African Caribbean Ethnicity Is an Independent Predictor of Significant Decline in Kidney Function in People With Type 1 Diabetes

For many people with type 1 diabetes, advanced stages of chronic kidney disease (CKD) and end-stage kidney disease (ESKD) remain a major cause of premature morbidity and mortality (1). Recent studies have reported a lifetime risk of ESKD between 10 and 15% for people with type 1 diabetes, despite renoprotective treatment (e.g., renin-angiotensin system [RAS] inhibitors), which suggest that although diabetes care has improved over the past 20 years, the risk of ESKD may be increasing in type 1 diabetes (1). Furthermore, varying rates of progression of CKD have been observed in people with type 1 diabetes, with some being more susceptible to faster decline in kidney function and rapid progression toward ESKD than others (2,3). Indeed, recent data suggest that such fast decliners have an interval of 2 to <10 years that separates normal kidney function from ESKD and account for the majority of cases of ESKD in type 1 diabetes (2). Most of these data come from studies in Caucasian cohorts, and there is limited information on the risks of progression of CKD in people with type 1 diabetes from other ethnicities (3,4).

People of African Caribbean origin with type 2 diabetes or hypertension appear to progress to advanced stages of CKD faster than people of Caucasian descent (5,6). There are no equivalent data on the impact of African Caribbean ethnicity on the progression of CKD in people with type 1 diabetes. The aim of this study was to evaluate the clinical and demographic features of an ethnically diverse cohort of people with type 1 diabetes with relatively preserved kidney function at baseline (all had eGFR >45 mL/min/1.73 m²) that were associated with progression of CKD.

Clinical and demographic data were collected from 6,368 adults with a clinical diagnosis of type 1 diabetes, as recorded in their primary care and eye screening records, who attended surveillance diabetes eye screening in southeast London between 2004 and 2018. The local general population of southeast London is ethnically diverse, with >30% of people being of African Caribbean heritage (7).

Exclusion criteria for this study included pregnancy, eGFR <45 mL/min/1.73 m² at baseline or documented history of non–diabetic kidney disease on hospital records. Following the exclusion of these individuals, 5,261 people with type 1 diabetes were eligible for analysis.

Anonymized diabetes-related clinical and biochemical data covering a time span between 2004 and 2018 were collected from electronic patient records at two large teaching hospitals in London. The following variables were available: date of birth, date of death (if applicable), sex, self-reported ethnicity, date of diabetes diagnosis, anthropometrics (weight and height), systolic and diastolic blood pressure, and laboratory measurements of serum creatinine, urine albumin-to-creatinine ratio (ACR), HbA_1c, LDL cholesterol, HDL cholesterol, total cholesterol, and triglycerides. We also collected diabetes eye screening status and results of retinopathy and maculopathy grade assessments (monoscopic fundus photography of dilated pupils with a nonmydriatic digital camera) using national diabetes eye screening classifications (8). To create our baseline data set, we considered the date of the first serum creatinine measurement for each patient as the initial date of entry into the study and extracted all other baseline values within a 2-year span. All other variables that had not been measured within that span were considered missing. The follow-up duration was defined as the last available creatinine measurement or date of death.

Serum creatinine measurements were used to calculate eGFR values according to the Chronic Kidney Disease Epidemiology Collaboration equation, with correction for African Caribbean ethnicity where applicable (9). Urinary ACR was captured from the routine clinical laboratory data. Albuminuria status at baseline was defined as normo-, micro-, or macroalbuminuric according to the ACR falling in the intervals 0–2.99 mg/mmol (A1), 3–29.99 mg/mmol (A2), or >30 mg/mmol (A3), respectively (10). Serum creatinine values that were measured during acute hospital admissions were excluded. Serum creatinine was measured at two central laboratories run by the same laboratory service provider using an isotope dilution mass spectrometry–traceable modified enzymatic method on the cobas c 702 platform (Roche, Basel, Switzerland). ACR was measured in urine samples taken from routine clinical care using immunoturbidimetry for albumin and an enzymatic method for creatinine on the cobas c 702 platform.

We measured socioeconomic status using the Index of Multiple Deprivation (IMD) derived from U.K. Office for National Statistics tables and based on the individual’s postal code. IMD scores were ranked according to population deciles, with 1 indicating the highest level of deprivation and 10 the most affluent (11). This retrospective study was conducted in line with local audit protocols using existing anonymized routine clinical data accessed directly by the clinical team approved by the hospital data governance committees.

