The prevalence of diabetes was reported by the World Health Organization from 180 million in 1980 to 422 million in 2014, differentiating those patients who have diabetes but it is not the cause of kidney damage and is actually in the presence of non-diabetic renal disease (NDRD) or who, in addition to having diabetic nephropathy, simultaneously suffer from another illness that is aggravating kidney function and is susceptible to a therapeutic intervention other than glycemic control that allows improving renal survival. 

The present analysis of 201 biopsies from native kidneys in diabetes mellitus patients from the pathology department of the University of Antioquia at the San Vicente Hospital clinical laboratory was a retrospective cohort study. The kidney biopsy report that 41% of patients had diabetic nephropathy, 16% mixed, and 43% NDRD, the most frequent histological finding was focal segmental glomerulosclerosis; in the univariate and multivariate analysis, two independent predictors were identified, each year above the mean age (56 years) increases the risk of presenting NDRD (OR, 1.05; 95% CI, 1.02–1.09; p = 0.002) in the KB. Diabetic retinopathy significantly decreases the occurrence of NDRD (OR, 0.23; 95% CI, 0.09–0.60; p = 0.002). Our findings on the potential predictive strategies, the model with the clinical variables age, diabetic retinopathy, and time of diabetes offered the best predictive performance. The area under the discrimination curve was 0.75 (95% CI, 0.67-0.81) with an acceptable Hosmer Lemeshow test, and calibration can be useful when deciding whether to perform a kidney biopsy.

The dramatic increase in the prevalence of diabetes reported by the World Health Organization from 180 million in 1980 to 422 million in 2014, plus the high projected growth for 2050 in low and middle-income countries (1, 2) and the association with cardiovascular outcomes, chronic kidney disease (CKD) and mortality (3,4,5), support the interest in differentiating those patients who have diabetes. Still, it is not the leading cause of kidney damage among those with non-diabetic renal disease (NDRD) or who, in addition to diabetic nephropathy, simultaneously suffer from other illnesses that are aggravating kidney function and are susceptible to a therapeutic intervention other than glycemic control that allows improvement in renal outcomes (6,7).

To be able to discriminate diabetic kidney disease, which is a clinical diagnosis, from diabetic nephropathy (DN) alone or combined with another kidney pathology (mixed), is only possible by a kidney biopsy (KB) (8,9). Still, given the risk of the procedure, the clinician’s skill is required to define the patients who benefit from an invasive diagnostic test. The indications described in the literature for type 1 diabetes mellitus include microhematuria, absence of diabetic retinopathy, unusual alteration of renal function or immunological alterations (10), and for type 2 diabetes mellitus are the sudden onset of proteinuria, proteinuria in the absence of diabetic retinopathy, active urine sediment, rapidly declining kidney function, and diabetes less than ten years (11). A meta-analysis that included 48 studies evaluating KB results in diabetic patients showed a wide range of prevalence of DN (6.5-94%) versus NDRD (3-82.9%) and mixed (4 -45.5%) in their analysis.  The most frequent histopathological findings are Focal Segmental Glomerulosclerosis (FSGS), IgA Nephropathy (IgAN), and acute tubulointerstitial nephritis (ATIN) (12). Therefore, there is still a lack of consensus on when to perform KB in this population. We aimed to identify the possible clinical and laboratory factors in diabetic patients associated with the occurrence of NDRD in KB. 

Methods

Design and patients

From January 2011 to February 2022, the pathology department of the University of Antioquia at the San Vicente Hospital received a total of 6780 kidney biopsies, of which 201 biopsies were of patients ≥14 years old with native kidneys and diagnosis of diabetes mellitus, that were included in this retrospective cohort study. The indications for KB included acute presentations with persistent renal impairment following acute kidney injury or non-acute presentations with atypical clinical features including (1) sub-nephrotic or nephrotic-range proteinuria or nephrotic syndrome and (2) the presence of microscopic hematuria; or (3) rapid progressive chronic kidney disease.

