The Modern Spectrum of Renal Biopsy Findings in Patients with Diabetes

Discussion

In this cohort of 620 patients with DM, the largest study to date of renal biopsy findings in diabetic patients, NDRD was identified in >60% of biopsies: 220 patients with NDRD alone and 164 patients with NDRD and coexistent DN. We found that duration of diabetes was the strongest predictor of whether NDRD or DN was identified on biopsy. The median duration of DM in patients with NDRD alone was 5 years, which was significantly shorter than in patients with DN alone (13 years) and DN plus NDRD (10 years), and DM duration ≥12 years emerged as the best predictor of DN alone. In addition, although the median proteinuria for the entire cohort was in the nephrotic range, heavier proteinuria was associated with a lower likelihood of finding NDRD alone. Taken together, these results suggest that the diabetic patient who is most likely to have NDRD alone has a short duration of DM and subnephrotic proteinuria.

Our results are an important addition to previously reported data from smaller cohorts on biopsy findings in diabetic patients and the significance of disease duration and proteinuria. A number of smaller studies, with cohort sizes ranging from 52 to 370, have reported no differences in the duration of DM between patients with DN and NDRD (7–9). Conversely, Tone et al. (3) found that short duration of diabetes (<5 years) carried high sensitivity (75%) and specificity (70%) for predicting NDRD, and Chang et al. (2) reported a mean DM duration of 5.9 years in patients with NDRD versus 10.6 years in patients with DN alone (P<0.001 for comparison). Similarly, previous reports have provided conflicting findings on the relative weight of proteinuria in predicting DN versus NDRD. Whereas some investigators have reported nephrotic-range proteinuria in patients with DN alone, DN plus NDRD, and NDRD alone (1,2), the majority of reports have suggested that a higher degree of proteinuria is found in DN than in NDRD (3,4,8,10,11).

In clinical settings, a number of serologic tests are often performed on diabetic patients with proteinuria, particularly those with either an acute onset of proteinuria or an abrupt increase in proteinuria. The value of such serologic testing in predicting NDRD has not been reported by others (1–4,8,10,11). In our cohort, although the finding of any positive serologic test performed at the time of biopsy was not predictive of NDRD, the specific findings of low complement levels (low C3 and/or low C4) and M-spike in either serum or urine, on univariate analyses, were associated with NDRD on biopsy. These serologies, however, were only predictive of biopsy findings in about 25% of cases, and in multivariate models, low complements and M-spikes were associated with NDRD only when cases of ATN, FSGS, and hypertensive nephrosclerosis were excluded. We hypothesize that some of these low serum complements may have been spuriously low due to the methods of serum collection and delayed delivery of the sample for laboratory analysis. We interpret the M-spikes in patients without relevant renal pathologies as representing monoclonal gammopathy of uncertain significance. Our patients are in an age group (median age 62 years) when monoclonal gammopathy of uncertain significance becomes increasingly prevalent in the general population.

Perhaps the most important aspect of this study is not only that NDRD was identified in the majority of diabetic patients who underwent kidney biopsy, but also the spectrum of NDRD seen on biopsies, either alone or superimposed on DN. In the cohort of patients with NDRD alone, FSGS (22%), hypertensive nephrosclerosis (18%), ATN (17%), IgAN (11%), MGN (8%), and pauci-immune GN (7%) comprised 80% of diagnoses, compared with ATN (43%), hypertensive nephrosclerosis (19%), FSGS (13%), and IgAN (7%) for NDRD with coexistent DN (Table 2). The overwhelming majority of these diagnoses would yield some change in treatment, ranging from the use of immunosuppression to titration of renin-angiotensin-aldosterone system blockade. The most common NDRDs previously reported in patients with DM are IgAN (3,11–13), FSGS (1,10), and MGN (10,14). In our series, FSGS was the most common finding in NDRD alone, identified in 48 patients (22%) and, in all but 6 patients, read as secondary or adaptive FSGS related to such comorbidities as obesity and systemic hypertension. Among all cases of NDRD in this cohort (i.e., with and without concomitant DN), however, ATN was the most common biopsy finding, seen in 28.4% of all NDRD patients. This is a unique finding of our study, because ATN has not been reported previously as a common NDRD, either alone or superimposed on DN. The higher incidence of ATN as a superimposed disease on DN may reflect reduced renal reserve and the greater potential for renal hypoperfusion and hemodynamic instability under renin-angiotensin-aldosterone system blockade, typically used as standard therapy to allay progression of DN. The high prevalence of ATN carries even more weight in light of evidence from a number of epidemiologic cohorts that AKI episodes are associated with a cumulative risk for progression to ESRD, particularly in patients with diabetes (15–17).

The strength of this retrospective study lies in its size and generalizability. Because our renal pathology laboratory is a referral laboratory for nephrologists over a broad geographic region, these findings are likely representative of current national trends and illustrate the spectrum of renal pathology in the modern era of the type 2 diabetes epidemic. Notably, approximately one of every four native renal biopsies is performed in a diabetic patient. The high frequency of NDRD in such biopsies suggests that nephrologists are appropriately selecting for biopsy those diabetic patients with clinical features considered atypical of DN alone or suspicious for NDRD. In addition, the high rates of FSGS and hypertensive nephrosclerosis highlight the importance of comorbidities such as obesity and hypertension.

Our study has a number of limitations. First, selection bias is inherent in any biopsy-based clinicopathologic study. Our results can only be interpreted as applicable to those patients whom a treating nephrologist would consider biopsy candidates. This bias should be toward the null, however, because most nephrologists would not biopsy patients for whom the pre-test probability of finding NDRD is low. Diabetic retinopathy has classically been shown to predict DN, but the clinical history provided at the time of biopsy only reported retinopathy results on 14% of this cohort. If the common clinical practice to not biopsy diabetic patients with retinopathy holds for our cohort, then including retinopathy data in our analyses may have introduced significant selection bias. Given the importance of proteinuria in this cohort and its association with biopsy findings, ideal analyses would not only include proteinuria at time of biopsy but trends in proteinuria before biopsy. Such data were not available and precluded us from analyzing whether changes in proteinuria were equally important as degree of proteinuria. Similarly, hemoglobin A1c values were missing for most patients and thus could not be included in our analyses.

In summary, our data indicate that approximately one-quarter of all renal biopsies in the modern era are performed in patients with diabetes. As the diabetes epidemic prevails, there is an increasing need for nephrologists to remain vigilant and weigh the risks and benefits of renal biopsy in the diabetic patient. In this cohort, >60% of biopsies in diabetic patients revealed NDRD, alone or alongside concomitant DN. Whereas AKI, low complements, and positive M-spike suggest an increased likelihood of finding NDRD on biopsy, long duration of DM emerged as the strongest predictor of finding DN alone. Therefore, the threshold to pursue kidney biopsy should increase in parallel with DM duration.