Cardiorenal prognosis by residual proteinuria level in diabetic chronic kidney disease: pooled analysis of four cohort studies

Abstract

INTRODUCTION

Over the past few decades, the prevalence of albuminuria in US adults with diabetes mellitus (DM) has declined significantly, likely due to better glycemic and hypertension control and wider use of drugs inhibiting the renin–angiotensin system (RAS) [1]. These preventive strategies have also modified the phenotype of diabetic renal disease, with an increasing incidence of non-albuminuric patients [2–6]. Nevertheless, the incidence of end-stage renal disease (ESRD) due to DM has not substantially changed [7, 8]. Furthermore, studies in the general population have demonstrated increased mortality and cardiovascular (CV) risk in the case of coexistence of DM and chronic kidney disease (DM-CKD) in compared with either DM or CKD alone [9–12].

These epidemiological features of diabetic kidney disease have been obtained in unselected patient populations and these findings deserve to be verified in the ideal setting of DM-CKD patients under nephrology care. Nephrologists, in fact, are the care providers for DM-CKD patients, especially in the presence of complications (anaemia, secondary hyperparathyroidism, electrolyte disturbances, etc.) or advanced CKD [13], and they can significantly improve cardiorenal prognosis when compared with other clinical settings, likely because of the more intensive therapy of risk factors, including more extensive use of RAS inhibitors [14–18].

The need for additional studies is also supported by the finding that observational studies in cohorts of diabetic subjects derived from unselected populations have shown that non-albuminuric patients with normal glomerular filtration rate (GFR) can still progress to overt renal impairment. These studies, however, did not evaluate the isolated role of diabetic status per se—that is, independent of proteinuria level [19–21]. Of note, the CKD Prognosis Consortium recently found that DM is associated with a higher risk of ESRD and death versus non-diabetic patients, even when controlling for albuminuria. However, the patients examined by the consortium were selected from various settings, with a large portion, >40%, not being followed by a nephrologist [22]. Finally, in all previous studies, prescription of RAS inhibitors—first-choice antihypertensive drugs in DM-CKD—was inconsistent [19–21] or not reported [22], thereby limiting interpretation of the results.

In the present study, we studied CKD patients treated with anti-RAS therapy and under nephrology care to compare mortality, CV risk and renal risk in diabetic versus non-diabetic CKD patients stratified by proteinuria level. The results provide novel information on the distinct effects of DM and proteinuria on cardiovascular and renal outcomes in the growing and high-risk population of diabetic CKD patients.

MATERIALS AND METHODS

This is a pooled analysis of four prospective cohorts enrolling consecutive patients with CKD Stages 1–5 referred to 40 Italian nephrology clinics [16, 23–25]. The four cohorts shared similar exclusion criteria: dialysis, kidney transplant, acute kidney injury, active malignancy and incomplete follow-up data; other study-specific exclusion criteria were normotension [blood pressure (BP) < 130/80 mmHg in the absence of antihypertensive therapy] and atrial fibrillation [24] and advanced heart failure (NYHA IV) [25]. The institutional review boards of the participating centres approved the four studies and patients gave written consent to use their clinical data.

For the purposes of the present study, diabetic and non-diabetic patients receiving angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers or both were stratified separately in four categories of proteinuria (<0.15, 0.15–0.49, 0.5–1.0 and >1.0 g/day). According to the Kidney Disease: Improving Global Outcomes classification of CKD [26], the first two categories correspond to normal-to-mild proteinuria and moderate proteinuria, respectively, whereas the two categories of 0.5–1.0 and >1.0 g/day match the definition of severe proteinuria.

In all four cohorts, participating nephrologists collected demographic information and history of CV disease, performed a physical examination with assessment of height, body weight, BP and medication profile and reviewed laboratory results. Data collected in anonymous electronic case reports were subsequently sent to the coordinating centre for analyses.

Laboratory protocols were standardized using in-house analytical measurements; 24-h urine collection was done to quantify proteinuria and evaluate adherence to the prescription of protein and salt intake. Collection was considered inaccurate, and repeated, if creatinine excretion was outside the 60–140% range of the value calculated according to Dwyer and Kenler [27]. Glomerular filtration rate (eGFR) was estimated using the Chronic Kidney Disease Epidemiology Collaboration equation; since creatinine was not standardized to isotope-dilution mass spectrometry values, we reduced creatinine values by 5% [28].

