Association of Estimated Glomerular Filtration Rate With Progression of Albuminuria in Individuals With Type 2 Diabetes

Investigation into the pathogenesis of diabetic nephropathy is urgently required because of global public health concerns (1–4). Therefore, early detection and treatment of diabetic nephropathy are particularly important to overcoming the current situation. A supraphysiologic increase in whole-kidney glomerular filtration rate (GFR), known as glomerular hyperfiltration, has been considered a characteristic of early-stage classical diabetic nephropathy (5–7). Meanwhile, a reduced number of nephrons in the latter phase cause single-nephron hyperfiltration to maintain whole-kidney filtration (6,8,9). According to the theory that glomerular hyperfiltration induces kidney damage by increasing the glomerular capillary pressure and consequently increasing the amount of ultrafiltrate in Bowman’s space and the flow into the proximal tubule (6,7,10), both high and low GFRs in the whole kidney may contribute to the progression of albuminuria in individuals with diabetes. However, this remains unclear because of the limited number of studies with a large sample size. We aimed to elucidate the association of GFR with subsequent albuminuria progression in individuals with diabetes.

This single-center retrospective cohort study, performed as part of the Cohort Study Elucidating Factors Associated With the Pathogenesis and Prognosis of Diabetic Kidney Disease at the Tokyo Women’s Medical University School of Medicine, adhered to the Declaration of Helsinki. The local ethics committee approved the protocol (Tokyo Women’s Medical University School of Medicine, Tokyo, Japan; approval no. 3932), in which the need for informed consent was waived because of the nonprospective interventional design. Instead, the website of the institution offered participants an opportunity to opt out.

Initially, we identified 8,810 Japanese individuals age ≥18 years with type 2 diabetes and without a history of chronic kidney replacement therapy (i.e., chronic dialysis or kidney transplantation) who visited our outpatient clinic at Tokyo Women’s Medical University School of Medicine between 1 August 2003 and 30 June 2017, previously described in detail (11). Of the 8,810 individuals, women pregnant at baseline (n = 48) and those with malignant disease at baseline (n = 160), history of unilateral nephrectomy at baseline (n = 7), biopsy-proven diagnosis of nondiabetic nephropathy at baseline (n = 2), acute kidney injury or postrenal failure at baseline (n = 2), no follow-up data on kidney function (n = 257), and missing baseline profile data (n = 14) were excluded, as described in our previous study (11). Participants with urinary albumin-to-creatinine ratio (UACR) ≥300 mg/g at baseline (n = 1,072) and no follow-up data on UACR (n = 630) were also excluded from the current study. Eventually, 6,618 individuals were deemed eligible. All relevant baseline data were available for the 6,618 participants. These 6,618 participants were classified into five groups based on category of baseline eGFR, as defined below.

Data obtained from blood and urine samples were measured using random and first morning samples, respectively. The formula proposed by the Japanese Society for Nephrology was adopted to estimate GFR (12). When HbA_1c levels were measured as Japan Diabetes Society values, they were converted into National Glycohemoglobin Standardization Program values using the formula proposed by the Japan Diabetes Society (13). On the basis of the clinical practice guidelines proposed by the Kidney Disease: Improving Global Outcomes (KDIGO) organization (14,15), albuminuria was classified into three categories: normoalbuminuria (UACR <30 mg/g), microalbuminuria (UACR ≥30 to <300 mg/g), and macroalbuminuria (UACR ≥300 mg/g). Hypertension was defined by blood pressure ≥140/90 mmHg (16) or use of antihypertensive drugs.

Exposure was baseline estimated GFR (eGFR) (mL/min/1.73 m²), treated as a categorical variable and classified into five categories (≥90, 75–90, 60–75, 45–60, and <45), as well as a continuous variable.

The outcome was progression of albuminuria category (i.e., from normoalbuminuria to micro- or macroalbuminuria or from micro- to macroalbuminuria), which was determined by at least two consecutive increases in UACR, considering the day-to-day variability of UACR. Furthermore, 1 week was set as the minimum interval between the dates of the two consecutive increases in UACR; in other words, the outcome was judged by two consecutive UACR increases at least 1 week apart to reduce the risk of misclassification.

The administrative censoring date was set as 30 September 2021. The last follow-up date for each participant was defined as the date when the individual reached the outcome or the date when UACR was last measured before the administrative censoring date without reaching the outcome.

All analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC). A two-tailed P value <0.05 was considered statistically significant.

