The Familiality of Rapid Renal Decline in Diabetes

Despite significant advances in the management of diabetes, including widespread implementation of renoprotective therapies when indicated, diabetic kidney disease (DKD) remains the leading cause of end-stage renal disease (ESRD) in the U.S. and is associated with excess morbidity and mortality (1,2).

Progressive glomerular filtration rate (GFR) decline precedes ESRD and has recently been established as the predominant clinical feature of DKD (3–7). For some patients with DKD, renal function declines gradually, whereas others experience a sudden decline, quickly reach ESRD (defined by a GFR <15 mL/min/1.73 m²), and require renal replacement therapy in the form of dialysis or the receipt of a functioning kidney through donor transplantation to survive. Despite interindividual variability, renal function decline progresses at a steady, or linear, rate over the course of DKD (2,5–7). In addition, those with more rapid renal decline have higher risk of all-cause mortality (8).

Wide individual variation in the rate of progression of renal function decline in diabetes has motivated recent studies to identify biomarkers that are associated with or predictive of rapid renal decline. Such biomarkers may have utility in patient surveillance and management. In addition to studies examining proteomic (9,10), metabolomic (11,12), and environomic (13,14) profiles associated with renal decline in diabetes, various studies are also underway to identify genetic factors that contribute to its risk (15). Although these genetic studies are likely to be more powerful than previous studies that have examined a spectrum of DKD phenotypes (16–19), the familial risk of rapid renal decline has not been investigated. Understanding the familiality of rapid renal decline is paramount to recognizing the magnitude of the influence of shared factors, both environmental and genetic, on rapid renal decline.

To advance this area of research, we set out to investigate the familial nature of progressive renal function decline in diabetes using the Utah Population Database (UPDB), a unique research resource that links genealogical data to electronic health record data. Through this population-based retrospective cohort study, we show that there is strong evidence of familial clustering of rapid renal decline. Our findings support the hypothesis that shared familial factors (e.g., shared environmental and genetic factors) play a role in its susceptibility.

Using electronic health information from the University of Utah Health Sciences Center (UUHSC) Enterprise Data Warehouse (EDW), patients with diabetes were identified using the following ICD-9 and ICD-10 codes: ICD-9 250.xx and ICD-10 E10.xx and E11.xx. In total, we identified >105,000 patients with diabetes in the UUHSC EDW.

To characterize estimated GFR decline (eGFR) trajectories, we retrieved all available outpatient serum creatinine measurements for these patients (Fig. 1). The date of each patient’s first available outpatient serum creatinine measurements was used to establish their date of entry to our study. The duration of follow-up was calculated as the time from a patient’s first available outpatient serum creatinine measurement until their last available outpatient serum creatinine measurement or until ESRD was reached. We restricted our analysis to patients whose first (i.e., baseline) serum creatinine measurement was measured when they were between 18 and 60 years old. To more accurately estimate kidney function decline, serum creatinine measurements that occurred after a patient reached ESRD, either observed (eGFR <15 mL/min/1.73 m²) or indicated by an ICD code for ESRD (ICD-9: 585.6; ICD-10: N18.6), dialysis (ICD-9: V45.11; ICD-10: Z99.2), or kidney transplantation (ICD-9: V42.0; ICD-10: Z94.0), were censored from our analysis.

eGFR was calculated using these serum creatinine measurements and the Chronic Kidney Disease Epidemiology Collaboration creatinine formula (20).

To model eGFR trajectory, we required that patients have ≥3 serum creatinine measurements and a follow-up time ≥1 year. eGFR trajectories for patients meeting these criteria were estimated using a simple linear regression model (5). Patients were then categorized into two groups according to their rate of renal function decline based on the Kidney Disease Improving Global Outcomes (KDIGO) guidelines’ (21) definition of rapid progression: slow decliners (eGFR slope ≥ −5 mL/min/1.73 m²/year) and rapid decliners (eGFR slope < −5 mL/min/1.73 m²/year). Clinical data (i.e., systolic and diastolic blood pressure, BMI, hemoglobin A_1c [HbA_1c], etc.) closest to each patient’s baseline eGFR determination were obtained from electronic health records.

This study was approved by the Institutional Review Board of the University of Utah and the Utah Resource for Genetic and Epidemiological Research, which are responsible for oversight of the UPDB. An Institutional Review Board Waiver of Consent and Authorization was obtained to conduct this study.