Outcomes

The three outcome measures were all-cause death, ESRD (start of chronic dialysis or kidney transplantation) and a composite CV endpoint of non-fatal CV events requiring hospitalization (myocardial infarction, congestive heart failure, stroke, revascularization, peripheral vascular disease and non-traumatic amputation) and CV death, whichever occurred first. To establish the underlying cause of death and to determine CV deaths, we used death certificates and autopsy reports, whereas hospital records were used to identify the diagnosis of non-fatal CV events based on the International Classification of Diseases, Ninth Revision, Clinical Modification [16, 23, 24]. Patients were followed-up until 31 December 2015, death or ESRD and censored on the date of the last nephrology clinic visit.

Statistics

Continuous variables were reported as either mean ± standard deviation (SD) or median and interquartile range (IQR) based on their distribution. Comparison between the two groups was performed by unpaired Student’s t-test or Mann–Whitney test, whereas changes across proteinuria categories in each group were assessed by either analysis of variance (ANOVA) or Kruskal–Wallis test. Categorical variables (expressed as a percentage) were analysed using the chi-square test. Among 2174 patients, 83.4% had complete data. Missing data included calcium (n = 207), phosphate (n = 79), albumin (n = 184) and triglycerides (n = 169). Since patients with one or more missing variables did not differ from those with complete data in terms of demographics, clinical characteristics and outcomes, we imputed missing data to include all patients in survival models by implementing multiple imputation [29].

Because ESRD and death before ESRD are competitive events—that is, occurrence of death prevents ESRD—we calculated the cumulative incidence of ESRD or death before ESRD by using the competing risk approach; for the CV endpoint, ESRD and non-CV death were considered as competing events. Cumulative incidence curves were compared by using the Gray test [30]. The risks of the three endpoints were estimated by multivariable Cox proportional hazards models stratified by cohort and adjusted for baseline covariates identified a priori (age, gender, active smoking, BMI, history of CV disease, systolic BP, cholesterol, triglycerides, phosphate, albumin, haemoglobin and GFR). Inclusion of these covariates and proteinuria categories by diabetic status (22 degrees of freedom) did not produce over-fitted survival models due to the large number of events recorded. We used Cox models because the cause-specific relative hazards are more appropriate for studying the cause of disease in the case of a competing event [31]. To evaluate whether the increase in hazard ratios (HRs) for each of the outcomes with increasing proteinuria differs according to diabetic status, we performed a priori contrast between diabetic and non-diabetic patients in each proteinuria category by means of the Wald test. We conducted three sensitivity analyses. First, because a history of myocardial infarction is strongly associated with mortality and CV outcomes in DM with or without CKD [12], we conducted a sensitivity analysis after excluding patients with previous coronary heart disease. Second, we separately examined DM-CKD and non-diabetic CKD patients by using as a reference the lowest proteinuria category. Third, we reran survival models by including only patients with complete data sets (n = 1814) in order to test the impact of imputed data on the major analyses. A two-tailed P-value <0.05 was considered significant. Data were analysed using STATA version 14 (StataCorp, College Station, TX, USA).

RESULTS

Clinical characteristics

From the initial pooled cohorts, we removed 1051 patients (28.4% with diabetes) based on the exclusion criteria (Figure 1). In the whole CKD population, the prevalence of diabetes was 31.9% (95.3% Type 2 DM). The distribution of CKD stages was as follows: Stages 1–2: 8.9%, Stage 3a: 16.2%, Stage 3b: 32.8%, Stage 4: 32.0% and Stage 5: 10.1%, similar in diabetic and non-diabetic groups (P = 0.103). Patients with DM-CKD were at higher risk compared with non-diabetic CKD patients. Indeed, they were older (68.8 ± 10.6 versus 65.3 ± 15.3 years; P < 0.001), had higher BMI (29.6 ± 5.4 versus 27.5 ± 4.8 kg/m²; P < 0.001) and systolic BP (143 ± 19 versus 136 ± 18 mmHg; P < 0.001) and had more frequent history of CV disease (48.5% versus 28.8%; P < 0.001). These differences persisted when patients were stratified by proteinuria (Table 1). The length of follow-up in nephrology clinics was 17 months (IQR 12–45), similar in diabetic and non-diabetic patients. Overall, in the DM-CKD group, proteinuria was higher than in the non-diabetic group [0.48 (IQR 0.14–1.40) versus 0.32 (IQR 0.11–0.90); P < 0.001] and eGFR was slightly lower (34.5 ± 17.9 versus 36.6 ± 19.0 mL/min/1.73 m²; P = 0.014). eGFR was progressively lower across proteinuria categories in both groups, with no difference between diabetic and non-diabetic CKD (Table 1).

FIGURE 1

Flow chart of the study. Acronyms of studies are: TArget BLood pressure LEvel in Chronic Kidney Disease (TABLE-CKD) [17], NEPHROlogy Second University of Naples (NEPHRO-SUN) [20], Ambulatory Blood Pressure in Chronic Kidney Disease (ABP-CKD) [18] and REporting COmorbidities in Renal Disease in ITaly (RECORD-IT) [19].