Differences in proportion, mean, or median of baseline data between the five groups were compared using the χ² test, ANOVA, or Kruskal-Wallis test as appropriate. Hazard ratios (HRs) for the outcome were estimated using the multivariable Cox proportional hazards model. The adjusted (standardized) cumulative incidence of the outcome in the five groups was calculated using regression estimates from the relevant Cox model (17). In the analysis treating exposures (i.e., baseline eGFRs) as a continuous variable, the multivariable-adjusted restricted cubic spline model was used, where four knots were placed at the 5th, 35th, 65th, and 95th percentile levels of baseline eGFR, and the reference was set to 90 mL/min/1.73 m². In all multivariable models used to estimate HRs for the outcome, the following 14 baseline variables were incorporated as covariates: sex, age, history of coronary artery disease or stroke, smoking status (current/former vs. never), use of insulin, use of other antidiabetic drugs, use of ACE inhibitors or angiotensin receptor blockers (ARBs), BMI, systolic blood pressure, HbA_1c, logarithmically transformed triglycerides, HDL cholesterol, LDL cholesterol, and logarithmically transformed UACR.

We conducted a series of sensitivity analyses. First, for determining the outcome, a 6-month interval was set as the maximum time period between the dates of two consecutive increases in UACR in addition to the minimum interval of 1 week, referring to the position statement of the American Diabetes Association (4). Second, when setting progression of albuminuria category as the outcome, a modest increase in UACR, such as an increase from 29 to 31 mg/g or 299 to 301 mg/g, could lead to reaching of the outcome. Considering this limitation, progression of albuminuria category accompanied by at least a 30% increase in UACR from baseline was used as the outcome instead of progression of albuminuria category alone, referring to previous studies (18–20). Third, association of baseline eGFR with the outcome was examined in each group classified as normo- or microalbuminuria at baseline. Fourth, association of baseline eGFR with the outcome was examined in each group classified as age ≤49, 50–64, or ≥65 years at baseline. Fifth, when participants with microalbuminuria at baseline experienced regression to normoalbuminuria afterward, the phenotype was no longer microalbuminuria at that time; therefore, in participants with microalbuminuria at baseline, the Fine and Gray subdistribution hazards model, which is the competing risks model, was adopted for comparison among the five groups. In the analysis, incidence of regression to normoalbuminuria before reaching the outcome was treated as a competing risk. Sixth, number of UACR measurements per person per year during the follow-up period was incorporated as a covariate in addition to the abovementioned 14 variables. Seventh, the analysis was conducted in participants without use of ACE inhibitors, ARBs, or sodium–glucose cotransporter 2 (SGLT2) inhibitors at baseline (n = 4,198), because these drugs can cause changes in intraglomerular pressure (6,21,22). In 2,420 individuals excluded from this analysis, 2,370, 23, and 27 used ACE inhibitors or ARBs alone, SGLT2 inhibitors alone, and their combination at baseline, respectively. Finally, baseline uric acid levels were incorporated as a covariate in addition to the 14 variables mentioned above in participants with these data available (n = 5,701), because uric acid levels have been shown to be associated with not only eGFR values but also progression of diabetic nephropathy (23,24).

In the eligible 6,618 individuals, comprising 2,459 women and 4,159 men with a mean (± SD) age of 60 ± 12 years, the mean baseline eGFR was 74.5 ± 19.3 mL/min/1.73 m². The number of participants using ACE inhibitors or ARBs at baseline was 2,397 (36.2%) (Table 1). Of the 2,474 participants with baseline UACR ≥30 mg/g and/or eGFR <60 mL/min/1.73 m², 1,313 (53.1%) used ACE inhibitors or ARBs at baseline. Comparing the five groups classified based on baseline eGFR, age, proportion of participants with a history of coronary artery disease or stroke, use of ACE inhibitors or ARBs, and use of other antihypertensive drugs increased as baseline eGFR decreased (Table 1). Meanwhile, HbA_1c levels decreased as baseline eGFR decreased (Table 1). Baseline UACR level and proportion of microalbuminuria were higher in participants with eGFR <60 mL/min/1.73 m² than in those with eGFR ≥60 mL/min/1.73 m² (Table 1).

Of the 4,999 participants with normoalbuminuria and 1,619 with microalbuminuria at baseline, 822 and 368 reached the outcome (i.e., progression of albuminuria category), respectively, during the median follow-up period of 6.3 years (interquartile range [IQR] 3.1–10.8 years). The median number of UACR measurements per person during the follow-up period was 1.1 per year (IQR 0.8–1.6 per year).

The adjusted cumulative incidence of the outcome in the five groups classified based on category of baseline eGFR is shown in Fig. 1A. The risk of the outcome in participants with baseline eGFR of ≥90, 60–75, 45–60, or <45 mL/min/1.73 m² was 1.38, 1.34, 1.81, or 2.37 times, respectively, compared with the risk in those with baseline eGFR of 75–90 mL/min/1.73 m² (Fig. 1B). When baseline eGFR was further classified as ≥105, 90–105, 75–90, 60–75, 45–60, 30–45, or <30 mL/min/1.73 m², a similar trend was seen (Fig. 1C). The inverse J-shaped curve was more clearly shown by the spline model, where baseline eGFR was treated as a continuous variable (Fig. 2).

