Risk of End-Stage Renal Disease in Patients With Systemic Lupus Erythematosus and Diabetes Mellitus: A Danish Nationwide Cohort Study

Abstract

Objective.

The risk of end-stage renal disease (ESRD) is increased in patients with systemic lupus erythematosus (SLE). This study was undertaken to determine whether diabetes mellitus (DM) increases ESRD risk in a large inception cohort of SLE patients.

Methods.

By means of the Danish National Patient Registry, we identified 3,178 adult patients diagnosed as having SLE between January 1, 1996, and July 31, 2018. DM was defined as the date of first hospital contact for DM or date of a first prescription of an antidiabetic drug. ESRD was defined as first registration of dialysis, renal transplant, or terminal renal insufficiency in the Danish National Patient Registry. ESRD incidence was compared between SLE patients with DM (SLE–DM) and those without DM (SLE–non-DM). Hazard ratios (HRs), adjusted for sex, age, educational level, and occupational status at baseline were calculated for sex, age, educational level, and hypertension (at baseline or during follow-up) strata. The overall hazard ratio (HR) was also adjusted for hypertension.

Results.

The SLE–DM group included 290 patients, of whom 77% were female, compared with 85% of the 2,859 patients in the SLE–non-DM group. SLE–DM patients had a 3 times higher risk of ESRD compared with SLE–non-DM patients (multivariable-adjusted HR 3.3 [95% confidence interval 1.8–6.1]). In stratified multivariable-adjusted analyses, DM increased the rate of ESRD in women and men, patients ≥50 years old at baseline, those with low educational level at baseline, and those with concomitant hypertension.

Conclusion.

Our findings indicate that SLE patients with DM have a markedly higher risk of developing ESRD compared with SLE patients without DM.

INTRODUCTION

End-stage renal disease (ESRD) is one of the most severe manifestations of systemic lupus erythematosus (SLE). Up to 40% of SLE patients develop chronic kidney disease (CKD) (lupus nephritis) over the course of the disease (1,2), and about 10% of those patients progress to ESRD (3,4). ESRD greatly reduces quality of life and increases both morbidity and mortality in SLE patients (5) as well as in the general population (6). The global prevalence of CKD is ~10% in the general population (6-8), and in most countries the leading cause of ESRD is diabetes mellitus (DM), accounting for up to 50% of ERSD cases in the general population (9). Improved screening, glycemic control, and new therapies for DM have reduced the risk of diabetic kidney disease (diabetic nephropathy). However, DM remains a global problem with its incidence increasing and many people living undiagnosed (7).

SLE is associated with the development of DM (10-12). During a 5-year follow up, ~10% of patients with incident SLE developed type 2 DM in a recent population-based study (13). Concurrent autoimmune diseases (including type 1 DM), metabolic syndrome, inflammation, and medications such as glucocorticoids may contribute to a higher prevalence of DM in SLE patients compared with the general population (14,15).

The pathogenic mechanisms of ESRD development in SLE and DM are different (16), and whether DM further increases the risk of ESRD in SLE has not yet been established. Another significant cause of ESRD in the general population is hypertension (8). Besides frequently coexisting with DM and synergistically worsening outcomes due to vascular disease, hypertension is also one of the most frequent comorbidities in SLE patients (10,11,17,18). In this study in a nationwide cohort of Danish SLE patients, we investigated whether DM is associated with an increased risk of ESRD and whether hypertension modifies this association.

PATIENTS AND METHODS

Data sources.

In this population-based study, we used prospectively collected national registry data. The Danish Civil Registration System was established in 1968; it assigns a central person registration number, issued by law, at birth, or upon immigration to all persons residing in Denmark, and contains continuously updated information regarding vital status (19). The following registries were linked using central person registration numbers.