Open in new tabDownload slide

Albumin, haemoglobin and calcium levels decreased with higher levels of proteinuria in both groups whereas serum phosphate and urinary sodium excretion increased (Table 1). Changes in cholesterol and triglycerides levels were more pronounced across proteinuria categories in DM-CKD. In this group, total cholesterol was lower whereas triglycerides levels were higher than in non-diabetic CKD (Table 1). In DM-CKD, haemoglobin A1c (HbA1c) was on average 7.1 ± 1.3% with optimal control (HbA1c <7%) and similar in the four proteinuria categories (54.4, 60.4, 54.7 and 48.0%, respectively; P = 0.312). The duration of diabetes was 15 years (IQR 8–21) with no difference between proteinuria categories (P = 0.378).

DM-CKD patients received more antihypertensive agents than non-diabetic CKD, with greater use of furosemide, calcium-channel blockers and beta-blockers (Table 2). Dual blockade was more frequently prescribed in DM-CKD (19.0% versus 13.9%; P = 0.002) as well as erythropoiesis-stimulating agents, statins and antiplatelet agents (Table 2).

Survival analyses

During a median follow-up of 4.07 years, we registered 466 ESRD, 334 all-cause death and 401 CV events (225 fatal) (Table 3). Corresponding annual rates were 5.26 per 100 patient-years for ESRD (6.04 in DM and 4.91 in non-DM), 3.77 per 100 patient-years for mortality (4.79 in DM and 3.32 in non-DM) and 4.86 per 100 patient-years for CV events (6.55 in DM and 4.13 in non-DM). DM-CKD patients had a greater absolute risk of death and CV events compared with non-diabetic CKD patients, but were not exposed to a greater risk of ESRD (Figure 2A). Specifically, crude CV and mortality risks were similar in non-proteinuric patients but higher in diabetic patients with proteinuria >0.15 g/day (Figure 2B and C). Renal risk was significantly higher in DM-CKD patients for the proteinuria stratum 0.5–1 g/day (Figure 2D). These results were confirmed by the analysis of cumulative incidence of ESRD and death, taking into account the competing risk of ESRD and death before ESRD (Supplementary data, Figure 1) and CV outcome (Supplementary data, Figure 2).

FIGURE 2

Absolute risks for all-cause death, fatal and non-fatal CV events and ESRD in diabetic (grey bars) and non-diabetic (white bars) CKD patients (A) overall and (B–D) by proteinuria level.

Open in new tabDownload slide

Adjusted survival analyses, considering non-proteinuric non-DM as the reference group, showed that mortality risk increased by 92% in diabetic patients with moderate proteinuria (0.15–0.49 g/day) and persisted at higher proteinuria levels, whereas in non-diabetic CKD the risk increased only for proteinuria >0.5 g/day (Table 3). A similar pattern of risk modification by proteinuria category and diabetic status was observed when examining CV risk. Adjusted survival analysis confirmed that renal risk increased when proteinuria was >0.5 g/day, with no difference between diabetic and non-diabetic patients (Table 3).

To evaluate the excess risk for each endpoint, we calculated the contrast between diabetic and non-diabetic patients in each proteinuria category using the Wald test. In DM-CKD, excess risk for mortality and CV events was evident when proteinuria was in the range 0.15–0.49 g/day [+65% (P = 0.027) and +90% (P = 0.002), respectively]. Indeed, in DM-CKD as compared with non-diabetic patients, a higher HR was found for mortality {1.92 versus 1.17, corresponding to an HR of 1.65 [95% confidence interval (CI) 1.06–2.57]} as well as for CV events [1.80 versus 0.95, corresponding to an HR of 1.90 (95% CI 1.26–2.85)] (Table 3). In other proteinuria categories, mortality and CV risk did not differ between diabetic and non-diabetic CKD patients. ESRD risk did not differ between diabetic and non-diabetic patients at any level of proteinuria. Other factors independently associated with outcomes are depicted in Supplementary data, Figure 3.

Finally, we conducted three sensitivity analyses. First, after excluding patients with a history of myocardial infarction (n = 329), our results were similar to those using the data from the full data set showing higher mortality and CV risk in DM-CKD patients with proteinuria 0.15–0.49 g/day (Supplementary data, Table 1). Second, we separately analysed DM-CKD and non-diabetic CKD patients by using as a reference the lowest proteinuria category (Supplementary data, Table 2). This analysis confirmed that CV and mortality risks became significant with lower proteinuria levels in DM-CKD, whereas ESRD risk did not differ between diabetic and non-diabetic patients, increasing significantly only when proteinuria was >0.5 g/day. Third, we reran all analyses by including only patients with complete data and found the same risks for each outcome (data not shown).