In the analysis where a 6-month interval was set as the maximum time period between the dates of two consecutive increases in UACR in addition to the minimum interval, 735 participants reached the outcome during the median follow-up period of 6.8 years (IQR 3.5–11.3 years). The result was similar to that of the primary analysis (Fig. 3A). In the analysis using progression of albuminuria category accompanied by at least a 30% increase in UACR from baseline as the outcome, 1,174 participants reached the outcome during the median follow-up period of 6.3 years (IQR 3.1–10.8 years), and a similar result was obtained compared with that of the primary analysis (Fig. 3B). In the analysis using the Fine and Gray subdistribution hazards model in 1,619 participants with microalbuminuria at baseline, 346 and 509 experienced the outcome and regression to normoalbuminuria before reaching the outcome during the median follow-up period of 3.1 years (IQR 1.2–6.3 years), respectively. The result was not changed (Supplementary Fig. 1A). Other sensitivity analysis results showed similar trends compared with the primary analyses findings (Fig. 3C–G and Supplementary Fig. 1B–D).

In this large single-center retrospective cohort study of 6,618 Japanese adults with type 2 diabetes with a median follow-up period of 6.3 years, we showed that both eGFRs ≥90 and <75 mL/min/1.73 m² at baseline were risk factors for the subsequent progression of albuminuria category, independent of known risk factors. In the multivariable-adjusted restricted cubic spline model treating baseline eGFR as a continuous variable, the result was consistent with the abovementioned findings. Sensitivity analyses also strengthened the robustness of these findings.

We considered that the significant association of higher baseline eGFR with the subsequent progression of albuminuria in the current study, at least in part, suggests harmful effects of whole-kidney hyperfiltration on kidneys. In that case, the relatively lower threshold value of eGFR (∼90 mL/min/1.73 m²) in this study, as shown in Figs. 1C and 2, must be discussed. Although no established definition of hyperfiltration exists, a systematic review regarding the assessment and definition of hyperfiltration, including 405 studies, reported the median threshold of hyperfiltration as 135 mL/min/1.73 m² (range 90.7–175 mL/min/1.73 m²) (25). In a meta-analysis of 37,872 participants from the general population, a higher eGFR based on serum creatinine, cystatin C, or their combination was a risk factor for end-stage kidney disease, and the threshold of eGFR was ∼105 mL/min/1.73 m² (26). The difference between the threshold in the current study and those in previous studies might be partly explained by the finding that the number of nephrons (i.e., inherent GFR) was lower in Japanese individuals than in those of other races (approximately two-thirds) (27). Nevertheless, we cannot deny the possibility that the higher eGFR in the current study was simply caused by loss of muscle mass resulting from malnutrition or sarcopenia (i.e., ostensibly higher eGFR), although BMI was incorporated as a covariate in the related analyses.

Meanwhile, the lower the eGFR, the higher the risk of albuminuria progression in the current study. Our previous study of individuals with type 2 diabetes showed that nonalbuminuric kidney insufficiency was associated with increased risk of subsequent kidney outcomes (11). A cohort study of 407 indigenous Australians, in which GFR was measured by using iohexol, showed that those with GFR <60 mL/min/1.73 m² at baseline were more prone to experiencing an increase in UACR than those with preserved kidney function during a 3-year follow-up period, regardless of the status of albuminuria at baseline (28). The U.K. Prospective Diabetes Study 74 also demonstrated that elevated plasma creatinine levels predicted the incidence of macroalbuminuria (29). These findings, together with the present findings, suggest that lower GFR itself promotes further kidney damage, the reason for which might be partly explained by single-nephron hyperfiltration.

The current study has several limitations: 1) eGFR based on serum creatinine used as a substitute for GFR; 2) the possibility that the progression of albuminuria category was underestimated because of the study design (i.e., retrospective cohort study); 3) the possibility that the categorization of albuminuria at baseline was improper as a result of the single UACR measurement, although the timing of urine collection was restricted to the first in the morning in this study to minimize the misclassification (30,31); 4) the inability to evaluate time-dependent changes in laboratory data, blood pressure, BMI, and medications, such as ACE inhibitors, ARBs, SGLT2 inhibitors, and other antidiabetic drugs, during the follow-up periods; and 5) an ethnically homogeneous population from a single Japanese university hospital, limiting the generalizability of the present findings.

In conclusion, the present cohort study of a large number of Japanese adults with type 2 diabetes suggests that both high and low GFRs are implicated in the pathogenesis of albuminuria progression. However, multicenter prospective cohort studies using measured GFR with larger sample sizes are required to confirm our findings.

Acknowledgments. The authors are grateful to Y. Yokoyama, E. Tauchi, S. Yamashita, and I. Nyumura (Division of Diabetology and Metabolism, Department of Internal Medicine, Tokyo Women’s Medical University School of Medicine) for their assistance in data collection.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.H. contributed to the conception and design of the study and the analysis and interpretation of data and was responsible for drafting the manuscript. K.H., T.M., Y.Y., N.Y., H.M., and T.B. contributed to the data collection and preparation. All authors revised the manuscript and approved the final version of the manuscript. T.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.