The primary end point was a ≥50% reduction of eGFR from baseline with a final value <30 mL/min/1.73 m², which was confirmed on repeat testing. Secondary end points included a ≥30% and ≥40% reduction of eGFR from baseline, with the final eGFR <30 mL/min/1.73 m². In view of the close association between CKD and mortality, we used a Fine-Gray model to adjust for death from any cause as a competing risk event, as previously described (12). The Fine-Gray approach allows one to relate the covariate effect directly to and calculate the cumulative incidence function for each of the events (13) while taking competing risks into account. We also evaluated the incidence of ESKD (defined as a sustained eGFR <15 mL/min/1.73 m² or need for kidney replacement therapy) in our cohort.

The prognostic value of covariates was tested by univariable Cox regression and log-rank tests, and those with P < 0.2 were considered for the multivariable Cox regression models, unless they were deemed clinically relevant, in which case they were included in the multivariable analysis irrespective of statistical significance in the univariable model. To identify variables and combinations of variables with the best prognostic value, a stepwise backward selection procedure was followed. After univariable and multivariable Cox regression analysis, the effect of all-cause mortality was examined as a competing risk using Fine-Gray regression models. Follow-up was to the last available date in the electronic patient data set, 2018, date of death, or date of incident primary event, whichever occurred first.

Only variables with <30% missing within the baseline data set were included in the analysis. All missing baseline data for continuous variables were imputed by predictive mean matching using the “mice” package in R (RStudio version 1.3.959; R Foundation for Statistical Computing, Vienna, Austria). Analyses were performed in data sets with and without imputation of missing values.

For statistical comparisons of variables between CKD stages and ACR status, one-way ANOVA or Wilcoxon comparison tests were used for continuous variables and χ² tests for categorical variables. For computation and presentation of the models, we used R statistical functions coxphuni, coxphmulti, and crrmulti, all integrated in the “finalfit” statistical package (version 1.0.2). For assessing performance, validation, and calibration of the models, packages “survival” (version 3.2-11), “survminer” (version 0.4.9), and “rms” (version 6.2-0) were used. Proportional hazard assumptions were checked by graphs of scaled Schoenfeld residuals (Supplementary Material 1, Tables 1 and 2) displaying that those assumptions were not violated. For model validation, we used resampling validation, which provides bias-corrected indexes. (Supplementary Material 1, Figs. 1 and 2). All statistical analyses were done within RStudio version 1.3.1073 under R version 4.0.2.

Baseline characteristics of the study cohort (N = 5,261, 49.3% male) are detailed in Table 1. The median age and interquartile range (IQR) was 34 (26, 46) years, and duration of diabetes was 11 (2,22) years. The median age at time of diagnosis of diabetes was 22 (12, 34) years, with >80% of people diagnosed at age <40 years.

The predominant ethnicity was Caucasian (77.8%) versus African Caribbean (13.4%), Asian (2.9%), and other (5.9%). The median (IQR) HbA_1c was 68.3 (57.0, 86.9) mmol/mol, and systolic and diastolic blood pressure were 122.0 (113.0, 132.0) and 74.0 (68.0, 80.0) mmHg, respectively. Of the cohort, 49% had evidence of retinopathy (the majority of whom [>85%] had background changes [grade R1]), 11.2% had normoalbuminuria, 45.5% microalbuminuria, and 43.3% macroalbuminuria. The median eGFR was 90.7 (72.5, 111.4) mL/min/1.73 m², and serum creatinine levels were 73 (50, 96) μmol/L. Median follow-up was 8 years. We did not have full information on medications such as blood pressure medication and lipid-lowering drugs or any data on compliance with medications in the database. The median IMD decile of the cohort was 3 (2, 5).

Of the cohort, 263 (5%) individuals reached the primary end point of eGFR decline ≥50% from baseline with a final eGFR <30 mL/min/1.73 m². People who reached primary end point were more likely to be of African Caribbean ethnicity, to be older, and to have a longer duration of diabetes, higher systolic blood pressure, more prevalent retinopathy, albuminuria, and lower eGFR (all P < 0.05) (Table 2). Similar characteristics were observed for baseline values for the secondary end points of ≥30% and ≥40% eGFR decline from baseline with a final eGFR <30 mL/min/1.73 m² (Supplementary Material 2, Tables 1 and 2). We observed a significantly higher risk of people of African Caribbean ethnicity with type 1 diabetes compared with non–African Caribbean people reaching the primary end point with a faster decline in kidney function (Fig. 1).