Data Collection

Data were collected from hospital records and the KB registry at the pathology department; final reports were checked individually. All biopsy reports included the results examined under light microscopy (stained with hematoxylin & eosin, periodic acid-Schiff, Masson’s trichrome, and Jones methenamine silver and with other histochemical stains Congo red if was necessary) and immunofluorescence (for IgA, IgG, IgM, C3, C1q, κ, and λ). Still, only 18.4% had electron microscopy to examine glomerular basement membrane thickness or clarify selected cases. Non-sclerosed and sclerosed glomeruli were counted to ascertain the degree of scarring. Glomeruli with global sclerosis and glomeruli with segmental lesions were quantified as percentages of total glomeruli or viable glomeruli, respectively. IFTA scores were classified according to the estimated rate seen in the cortical area of the biopsy sample as follows: absent (grade 0) as 0%; mild (grade 1) <25>50% of the total area (13). DN was diagnosed and graded according to the Renal Pathology Society classification in two groups: 1) combine grades 1 and 2 described as early DN, and 2) grades 3 and 4 Kimmelstiel-Wilson nodules described as advanced DN. (14)

Clinical variables:  we collected patients’ demographic information (age, sex), prespecified laboratory and clinical variables, duration of diabetes, and retinopathy status at the biopsy. Baseline renal function was recorded using the serum creatinine measurements at least three months before the KB. Glomerular filtration rate was estimated using the Chronic Kidney Disease Epidemiology Collaboration method (CKD-EPI) (15). Retinal status was recorded from clinical records at the last assessment. We also follow the previously available serum creatinine before End-Stage Kidney Disease (ESKD) or death. Our main objectives of interest were a) to identify the possible clinical and laboratory association factors of diabetic patients for the appearance of NDRD in KB; b) To compare patient and ESKD (defined as the need for renal replacement therapy) in patients with diabetic versus nondiabetic renal disease; c) Identify risk factors associated with all-cause mortality in the cohort. Patients were followed from the time of renal biopsy until study endpoints or February 2022. The following clinical definitions were used nephrotic-range proteinuria,24-hour urine protein >3500 mg/d; nephrotic syndrome as defined by the KDIGO guidelines (16), 24-hour urine protein >3500 mg/d plus hypoalbuminemia (less than 3 mg/dl) plus hyperlipidemia and edema; hematuria, >5 red blood cells per high-power field; pyuria, >5 white blood cells per high-power field (17); acute kidney disease according to KDIGO guidelines (18), >0.3 mg/dl baseline serum creatinine in less than three months.

Statistical analysis

Qualitative variables were compared using Fisher's test or chi-square as appropriate. Continuous variables depend on whether they followed a normal distribution with means and standard deviations and those did not differ by median and interquartile range. To compare quantitative variables, Student's t-test was used for variables that followed normality and Wilcoxon's t-test for those that did not follow normality.

A logistic regression analysis was carried out for the primary outcome. The dependent variable was NDRD, and the following variables were entered into the model, conforming to their behavior in the univariate analysis. According to the background in the literature (age, glomerular filtration rate, hematuria, retinopathy, and albuminuria) for this, a stepwise logistic regression model was assessed with the clinical variables of interest, and those that in the univariate analysis had a lower p-value of 0.25. 

For the second aim, we performed a Kaplan-Meier survival curve using as start date, the time of the biopsy until the ESKD or death from all causes or censoring by the last follow up, comparing DN with NDRD by log-rank test. A Cox proportional hazards regression analysis was performed for the third objective and identified possible factors associated with renal survival and patient survival using the time of biopsy as the start date until death from all causes for patient survival or ESKD. The variables in the univariate analysis had a p of less than 0.25. According to the literature for the event's development, clinically relevant ones were entered into the Cox regression model. Finally, several prediction models were proposed in an exploratory manner through various logistic regression analyses. The outcome variable was defined as the final biopsy result of any NDRD or mixed versus DN alone in all cases. The independent variables were initially included according to a previous literature review. The variables were selected by a step-by-step method using a significance level of p < 0>

Figure 1: Sampling methodology for patients with kidney biopsy diagnosed with diabetes mellitus from January 2011 to February 2022 at the Department of Pathology Antioquia at University and San Vicente Hospital.