The Danish National Patient Registry contains International Classification of Diseases (ICD) codes of diagnoses and procedures recorded by physicians during discharges from inpatient wards (since 1977), emergency wards (since 1995), and outpatient hospital clinics (since 1995) (20,21). The data are organized as patient “contacts,” wherein inpatient and emergency care contacts always contain only 1 visit and include a start and end date. However, individual outpatient contacts are initially assigned a start date and may contain 1 or several follow-up visits under the same primary diagnosis. For each patient care contact, ≤20 diagnoses can be registered, including a primary diagnosis and any secondary, optional diagnoses. ICD, Eighth Revision (ICD-8) codes were used from 1977 until 1993, followed by ICD, Tenth Revision (ICD-10) codes since 1994.

The Danish National Prescription Registry was established in 1995 and contains all prescription dispensing data from Danish community pharmacies (21). The Income Statistics Registry and the Population’s Education Registry contain data concerning the occupational status (affiliation to the labor market) and educational status, respectively, of Danish residents (21).

Study population.

We established a national SLE inception cohort comprising patients ≥18 years old with a first-time Danish National Patient Registry registration of SLE between January 1, 1996, and July 31, 2018. Registration codes used to define SLE included 73419 (ICD-8) as well as M32.1, M32.8, and M32.9 (ICD-10). We required ≥1 SLE ICD code (primary or secondary) registered at a rheumatology, nephrology, or dermatology department (inpatient or outpatient) or a primary code registered at any inpatient department. Baseline was the date when this definition was fulfilled. Using a previously reported data set for validation of SLE registration in the Danish National Patient Register (22), we identified 194 individuals fulfilling our case definition. Among these, 145 (75%) fulfilled the American College of Rheumatology classification criteria for SLE (23,24) and 160 (83%) had a physician-based clinical diagnosis of SLE.

Patients with prevalent ESRD at baseline were excluded (n = 29). The positive predictive value of registration of moderate-to-severe renal disease in the Danish National Patient Register is 100% (25).

Study variables and outcomes.

Outcome and follow-up.

We defined incident ESRD as the first registration of terminal (stage 5) kidney insufficiency, chronic dialysis, or renal transplant according to ICD diagnosis and procedure codes (see Supplementary Table 1, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.25091). Individuals contributed person-time until ESRD, death, emigration, or July 31, 2018, whichever occurred first.

Clinical factors: DM and hypertension.

To define DM (both types 1 and 2) and hypertension, we used the first relevant ICD codes in the Danish National Patient Register or the date of the first filled prescription (available from 1995–2018) for these conditions, whichever came first (see Supplementary Table 1, at onlinelibrary.wiley.com/doi/10.1002/acr.25091). The primary exposure, DM, was a time-dependent factor. When categorized, DM was presented as prevalent at baseline, incident during follow-up, and no DM. The exposure variable (DM) was considered a risk factor for developing ESRD, and hypertension was regarded as a clinical factor that could be an effect modifier. In patients with incident DM, hypertension was considered “present” if it was prevalent at baseline or diagnosed prior to incident DM. In the time-dependent analyses, both DM and hypertension are covariates considered absent until they were diagnosed.

Parameters of socioeconomic status.

The highest level of education achieved and occupational status at baseline were used as socioeconomic factors. Education was divided into 5 categories based on International Standard Classification of Education (ISCED) 2011 classification: 1) early childhood education, primary, and lower secondary education (ISCED levels 0–2); 2) general upper secondary education, high school (ISCED level 3); 3) vocational upper secondary education (ISCED level 4); 4) short- or medium-length tertiary education, bachelor, or equivalent (ISCED levels 5 and 6); and 5) long-length higher education or master’s degree, doctorate, or equivalent (ISCED levels 7 and 8) (26).

For adjustment purposes, all 5 educational level categories were used. When stratifying, educational-level categories were dichotomized as primary and lower secondary education (1) compared with upper secondary and higher education (2-5). In the educational-level stratified analyses, patients with missing data regarding educational level were excluded. In the adjusted analyses, “missing” was considered to be a separate educational level category.