In the multivariable Cox regression model of the primary outcome, African Caribbean ethnicity demonstrated a hazard ratio (HR) of 1.57 (95% CI 1.19, 2.08; P < 0.001) compared with other ethnicities and emerged as a significant risk factor for the primary end point. This effect was independent of established risk factors, such as age >60 years (HR 5.56 [95% CI 3.71, 8.34]; P < 0.001), HbA_1c (HR 1.02 [95% CI 1.01, 1.02]; P < 0.001), systolic blood pressure (HR 1.01 [95% CI 1.01, 1.02]; P < 0.001), and macroalbuminuria (HR 2.92 [95% CI 1.89, 4.53]; P < 0.001) (Table 3 and Fig. 2). We also observed a similar significant independent impact of African Caribbean ethnicity on secondary end points of a ≥40% decline in eGFR (HR 1.56 [95% CI 1.19, 2.05]; P = 0.001) and ≥30% decline in eGFR (1.61 [1.23, 2.11]; P = 0.001) with a final eGFR <30 mL/min/1.73 m² (Supplementary Material 2, Tables 3 and 4).

African Caribbean people with type 1 diabetes were younger and had a shorter duration of diabetes but similar levels of systolic and diastolic blood pressure and baseline eGFR compared with non–African Caribbean people (Supplementary Material 2, Table 5). Despite this, people of African Caribbean ethnicity with type 1 diabetes demonstrated significantly higher ACR and HbA_1c and a higher prevalence of macroalbuminuria (A3) at baseline than non–African Caribbean people.

We did not observe any significant impact of deprivation as determined by IMD score on the primary end point or secondary end points. We also did not observe any differences in IMD scores between African Caribbean and non–African Caribbean people (Supplementary Material 2, Table 5).

The incidence rate for the primary end point was 8.8 per 1,000 patient-years of follow-up. The incidence rate was more than double in African Caribbean people compared with non–African Caribbean people (16 vs. 7.7 events per 1,000 patient-years, P < 0.001).

In our cohort, ESKD (defined as sustained eGFR <15 mL/min/1.73 m² or need for kidney replacement therapy) developed in 133 (2.3%) people over 14 years of follow-up, with an incidence rate of 3.3 per 1,000 patient-years. We also observed that the incidence rate for ESKD was more than three times higher in African Caribbean people than in non–African Caribbean people (8 vs. 2.6 per 1,000 patient-years, P < 0.001).

Of the entire cohort, 250 people died before the end of follow-up and before reaching the primary end point. To confirm our observation, we also performed all the reported analyses on data sets, including all-cause mortality as a competing event. In these competing risk analyses, we observed similar consistent results, with a significant independent effect of African Caribbean ethnicity on the primary end point and secondary end points (Table 3, Fig. 2, and Supplementary Material 2, Tables 3 and 4 and Figs. 1–4).

We also repeated the analyses without the ethnicity correction for eGFR using the 2021 Chronic Kidney Disease Epidemiology Collaboration equation that was recently recommended to estimate kidney function without a race variable (14). In these further analyses, we consistently observed a significant independent effect of increased hazard of African Caribbean ethnicity on the primary end point (HR 1.57 [95% CI 1.19, 2.08]; P = 0.001) and secondary end points, including the competing risk analyses (HR 1.60 [95% CI 1.21, 2.11]; P = 0.001) (Supplementary Material 2, Tables 6–9).

In this study using longitudinal data from an ethnically diverse cohort of 5,261 people with type 1 diabetes, we demonstrate a robust independent relationship between African Caribbean ethnicity and significant decline in kidney function. More specifically, the study reveals the novel finding that African Caribbean ethnicity is a predictor of eGFR decline ≥50% from baseline with a final eGFR <30 mL/min/1.73 m² independent of traditional risk factors associated with kidney failure risk (e.g., age, duration of diabetes, systolic blood pressure, retinopathy, albuminuria) and socioeconomic status in a publicly funded health system (15–17). In secondary analyses, we also observed a similar significant independent impact of African Caribbean ethnicity on eGFR decline ≥30% and ≥40% from baseline with a final eGFR <30 mL/min/1.73 m². These data highlight the consistent impact of African Caribbean ethnicity on advanced and clinically important declines in kidney function.