Results

A total of 201 patients diagnosed with diabetes mellitus older than 14 years underwent KB during the study period from January 2011 to February 2022 (fig 1). The mean ± SD age was 56 ± 13 years, the median diabetes mellitus duration 7 (IQR 3 – 13) years, with glycosylated hemoglobin (HbA1c) 7.2(6,4-9,0), and the median serum creatinine was 2 mg/dl, and retinopathy was present in 33% of the cohort.  The KB reports that 41% of patients had DN, 16% mixed, and 43% NDRD. 84% of the patients had high blood pressure, and 76% received Angiotensin-Converting Enzyme (ACEIs) or Inhibitors Angiotensin II receptor blockers (ARBs). The 3 most common causes leading to biopsy were nephrotic syndrome 36.3% (n=73), proteinuria 30.3% (n=61) and acute kidney injury 18.9% (n=38), table 1.

Histopathological findings 

Of the eighty-three DN alone, 92.7% (n=77) present advanced DN classified as grades 3 and 4 containing Kimmelstiel-Wilson nodules, and only 50.6% had diabetic retinopathy. The most frequent histological finding in the KB from the NDRD patients was focal segmental glomerulosclerosis (FSGS), found in twenty-three patients (26.7%), followed by IgA nephropathy. Mixed (NDRD plus DN) most prevalent was tubulointerstitial nephritis with fourteen (45%) individuals, and just six patients (18.7%) had a glomerulopathy in this subgroup, see table 2.

Table 1: Descriptive analysis of patients diagnosed with diabetes mellitus with a kidney biopsy.

Table 2. Histopathological findings in patients with non-diabetic renal disease and diabetic nephropathy plus another diagnosis.

Factors associated with non-diabetic kidney disease in kidney biopsy

Two independent predictors were identified in the univariate and multivariate analysis; each year above the mean age (56 years) increases the risk of presenting NDRD (OR, 1.05; 95% CI, 1.02–1.09; p = 0.002) in the KB. The diabetic retinopathy significantly decreases the occurrence of NDRD (OR, 0.23; 95% CI, 0.09–0.60; p = 0.002), see table 3.

Table 3. Univariate and multivariate logistic regression analysis of diabetic nephropathy versus non-diabetic renal disease. Regression with all continuous quantitative variables (albumin and proteinuria do not show collinearity).

Patient and renal-censored survival diabetes nephropathy versus non-diabetic renal disease 

Fifty-seven (28.3%) patients died during the follow-up, the most common cause was infection (n=30, 52.6%), continued by cardiovascular cause (n=16, 28%). The patient survival with DN rate was 52%, and for NDRD, 60% in the first year, 25% and 28% in the third year, and 12,5% and 15% in the fifth year, respectively, figure 2a.  The renal survival with DN rate was 40% and for NDRD plus mixed 38% in the first year, figure 2b. There was no significant difference between patients and renal survival if they presented DN or NDRD.

Figure 2. (a) Patient survival rates for diabetic nephropathy versus mixed (diabetic nephropathy plus non-diabetic renal disease). (b) Renal survival rates diabetic nephropathy versus mixed (diabetic nephropathy plus non-diabetic renal disease)

Patient and renal-censored survival associated factors.

Univariable and multivariate Cox proportional hazards regression models were performed to estimate the adjusted risk for lower patient and renal survival. Patient survival multivariate analysis showed albumin >3 g/dL increases the survival of overall patients that were biopsied; HR, 0.29 (95% CI, 0.12 – 0.68; p = 0.005), as does having proteinuria of less than 3.5 g/24 hours; HR, 0.37 (95% CI, 0.15 – 0.95; p = 0.030). On the contrary, age over 56 years decreases patient survival, HR, 1.04 (95% CI, 1.01–1.06; p = 0.05). These findings were similar in the subgroup of only type 2 diabetic patients. The difference that in this population, not having hematuria was also related to better survival, HR, 0.37 (95% CI, 0.15 – 0.95; p = 0.030). Multivariate analysis showed a marked association with the decreased renal survival associated with diabetic retinopathy, HR, 3.37 (95% CI, 1.36 – 8.33; p = 0.009).

Table 4. Univariable and multivariable Cox analysis for overall patient survival and renal survival.

During the exploration of potential predictive strategies, the model with the clinical variables age, diabetic retinopathy, and time of diabetes offered the best predictive performance. The area under the discrimination curve was 0.75 (95% CI, 0.67-0.81) with an acceptable Hosmer Lemeshow test and calibration plot, figure 3.