The UPDB is a unique population-based genealogy resource of Utah pioneers and their descendants along with all later in-migrants that is linked to electronic health record data from the UUHSC. Established in the 1970s, the UPDB currently contains family histories and demographic data for >11 million individuals, including some who lived as long ago as the 17th century. The UPDB comprises descendants of individuals with at least one vital event (birth, marriage, or death) in Utah or on the Mormon Pioneer Trail (22). The majority of families living in Utah are represented in the UPDB’s multigenerational pedigrees. Most families can be linked across 5 generations, and some span as many as 17 generations. Individuals included in the UPDB reflect the Western and Northern European roots of its early settlers (23). Information derived from birth, death, marriage, and divorce certificates, driver licenses, censuses, and other records further enriches this valuable data set.

The pedigrees and other information associated with each individual in the UPDB are updated annually as new records become available (24,25). Patients with rapid renal decline identified through the UUHSC EDW were linked to individuals in the UPDB prior to the familial analyses.

We quantified kindred-specific risk of rapid renal decline using the familial standardized incidence ratio (FSIR) (26). The FSIR computes an individual family’s risk of disease by using the number of blood relatives in the pedigree, their degree of relatedness to the proband, and whether these relatives are observed to have the disease or phenotype of interest. In this study, FSIR was calculated as the ratio of the observed number of persons with rapid renal decline in a pedigree member relative to the expected number where the expectation is based on population estimates from the UPDB that adjust for age and sex compositions. To identify pedigrees with excess clustering of rapid renal decline, we selected those with ≥5 observed individuals with rapid renal decline, an FSIR >2, and a P value <0.05.

Baseline clinical characteristics of patients were summarized by frequencies and percentages for categorical variables and as medians and first and third quartiles for continuous data. Patient-specific eGFR trajectories (i.e., eGFR slopes) were estimated with linear regression in which (the repeating) eGFR is the dependent variable and time is the sole covariate. All familial analyses were conducted using a suite of kinship analysis software developed in-house at the UPDB and the software package R (27). Logistic regression models were used to estimate the magnitude of familial risk of rapid renal decline among relatives of rapid renal decline probands. These (familial recurrence risk as measured by odds ratios [ORs]) logistic regression models are based on first identifying all persons in the sample with rapid renal decline (probands). All of the probands’ first- and second-degree relatives are then identified. Next, a set of sex- and birth-year–matched population (unaffected) control subjects was selected at a 5:1 (control/proband) ratio who were alive when their matched probands were measured for renal function. All of the control subjects’ first- and second-degree relatives are also identified. The two sets of relatives for the probands and the control subjects are the basis for estimating ORs. Logistic regressions are estimated in which the outcome is whether or not the relative has rapid renal decline and the key predictor is whether the individual is a relative of a proband or a control subject. We conduct conditional logistic regression to generate ORs and to account for matching done separately for first- and second-degree relatives. We acknowledge the inclusion of multiple sightings of the same relative in the calculations, which has been shown to produce unbiased estimates (28). Statistical analyses were conducted using the software package R (27).

We identified 105,335 patients with diabetes with 997,739 serum creatinine measurements in the UUHSC EDW. Among these patients, 15,612 were between 18 and 60 years old, had a minimum of 3 serum creatinine measurements, and had a follow-up time ≥1 year (Fig. 1). Together, these patients comprise the UUHSC Renal Function Decline (RFD) cohort.

Clinical characteristics of patients included in the UUHSC RFD cohort are shown in Table 1. This cohort includes 8,316 (53.3%) men and 7,296 (46.7%) women, 97.4% of whom are of white ethnicity. The median age, HbA_1c, and BMI of patients in this cohort were 47 years, 6.6%, and 32 kg/m², respectively. Their median blood pressure was 127/78 mmHg. A total of 87.1% of patients in the UUHSC RFD cohort were diagnosed with type 2 diabetes, whereas 12.9% had a diagnosis of type 1 diabetes. More than half (53.6%) were treated with insulin. The median baseline urinary albumin-to-creatinine ratio (UACR) was 13 mg/g, and the median baseline eGFR was 97.1 mL/min/1.73 m². The majority of patients had normal to mild loss of renal function at baseline median (60.5% had an eGFR ≥90 mL/min/1.73 m², and 32.6% had an eGFR 60–89 mL/min/1.73 m²), and 45.3% were treated with ACE inhibitors (ACE-I) or angiotensin II receptor blockers (ARB).

