: the CRIC (Chronic RenalInsufficiency Cohort) Study. Am J Kidney Dis2018;72:653–66142. Ng JM, Cooke M, Bhandari S, Atkin SL,Kilpatrick ES. The effect of iron and erythropoi-etin treatment on the A1C of patients withdiabetes and chronic kidney disease. DiabetesCare 2010;33:2310–2313

902 Nonproteinuric Diabetic Kidney Disease Diabetes Care Volume 42, May 2019