Occupational status was categorized as being 1) affiliated to the labor market, 2) under education, 3) retired, and 4) unemployed or on welfare. Information regarding occupational status was available through 2016; thus, for patients with baseline visit dates in 2017 or 2018, we carried forward their 2016 data.

Data and statistical analysis.

Incidence of ESRD.

We estimated the crude incidence rate (IR) of ESRD per 1,000 person-years as the number of ESRD cases divided by DM-exposed and DM-unexposed person-years at risk separately. In those with incident DM during follow-up, person-time between baseline and date of incident DM was classified as non-DM person-time. We calculated crude IR ratios (IRRs) and corresponding 95% confidence intervals (95% CIs), stratified by sex, age, educational level (at baseline), and hypertension, and we used tests of homogeneity to assess differences between strata.

Multivariable-adjusted analyses.

Multivariable-adjusted Cox proportional hazards models were used to estimate hazard ratios (HRs) for ESRD relative to the time-dependent exposure of DM, adjusting for confounders that were determined based on a priori knowledge of the DM–ESRD association. The models were then stratified by sex, age at baseline (<50 years versus ≥50 years), educational level category at baseline, and hypertension (prevalent at baseline or incident prior to incident DM versus no hypertension). All HRs were adjusted for sex, age, educational level, and occupational status at baseline. Further, the overall HR was also adjusted for hypertension at baseline or prior to incident DM. Both DM and hypertension were time dependent in these analyses.

To evaluate if further development of hypertension could act as an effect modifier, multivariable-adjusted Cox modeling was extended to also include a time-dependent interaction term between DM and hypertension. To account for the possible competing risk of death (death is a competing risk for ESRD because ESRD can no longer occur if a patient dies), we used 2 different Cox proportional hazards models: cause-specific hazards and cumulative incidence (proportional subdistributional hazards) (27). Data analyses were performed using SAS software version 9.4.

Ethics and approvals.

The study was approved by the Danish Data Protection Agency (approval no. VD-2018-175). Informed participant consent was not required due to the anonymized, registry-based data retrieval provided by Statistics Denmark.

RESULTS

Baseline characteristics.

In this cohort of SLE patients, 290 had DM at baseline or developed DM during follow-up, whereas 2,859 did not have DM. Those with DM were more likely to be men (23% versus 15%) and older (mean age at baseline 53 versus 46 years) compared with the group of SLE patients without DM (Table 1). Among the SLE patients with DM, 119 (41%) already had a DM diagnosis at baseline. In the remaining 171 SLE patients (59%), DM presented during follow-up a mean ± SD of 6.4 ± 5.8 years after baseline. Hypertension was more common at baseline in those with DM compared with patients who did not have DM (36% versus 23%).

Table 1.

Characteristics of systemic lupus erythematosus (SLE) patients stratified by concomitant diabetes mellitus (DM)^*

	DM (n = 290)	No DM (n = 2,859)
Female sex	224 (77.2)	2,441 (85.4)
Age, mean ± SD
At SLE diagnosis	53.4 ± 14.8	46.2 ± 16.4
At DM diagnosis	54.3 ± 15.8	NA
Occupational status at baseline
Affiliation to the labor market	107 (36.9)	1,378 (48.2)
Under education	3 (1.0)	198 (6.9)
Unemployed or on welfare	51 (17.6)	489 (17.1)
Retired	128 (44.1)	790 (27.6)
Unknown or missing data	1 (0.3)	4 (0.1)
Highest educational level at baseline
Primary and lower secondary	114 (39.3)	844 (29.5)
General upper secondary (high school)	14 (4.8)	238 (8.3)
Vocational upper secondary	92 (31.7)	889 (31.1)
Short- or medium-length tertiary	42 (14.5)	640 (22.4)
Long-length higher (master’s or doctorate)	15 (5.2)	143 (5.0)
Unknown or missing data	13 (4.5)	105 (3.7)
Hypertension
At baseline	105 (36.2)	668 (23.4)
During follow-up†	54 (18.6)	1,215 (42.5)
At baseline or during follow-up	159 (54.8)	1,883 (65.9)
ESRD during follow-up	13	63

^*Except where indicated otherwise, values are the number (%) of patients. NA = not applicable.