Figure 3. Predictive strategies, a model with clinical variables age, diabetic retinopathy, and time of diabetes

We present our retrospective cohort study in a reference pathology center of diabetic patients who underwent KB in the last decade. The most frequent indication for KB was proteinuria, whether in the nephrotic range or not. In general, KB was a safe procedure, with just 3.98% minor complications and no death caused by it. 

NDRD in patients with diabetes has a wide range in prevalence (12,19-23); this phenomenon is perhaps due to the lack of consensus on the indications for KB from each center (24). In our cohort, the prevalence of NDRD alone was 43%, like those found in the largest cohorts in Europe and the USA, 49.6% and 35.4%, respectively (25,26). Focal segmental glomerulosclerosis was the most prevalent glomerulopathy within the NDRD alone in our study, followed by ATIN, IgAN, and membranous nephropathy, which differs from previous reports literature, in which IgAN was the main finding (27-34). In a study previously carried out by our group that included the review of glomerular disease in 1040 kidney biopsies in the general population, this same distribution of prevalence was also found, which is a similar tendency in Latin America and black race (35). These results are important because they are pathologies susceptible to other treatments that impact better kidney outcomes (36,37).

In the logistic regression model for the occurrence of NDRD, we identified as associated factors the older age (OR, 1.05; 95% CI, 1.02–1.09; p = 0.002) and the absence of diabetic retinopathy (OR, 0.23; 95% CI, 0.09–0.60; p = 0.002), this data is also support by the meta-analysis published by Liang et al. (38), who reported an inversely proportional relationship between the presence of retinopathy and the diagnosis of NDRD. Many researchers have long considered diabetic retinopathy a clinical characteristic of advanced diabetes that would rule out the need for KB in diabetic patients with renal deterioration signs (39,40). But older age (>56 years) has not been previously described as an associated factor as far as we know from the review carried out by our group.

The Cox analysis showed that a normal albumin level (HR, 0.29; p = 0.004) and sub-nephrotic proteinuria (HR, 0.37; p = 0.030) had better patient survival. The predictive component of proteinuria with the progression of kidney disease and cardiovascular mortality has been widely described as worse renal survival with higher amounts [41]. While advanced age (HR, 1.04; p = 0.05) was established as a factor associated with less survival, older individuals are usually linked to other comorbidities that are also frequent in both diabetic patients and those with chronic kidney disease (42-44), which can contribute to the sum of mortality rates (45,46). To our knowledge, after reviewing the available literature, our study was the first to describe independent factors related to better patient survival in a cohort of diabetic patients with KB. 

On the other hand, patients with an eGFR >60 ml/min had better survival, while the presence of diabetic retinopathy was the only independent factor related to worse renal survival. Tan et al. (47) compared NDRD with ND, finding that renal prognosis was generally better with NDRD without specifying associated aspects. Also, Bermejo et al. (48) identified the presence of DN or NDRD plus DN as factors associated with higher mortality in their cohort. Also, in another study, Bermejo et al. (49) related advanced age, peripheral vascular disease, increased creatinine levels, and DN as risk factors for poor renal survival. 

We analyzed the type 2 DM subgroup again older age was correlated with less patient survival. Likewise, the absence of hematuria stands out as a protective factor in terms of survival (<5>10 erythrocytes/field). This presence of red blood cells of glomerular origin in DN is due to alterations in the glomerular basement membrane or to microaneurysms that can rupture [51-55]

Since no single variable has sufficient concordance with the final histological result, it is impossible to rely on a single parameter to make the final clinical decision to perform a KB. However, there is the possibility of building multivariate models that take advantage of the predictive properties of a set of variables simultaneously and support clinical decision-making [56]. In our study, the model with the variables age, retinopathy, and years with diabetes seems to work as a sensitive strategy to rule out DN alone and more strongly justifies a KB in diabetic patients.

Our study has limitations related to being retrospective. In addition, the subjectivity is related to histopathological studies. On the other hand, there is a selection bias since the biopsies are only from diabetic patients with a high suspicion of NDRD. The presence of proteinuria was one of the indications, so our results could be overestimated the true prevalence of NDRD.

Finally, NDRD is a frequent condition, as demonstrated by this cohort of Mestizo (Latino) patients; in addition, the identification of some characteristics such as the older age of the patient as well as the absence of retinopathy can be helpful when deciding whether to perform a KB because the high correlation to have other findings than DN on the histopathological results.