Our cohort of people with type 1 diabetes were ethnically diverse, with 684 (13.3%) of African Caribbean descent. To our knowledge, no previous study has described the impact of ethnicity or race on kidney outcomes in type 1 diabetes. The majority of people we studied had micro- or macroalbuminuria, and nearly one-half had evidence of retinopathy and, hence, a possible higher risk phenotype than other recent cohorts that reported kidney outcomes in Caucasian people with type 1 diabetes (2,3). This finding may be partly due to several factors, such as possible selection bias of a cohort attending specialist hospital-based clinics, the use only one ACR sample result to ascertain albuminuria status, and differing selection criteria for national cohorts that may be more likely to include people with lower microvascular risk.

In a recent international collaborative study of 1,518 Caucasian people with type 1 diabetes and persistent macroalbuminuria (all with baseline eGFR >30 mL/min/1.73 m²) followed for 3–18 years, the authors reported that despite the use of renoprotective RAS inhibition, progression to ESKD was high. A total of 505 incident ESKD cases were noted (rate 32 per 1,000 patient-years), with higher rates observed in the U.S. than in Europe (3). In comparison, in our larger cohort, the rates of progression to the primary end point (≥50% fall in eGFR for a final eGFR <30 mL/min/1.73 m²) were 8.8 per 1,000 patient-years of follow-up and for ESKD, 3.3 per 1,000 patient-years. However, we observed that incidence rates for ESKD and our study primary end point were more than double in African Caribbean people compared with non–African Caribbean people.

In Caucasian people with type 1 diabetes and macroalbuminuria (A3), a rapid decline in kidney function has been observed and associated with a >50% loss of eGFR within 2–10 years (2). The exact pathophysiological mechanisms that explain this high-risk phenotype remain unclear because it was not influenced by sex, age at diabetes onset, or duration of diabetes but was associated with higher HbA_1c levels and more pronounced in those with higher levels of albuminuria (2). In our cohort of people who reached the primary end point, 30% had microalbuminuria at baseline, which demonstrates the continued importance of this risk predictor and highlights the importance of early identification and intensification of multifactorial treatment.

To date, no study in type 1 diabetes has described the impact of ethnicity on progression to kidney outcomes. In our multivariable regression analyses, African Caribbean ethnicity emerged as an independent risk factor for primary and secondary kidney outcomes along with more established risk factors that have been described previously, such as older age, HbA_1c, systolic blood pressure, and macroalbuminuria (A3).

When comparing different ethnicities, African Caribbean people with type 1 diabetes demonstrated significantly higher ACR, a greater prevalence of albuminuria (A3), and higher HbA_1c at baseline compared with non–African Caribbean people. However, we did not observe any significant differences in systolic or diastolic blood pressures, which are also important predictors of progression of kidney disease.

The mechanisms that explain our results of increased kidney risk in African Caribbean people with type 1 diabetes need to be further investigated. Our study was not designed to identify the putative reasons or mechanisms; however, we can speculate that there may be multiple explanations for our results. In studies of people with kidney disease due to type 2 diabetes or hypertension, people of African Caribbean ethnicity are at enhanced risk of kidney failure. Several hypothesis have been proposed as possible explanations for this higher risk as detailed below (18–22).

One potential factor is the impact of historical levels of blood glucose control. Several studies in adolescent and pediatric populations with type 1 diabetes have demonstrated higher HbA_1c in African Caribbean children with type 1 diabetes compared with their White counterparts, which persist after adjustment for multiple other factors, including socioeconomic status and diabetes duration (23–25). All these studies have been conducted in North America, in numerically smaller cohorts than ours, and were of relatively short duration. Hence, these studies were unable to assess the impact of ethnicity on advanced kidney or other microvascular complications. Several explanations have been proposed for these differences observed in children, including disparities in access to treatment, such as glucose monitoring technology; high rates of acute complications, such as hypoglycemia and ketoacidosis; and socioeconomic factors (18,23–27).

A recent study of adolescents with type 1 diabetes observed that heart rate variability and cardiorespiratory fitness were lower in Black than in non-Hispanic Caucasian participants (28). Interestingly, heart variability has been demonstrated to be an early predictor of early progressive kidney function decline in Caucasian adults with type 1 diabetes (29). People with albuminuria and type 1 diabetes are also more insulin resistant and have a greater propensity to salt sensitivity and related vascular injury (17). Mechanistic studies that evaluate whether these pathophysiological processes are evident or more prevalent in African Caribbean people with type 1 diabetes may also help to explain our findings.