^†Prior to incident DM (excludes hypertension following incident DM); in patients with end-stage renal disease (ESRD), incident before ESRD.

Incidence of ESRD.

There were 13 cases of incident ESRD during 1,672 person-years at risk in the SLE–DM group (7.8 cases per 1,000 person-years) and 63 ESRD cases during 26,905 person-years at risk in the SLE–non-DM group (2.3 cases per 1,000 person-years). The overall ESRD incidence rate among SLE patients was 2.7 (95% CI 2.1–3.3) cases per 1,000 person-years. ESRD occurred more frequently in SLE patients with DM compared with those without DM, irrespective of stratification for age, sex, or education (Table 2). At baseline or before incident DM, the unadjusted association between DM and ESRD was higher in SLE patients with hypertension (IRR 5.8 [95% confidence interval (95% CI) 2.6–13.0]) compared with those without hypertension (IRR 1.8 [95% CI 0.74–4.35]) (Table 2).

Table 2.

Measures of occurrence and association of ESRD (per 1,000 person-years) in SLE patients stratified by concomitant DM and by age, sex, educational level, and hypertension^*

	DM			No DM
	No. ESRD/ total no.	Person-years	IR (95% CI)	No. ESRD/ total no.	Person-years	IR (95% CI)	IRR (95% CI)	P †
Age at baseline								0.405
≥50 years	8/172	843	9.5 (4.7–19.0)	22/1,181	9,824	2.2 (1.5–3.4)	4.2 (1.9–9.5)
<50 years	5/118	829	6.0 (2.5–14.5)	41/1,678	17,081	2.4 (1.8–3.3)	2.5 (1.0–6.4)
Sex								0.612
Male	6/66	324	18.5 (8.3–41.2)	17/418	3,350	5.1 (3.1–8.2)	3.6 (1.4–9.3)
Female	7/224	1,348	5.2 (2.5–10.9)	46/2,441	23,546	1.9 (1.5–2.6)	2.7 (1.2–5.9
Educational level at baseline‡								0.278
Primary and lower secondary	8/114	595	13.4 (6.7–26.9)	28/844	8,457	3.3 (2.3–4.8)	4.1 (1.8–8.9)
Upper secondary and higher	4/163	1,043	3.8 (1.4–10.2)	34/1,910	17,619	1.9 (1.4–2.7)	2.0 (0.7–5.6)
Hypertension at baseline or prior to incident DM								0.048
Yes	7/159	618	11.3 (5.4–23.8)	36/1,883	18,378	2.0 (1.4–2.7)	5.8 (2.6–13.0)
No	6/131	1,066	5.7 (2.5–12.7)	27/976	8,528	3.2 (2.2–4.6)	1.8 (0.7-4.4)

^*95% CI = 95% confidence interval; DM = diabetes mellitus; ESRD = end-stage renal disease; IR = incidence rate; SLE = systemic lupus erythematosus.

^†P value for the test of homogeneity of incidence rate ratio (IRR) in each stratum.

^‡Patients with unknown or missing education level were excluded (n = 118).

Multivariable-adjusted analyses.

SLE patients with DM had a 3-fold higher rate of ESRD compared with SLE patients without DM (multivariable [sex, age, hypertension, education, and occupational status at baseline] adjusted HR 3.3 [95% CI 1.8–6.2]) (Table 3). In stratified, multivariable-adjusted Cox proportional hazards models, DM increased the rate of ESRD in women and men, individuals in the older age group at baseline, those with low educational level at baseline, and those with hypertension at baseline or during follow-up (incident prior to incident DM) (Table 3). Although the highest incidence rate for ESRD was found in SLE patients with both DM and a history of hypertension (Table 2), no time-dependent interaction between DM and hypertension was observed in the adjusted Cox proportional hazards models (HR 0.6 [95% CI 0.2–1.9]).