Patients included in the UUHSC RFD cohort had a median duration of follow-up of 5.7 years, and the median number of eGFR measurements per patient was 7.0, yielding a median density of 1.4 eGFR measurements/year (Table 1). These data were used to characterize renal function trajectories of this cohort, as the slope associated with time at which the repeated (within subject) measure of eGFR is the dependent variable. Figure 2 illustrates the eGFR trajectories of select individuals from the UUHSC RFD cohort.

Although the median rate of eGFR decline of this cohort was −1.0 mL/min/1.73 m²/year, the overall distribution of eGFR decline slopes of this cohort was skewed toward steeply negative slopes (Fig. 3). We stratified patients in the UUHSC RFD cohort to slow (n = 13,485) and rapid (n = 2,127) decliners (Fig. 3 and Table 1) based on the KDIGO guidelines’ definition of rapid progression.

Slow decliners in this cohort had a median eGFR slope of −0.7 mL/min/1.73 m²/year, whereas rapid decliners had a median eGFR slope of −8.0 mL/min/1.73 m²/year (P value <0.001). Median age, blood pressure, BMI, and baseline eGFR were similar between slow and rapid decliners. However, rapid decliners had higher HbA_1c (median 7.2% vs. 6.5%; P value <0.001) and higher baseline UACR (median 32 vs. 12; P value <0.001) than slow decliners. Seventy-seven percent of rapid decliners had either normoalbuminuria (UACR <30 mg/g) or microalbuminuria (UACR 30–299 mg/g) at baseline. A higher proportion of rapid decliners were treated with insulin and ACE-I or ARB (P value <0.001). Rapid decliners had a shorter duration of follow-up compared with slow decliners (median 2.9 years vs. 6.5 years; P value <0.001), and more rapid decliners progressed to ESRD over the course of their follow-up (83 [0.6%] vs. 207 [9.7%]; P value <0.001). Among the slow decliners who progressed to ESRD, 60% had impaired renal function (chronic kidney disease [CKD] stage 3 or higher) at baseline, whereas those who entered the study in CKD stages 1 or 2 had a median of 15.2 years of follow-up along with a median eGFR slope of −2.3 mL/min/1.73 m²/year.

Rapid decliners were linked to >3,800 pedigrees in the UPDB whose founders were born as early as 1582 and have as many as 124,000 descendants spanning up to 12 generations (Supplementary Fig. 1). Among the pedigrees identified in the UPDB, we identified 51 high-risk pedigrees with at least five individuals with rapid renal decline and an excess risk of rapid renal decline, defined by an FSIR >2 and a P value <0.05 (Table 2 and Supplementary Table 1). The median birth year for the founders of these high-risk pedigrees was 1803. The number of descendants from these founders ranged from 1,450 in the smallest pedigree to >24,000 in the largest pedigree (median 7,133). These pedigrees included as many as 15 rapid decliners (range 5–15; median 6). The FSIRs for these families ranged from 2.16 to 16.19, and the median FSIR was 3.28, demonstrating strong familial aggregation of rapid renal decline in these high-risk pedigrees. Two high-risk rapid renal decline pedigrees that are representative of those that we identified in the UPDB are presented in Fig. 4.

Using the UPDB, we estimated the OR of rapid renal decline in relatives of individuals with rapid renal decline in relation to the relatives of individuals of control subjects. In total, 5,934 first-degree relatives and 9,567 second-degree relatives of rapid decliners were identified in the UPDB; there were 32,741 first-degree relatives and 52,892 second-degree relatives of matched unaffected population control subjects. The OR of rapid renal decline among first-degree relatives of rapid decliners compared with that of first-degree relatives of control subjects in the UPDB was 9.46 (P = 2.46 × 10⁻¹¹ [95% CI 4.89–18.31]) (Table 3). When we restricted the analysis to control subjects with diabetes with observed slow renal decline, the OR of rapid renal decline among first-degree relatives of rapid decliners was 1.92 (P = 0.007 [95% CI 1.19–3.09]). Second-degree relatives of rapid decliners were not shown to be at greater risk of experiencing rapid renal decline in relation to second-degree relatives of slow decliners (OR 0.89; P = 0.79 [95% CI 0.37–2.13]).