Our study has several strengths and limitations. Strengths include its contemporaneous nature and ethnic diversity representative of a cohort of urban-dwelling adults with type 1 diabetes followed for a median of 8 years with real-world clinic-based measures and assessments. We were able to access a wider range of linked data, including both biomedical data and socioeconomic status. In contrast to many cohort studies, we had the advantage that all laboratory data were obtained from a single provider using standardized processes, which can limit variability.

Limitations of our study are that this urban-dwelling cohort of ethnically diverse people is unlikely to be representative of national cohorts or registries. Furthermore, there may be a selection bias of those with more advanced disease or complex diabetes attending specialist services. We report a high prevalence of albuminuria, which may be due to the above selection bias as well as to the imputation of missing ACR data for 25% of total cohort, as omissions may be higher in those with normal ACR. Our data for missing ACR numbers are comparable or even lower than recent national surveys in the U.K., where up to 40% of ACR data may be unavailable (4). Our median length of follow-up was 8 years; this is similar to other recent cohorts, but we acknowledge the need for longer follow-up studies to confirm our results.

We measured deprivation using nationally approved IMD deciles and did not observe a significant impact of IMD on kidney end points. Based on median IMD scores, our cohort was relatively deprived compared with other populations in the U.K. However, we acknowledge that more nuanced indices of socioeconomic and health care indices are needed to assess the impact of socioeconomic factors on the kidney outcomes we observed.

A major limitation of our study is the lack of information on medications, including RAS inhibition. These data, unfortunately, were not captured in the data set, which was derived from diabetes eye screening and linked hospital laboratory data. We cannot exclude that the observed results may be related to differences in use of medications, including renoprotective therapy such as RAS inhibitors.

In our cohort, baseline blood pressure was modestly higher in those with progression to kidney end points; however, we did not observe African Caribbean ethnicity as being associated with higher blood pressure at baseline. The prescribing patterns and, importantly, compliance of renoprotective and other medications are clearly vital areas for further study. Our findings in type 1 diabetes are similar to the results of a recent study where we observed in people with type 2 diabetes on RAS inhibition that a rapid and nonlinear loss of eGFR and fast progression to ESKD, independent of blood pressure, was prevalent in people of African Caribbean ethnicity (6).

We had no formal laboratory confirmation of diabetes subtype and relied on medical records and self-reporting of diabetes status when people attended for eye screening. It is possible that people with ketosis-prone diabetes may have been labeled with type 1 diabetes. However, the baseline data comparing people of African Caribbean ethnicity with those of non–African Caribbean ethnicity did not demonstrate any significant differences in features such as higher BMI, greater body weight, or older age, which may be more prevalent in the ketosis-prone phenotype (30).

It is increasingly apparent that there is significant heterogeneity in the underlying pathophysiology (2,6) and consequent risk of progression toward advanced CKD and ESKD. Recent publications aimed to estimate lifetime risk of ESKD in type 1 diabetes demonstrate significant geographical and national variations in ESKD risk. Our study focused more on a distinctive cohort living in an ethnically diverse urban setting. Such studies in high-risk groups can help to identify unique risk factors that may identify people at high risk of kidney events who may require earlier identification and optimization of modifiable risk factors with newer or more effective interventions (31).

In conclusion, in an ethnically diverse cohort of people with type 1 diabetes, we report the novel finding of African Caribbean ethnicity as a predictor of ≥50% eGFR loss, which was independent of traditional risk factors. Our findings establish the rationale for further research on the impact of ethnicity on onset and progression of kidney disease in type 1 diabetes. Further studies are needed to examine and understand the associated pathophysiology, mechanisms, and clinical features that may explain the association between African Caribbean ethnicity and increased risk of kidney disease in type 1 diabetes.

Funding. This work was funded by a research grant from Guy’s and St. Thomas Charity, London, U.K., grant JJ180101. S.A. was supported by the National Institute for Health Research Biomedical Research Centre based at Guy’s and St. Thomas NHS Foundation Trust and King’s College London.

The views expressed are those of the authors and not necessarily those of the National Health Service, National Institute of Health Research, or Department of Health.

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

Author Contributions. A.M., N.F., S.T., S.A., and J.K. designed the research study, collected and interpreted the data, and drafted the manuscript. J.C., P.V., A.C., S.H., D.H., and L.G. collected and interpreted the data and contributed to the manuscript. A.M. and S.A. contributed to and led the data analysis and interpretation. All authors reviewed the manuscript and approved the final draft. A.M. and J.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.