Table 3.

Hazard ratios (HRs) of ESRD in SLE patients with DM, stratified according to age at baseline, sex, education category at baseline, and hypertension^*

	HR	95% CI
Overall, multivariable-adjusted†	3.3	1.8–6.2
Overall, including adjustment for hypertension†	3.3	1.8–6.1
Age at baseline‡
≥50 years	4.0	1.8–9.0
<50 years	2.6	0.99–6.7
Sex§
Male	3.3	1.2–8.9
Female	3.2	1.4–7.2
Educational level at baseline¶
Primary and lower secondary	3.8	1.7–8.4
Upper secondary and higher	2.2	0.8–6.5
Hypertension at baseline or prior to incident DM#
Yes	4.9	2.1–11.3
No	2.2	0.9–5.9

^*Diabetes mellitus (DM) is time dependent; hypertension is time dependent when adjusting. 95% CI = 95% confidence interval; ESRD = end-stage renal disease; SLE = systemic lupus erythematosus.

^†Adjusted for sex, age, education, and occupational status at baseline.

^‡Adjusted for sex, education, and occupational status at baseline.

^§Adjusted for age, education, and occupational status at baseline.

^¶Patients with unknown or missing educational level excluded (n = 118). Adjusted for sex, age, and occupational status at baseline.

^#Hypertension was not time dependent in this stratum. Adjusted for sex, age, education, and occupational status at baseline.

ESRD was associated with DM in SLE patients using either a model of cause-specific hazard by treating competing risks (death) as a censored observation (HR 2.9 [95% CI 1.6–5.3]) or a model based on proportional subdistribution hazard ratios (cumulative incidence; HR 2.4 [95% CI 1.3–4.3]). Thus, the observed increased incidence of ESRD in SLE–DM patients could not be explained by competing risks.

DISCUSSION

In this nationwide cohort of 3,178 patients with incident SLE, we found that the SLE patients with DM developed ESRD at a 3-fold higher rate compared with those without DM. It is well known that DM and autoimmune diseases, such as SLE, are risk factors for developing CKD and ESRD (1,3,4,28). Few studies have shown an increased risk of ESRD in SLE patients with DM. In a cohort study of 601 adults with biopsy-proven glomerulonephritis, which included individuals with lupus nephritis (21%), DM was independently associated with a 2-fold greater risk of ESRD during a median of 39 months of follow-up (29). In a study of 1,317 patients with SLE and incident DM, matched with 1,317 patients with SLE and without DM, the crude HR for ESRD by DM was 2.7; the HR was reduced to 1.6 when adjusted for confounders such as age, sex, and comorbidities, including hypertension (30). Thus, the slightly varying estimates of ESRD risk by DM could partly be attributable to varying definitions and incidences of DM and hypertension in the reported cohorts.

In our study, the rate of ESRD was highest among SLE–DM patients in the lowest educational stratum, which in part could be related to lower educational level being reported as a determinant of nonadherence to antimalarials in SLE (31). Adherence to antimalarials in SLE has been found to be protective of type 2 DM (13) as well as renal damage (32).

The highest incidence rate of ESRD was observed in male SLE–DM patients, but the multivariable-adjusted HRs were nearly equivalent in women and men. Although male sex in some studies has been associated with poor renal outcome in SLE (33,34), this is not a general finding (35) and in diabetic nephropathy, the risk of ESRD does not appear to be sex dependent (36).

We also found that DM was associated with ESRD in the older group at baseline. Whether this is due to a longer period of DM exposure, even before SLE diagnosis, or other factors, is unknown. The durations of DM and hypertension—prior to or following SLE diagnosis—are also likely to influence the risk of developing lupus nephritis and ESRD. Our study design could not address this accurately since the start of follow-up begins with the SLE case definition and not DM/hypertension definition.