This is the first study to demonstrate that rapid renal decline, the predominant clinical feature of DKD, clusters in families with diabetes. By linking renal function trajectories in >15,000 patients with diabetes and a large genealogical database, we were able to identify 51 large, extended pedigrees with a significant excess of individuals with rapid progression of renal decline. Significantly elevated relative risk of rapid renal decline was also observed in first-degree relatives within this cohort. Together, these data suggest that shared familial factors, perhaps environmental, genetic, or both, contribute to the risk of rapid renal decline in patients with diabetes. Further study of these families is necessary to understand the magnitude of the influence of these factors on rapid renal decline.

The annual loss of eGFR that is part of the normal physiology of aging is estimated to be <1 mL/min/1.73 m²/year (29,30). Our goal was to identify pedigrees at significant risk of rapid renal decline. As such, we used a cutoff five times this rate, and one that is in line with the current KDIGO guidelines’ definition of rapid progression, to distinguish between rapid renal decline and more modest changes in GFR. Within the UUHSC RFD cohort, 14% of patients with diabetes experienced renal function loss at a rate of ≥5 mL/min/1.73 m²/year. This observation is in line with previous data from patients with diabetes in the Joslin Kidney Studies in which ∼11% of patients with type 1 and 13% of patients with type 2 diabetes were reported to experience similar rates of renal decline (2). Interestingly, 77% of rapid decliners in the UUHSC RFD cohort had either normoalbuminuria or microalbuminuria (UACR <299 mg/g) at baseline. Similarly, this observation is supported by data from previous studies that have reported early progressive renal decline in patients with type 1 diabetes with normoalbuminuria or macroalbuminuria (31). Importantly, this suggests that early renal decline is the primary disease process of impaired renal function and that increased albuminuria is either a consequence of this or develops along with early renal decline as this disease progresses.

Surrogate markers of DKD, including albuminuria (32–42) and eGFR (43,44), have previously been shown to aggregate in families. In the earliest investigation of familial clustering of DKD, Seaquist et al. (32) examined the risk of DKD between 2 sets of families with type 1 diabetes: 11 with probands who were free of DKD and 26 with probands who had undergone kidney transplantation due to DKD. Although only 17% of siblings with diabetes of probands without DKD had evidence of DKD (defined as overt proteinuria), the investigators found that >80% of siblings with diabetes of probands with DKD went on to develop DKD.

Increased familial risk of DKD has been confirmed in several other studies. In a study of patients with type 1 diabetes attending the Steno Diabetes Center in Denmark and having siblings with diabetes, 33% of siblings of patients with nephropathy had DKD, whereas only 10% of siblings of normoalbuminuric patients had DKD (34). In patients from the Joslin Clinic, the cumulative risk of advanced DKD (i.e., persistent proteinuria or ESRD) in siblings after 25 years of type 1 diabetes was 72% if the proband had persistent proteinuria but only 25% if the proband did not (36). Similarly, the Diabetes Control and Complications Trial (DCCT) reported a 5.4-fold increased risk of DKD in relatives of DKD-positive subjects compared with DKD-negative subjects (37). These findings are consistent with the increased risk to siblings of DKD-positive subjects identified in a Finnish population (45).

In the current study, we expand upon these earlier studies by demonstrating familial clustering of rapid renal decline. Taken together with our findings, it is clear that shared familial factors contribute to the risk of DKD and its related traits, including rapid renal decline. Like DKD, the pathogenesis of rapid renal decline in diabetes is complex and multifactorial. In addition to major risk factors that include hyperglycemia, hypertension, dyslipidemia, and albuminuria, various environmental factors (e.g., employment status, education level, and access to health care) have all been shown to contribute to DKD risk (46). Additionally, estimates of the heritability (h²), or the proportion of total variation due to genetic effects, of DKD and several DKD-related traits (e.g., urinary albumin excretion rate [AER] and eGFR) have also established that genetic factors contribute to increased risk of DKD in families with diabetes (18,40–43,47,48). Although several candidate genes and associated variants have been identified to date, our understanding of the genetic basis of DKD is far from complete. Family-based genetic studies involving large pedigrees, similar to those identified in our study, are likely to prove more powerful than previous studies involving unrelated case and control subjects (e.g., genome-wide association studies). Such studies, when coupled with next-generation sequencing technologies, are especially well powered to identify low-frequency and rare genetic variants in novel genes and pathways that are likely to contribute most to the overall risk of DKD and its related traits.