The most frequent cause of kidney damage in SLE is lupus nephritis (glomerulonephritis). In a recent review, 7–31% of patients had lupus nephritis at SLE diagnosis; 31–48% developed lupus nephritis after SLE diagnosis, most within 5 years (37). Less frequent causes of ESRD in SLE include thrombotic microangiopathy/antiphospholipid nephropathy, non–immune complex podocytopathy, tubulointerstitial nephritis, acute tubular necrosis, renovascular disease, or nephrotoxicity from medications (38). Recently, a case of coexistence of diabetic nephropathy and lupus-related renal manifestation, lupus podocytopathy, was reported (39).

Patients with lupus nephritis often receive aggressive treatment with immunosuppressive medications and high doses of glucocorticoids. It is well known that glucocorticoid use is associated with various complications, including DM (40). Thus, in some SLE patients in our study, DM might have developed as a consequence of lupus nephritis treatment and not been related to the development of ESRD. Some ESRD cases in our study could also have been caused or exacerbated by other, less frequent, lupus-related renal manifestations, as mentioned above.

In this study, we aimed to evaluate the association between DM and ESRD in SLE in general. To what extent DM is an independent risk factor or has a synergistic effect with lupus nephritis or other possible causes of ESRD, leading to a progression from CKD to ESRD is yet to be investigated. Due to the register-based nature of the study, we were not able specifically to address relevant clinical descriptors of lupus nephritis. Future studies evaluating the potential influence of renal changes as determined by histologic changes on the association between diabetes and hypertension in SLE patients are warranted.

In our study, concomitant hypertension was associated with a higher risk of ESRD in SLE–DM patients compared with SLE–non-DM patients. Hypertension is currently the second leading cause of ESRD in the general population (following DM) (6) and also a potential adverse event in glucocorticoid use (40). Some studies report an additional effect of DM (diabetic nephropathy) and hypertension (hypertensive nephrosclerosis) on the risk of developing ESRD in the general population, due to effects on the macrovasculature and microvasculature (8,18). In our study, adjustment for hypertension did not change the HR for development of ESRD, nor did we find an interaction between DM and hypertension on the risk of ESRD in SLE patients.

We did not have access to any blood pressure measurements, laboratory results, or kidney biopsy results in our registry-based study and were thus not able to assess effectiveness of DM and hypertension management or specify the histologic patterns in the kidneys. African ancestry is associated with a markedly increased risk of ESRD in SLE patients (41,42), hence evaluating race as a possible effect modifier may have been of interest. However, this ancestral group constitutes only 1% of SLE patients in Denmark (43), and therefore our study was not sufficiently powered for such an analysis.

Strengths of this study include the large inception cohort of SLE patients with prospectively collected, population-based longitudinal data, including inpatient and outpatient diagnoses, and medication data. The positive predictive value of registration of moderate/severe renal disease and DM in the Danish National Patient Register is high (100% and 96%, respectively) (25). Use of medication prescription data in addition to the ICD codes reduced misclassification in defining DM and hypertension.

In our study we found that DM is a significant risk factor for ESRD in SLE. Whether it is independent or related to specific SLE phenotypes, other autoimmunity, or treatment remains to be investigated. Our study also emphasizes the need for the assessment and treatment of DM and hypertension in the care of patients with SLE. Further studies are warranted to explore how the risk of ESRD is influenced by glomerulonephritis, DM, and hypertension in SLE patients.

SIGNIFICANCE & INNOVATIONS.

• Both diabetes mellitus (DM) and systemic lupus erythematosus (SLE) are associated with developing end-stage renal disease (ESRD).

• The risk of developing ESRD is 3 times higher in patients with SLE and DM compared with SLE patients without DM.

• DM increases the rate of ESRD in female and male SLE patients ≥50 years old at baseline, those with low educational level at baseline, and those with concomitant hypertension.

• Future studies are warranted to further explore how the risk of developing ESRD is influenced by lupus nephritis, DM, and hypertension in SLE patients.