The biggest challenge facing researchers studying DKD is the fact that DKD is a mosaic of subphenotypes (49). Patients with DKD present with varying levels of albuminuria, at various stages of CKD, and experience different rates of progression of renal function decline, all of which are likely to be influenced by a myriad of risk factors, including some that are shared and some that are distinct, that contribute to their pathogenic process. In fact, although the appearance of albuminuria is often considered the first clinical sign of DKD, in patients with diabetes, renal function decline frequently occurs prior to the onset of proteinuria and has even been shown to precede the development of microalbuminuria (2,5,6,29). Although its prevalence is higher among patients with type 1 diabetes with proteinuria, being present in 50% of such patients, ∼10% of those with normoalbuminuria and 35% of those with microalbuminuria experience early renal function decline at a rate of eGFR loss ≥3.3 mL/min/1.73 m²/year (6,29). Similarly, >60% of patients with type 2 diabetes with proteinuria, and as many as 20% with normoalbuminuria and 33% with microalbuminuria, develop progressive renal function decline (2). It is very likely that the familial risk factors, both environmental and genetic, contributing to the etiology of DKD and its related traits, including rapid renal decline, contribute to various phases of this disease process. Although the families identified in this study are enriched for rapid renal decline, given the complex nature of DKD, it may be difficult to dissect the precise contribution of their shared risk factors on this disease process (e.g., whether the risk factors identified in these families contribute to the risk of DKD’s initiation or the rate of its progression once the disease has developed).

With relatively few exceptions, cohorts used to investigate risk factors that contribute to DKD have not been well characterized, and most have been exclusively cross-sectional in nature. Given the complexity of this disease, the most fruitful approach to identifying novel risk factors for DKD is likely to be one that focuses on homogeneous subphenotypes of this disease (50). Because progressive renal function decline initiates early in the natural history of DKD, this feature has been considered to be the primary disease process that leads to impaired renal function, and eventually ESRD, in patients with diabetes (29). As such, approaches to identify risk factors for DKD should focus on patients who are at the greatest risk of rapid renal decline. In line with this concept, our findings suggest that shared familial factors play a role in rapid renal decline in DKD and may predispose a subset of individuals to rapid loss of kidney function and increased risk of ESRD, highlighting this phenotype as a strong candidate for future studies aimed at understanding its basis.

Our study has some limitations. First, in an effort to identify a phenotypically homogenous study population, we selected only ∼15% of all available patients with diabetes in the UUHSC EDW; therefore, our study may not be fully generalizable to patients not included in this study. The familiality of rapid renal decline could be assessed in a more general population after specific associations are identified. Second, inherent in the design of the study is a bias toward patients who are receiving, at the very least, intermittent health care from a medical specialist, and, therefore, the exclusion of patients who do not receive health care presents a limitation to our findings. Third, as discussed above, both environmental and genetic factors tend to cluster in families. Due to the nature of our familial analysis, we were not able to distinguish between the contributions of these shared factors. Further study of these families is required to understand the magnitude of the influence of various shared familial factors on rapid renal decline. Lastly, other risk factors of the progression of renal function decline, including blood pressure, glycemic, and UACR, may cluster in the high-risk families identified in this study. These additional risk factors were not examined in the familial analysis performed in this study. Further studies are currently underway to investigate whether these additional risk factors are enriched in these same pedigrees and to examine their contribution to the familial clustering of rapid renal decline observed in this study.

Funding. The authors received grant support from the National Institute of Diabetes and Digestive and Kidney Disease’s Diabetic Complications Consortium (DiaComp) (32307-1/U24-DK-115255 to M.G.P.), the National Kidney Foundation of Utah and Idaho (to M.G.P.), and Driving Out Diabetes, a Larry H. Miller Family Wellness Initiative (to M.G.P.).

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

Author Contributions. S.G.F., Z.Y., and M.H.P. researched the data. S.G.F. and M.G.P. conceived the idea and designed the study. S.G.F. and M.G.P. analyzed the data with assistance from Z.Y., A.M.L., A.A., J.Y., and T.G. S.G.F. and M.G.P. drafted the manuscript with assistance from A.A., T.R.S., K.L.R., and K.R.S. S.G.F., Z.Y., A.M.L., A.A., M.H.P., L.D., T.R.S., J.Y., T.G., K.L.R., K.R.S., and M.G.P. revised the manuscript and approved the final version. M.G.P. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of its analysis.