DISCUSSION

This is the first study to dissect the prognostic role of diabetes and proteinuria in diabetic CKD patients under nephrology care, all receiving anti-RAS agents. To this end, we compared cardiorenal outcomes in diabetic versus non-diabetic CKD patients stratified by residual proteinuria. Our results support several conclusions. First, non-proteinuric diabetic CKD patients (<0.15 g/day) do not have higher cardiorenal risk when compared with their counterpart non-diabetic CKD patients. This novel observation constitutes a major difference with findings in the general population, where diabetes per se heralds poor cardiorenal survival [7, 9–12]. Second, the presence and extent of proteinuria modulate the risk related to diabetes; specifically, moderate proteinuria (0.15–0.49 g/day) heralds a higher CV morbidity and mortality only in DM-CKD patients, whereas severe proteinuria (≥0.5 g/day) levels off the renal risk of diabetics to that of non-diabetic CKD patients.

These data are useful for research and clinical practice. In particular, the prognosis of non-proteinuric diabetic CKD patients has remained unknown until now, despite their large and increasing prevalence [2–6]. This problem is amplified by the systematic exclusion of these patients from clinical trials [32]. The reason for the lack of increased cardiorenal risk in DM-CKD versus non-diabetic CKD patients without proteinuria is not readily apparent. One explanation could be a relatively higher prevalence of hypertensive nephropathy as the cause of CKD in this group. Indeed, it has been proposed that normoalbuminuric renal impairment is secondary to macroangiopathy rather than the effect of typical DM-related microangiopathy [32, 33]. However, we cannot exclude that, in our non-proteinuric DM-CKD patients, a greater use of statins, anti-platelet agents, beta-blockers and furosemide may have provided greater protection against CV and mortality risk. This more intensive treatment may have actually limited the potentially greater risk due to the older age and the higher prevalence of obesity and CV disease. Although observational, our results suggest that these patients have a relatively good prognosis when polytherapy including anti-RAS agents is implemented. Finally, we cannot exclude that a more precise assessment of BP load through ambulatory monitoring (not available in the present study) would have contributed to explaining our findings. This is particularly true for patients with high ambulatory BP and normal office BP (masked hypertension), which is more prevalent in diabetic patients [34] and associated with a higher albuminuria level [35] and an adverse outcome in CKD patients [36].

The finding that proteinuria was a major modifier of prognosis was also clinically relevant. This is in agreement with current knowledge; however, the novelty of our results is that we compared the risk in diabetic versus non-diabetic patients with similar proteinuria. Previous cohort studies and meta-analyses in DM-CKD have highlighted that the risk of adverse outcomes increases with higher proteinuria levels [9, 22, 37, 38]. Nevertheless, these studies were unable to clarify whether the risk was different from that of non-diabetic CKD patients with a similar severity of renal damage. In our study, we utilized non-diabetic patients with the same degree of proteinuria and eGFR as a reference group to optimally compare cardiorenal risk and risk determinants. The CKD Prognosis Consortium found a higher ESRD risk in diabetic versus non-diabetic CKD patients when albuminuria was moderate to severe [22]. This association was strongly modified by the presence of hypertension because an increased ESRD risk in DM-CKD was evident only in untreated patients with BP <140/90 mmHg, whereas renal risk was similar in hypertensive CKD patients with and without diabetes [22]. Our results are in line with the consortium study when considering that only 0.1% of diabetic patients in our study are normotensive. A negative role of proteinuria on renal outcome independent of diabetic status has been reported by Lorenzo et al. [39] in a small study including 153 diabetic and 180 non-diabetic CKD patients. Their study, besides the small sample size, could not dissect the renal risk according to albuminuria level because the albumin:creatinine ratio was used as an adjustment variable rather than as a stratification variable; furthermore, the inconsistent use of anti-RAS therapy limited data interpretation.

In the last two decades, despite an increase in age-standardized death rate attributable to diabetes, particularly in those with CKD [38], the rates of CV diseases have been decreasing in diabetic CKD patients but continue to be higher in non-diabetic CKD individuals [7, 9]. Atherosclerotic CV disease is the principal cause of death and disability among diabetic patients, in whom death typically occurs ∼15 years earlier than the general population [7, 40, 41]. In contrast, CV complications in diabetic and non-diabetic CKD patients may be influenced by the remarkable risk conveyed by CKD per se [12]. Navaneethan et al. [42], using an electronic medical record 48 South River Road based CKD registry including patients with GFR < 60 mL/min, found no difference in CV mortality risk between DM-CKD and non-diabetic CKD. However, survival models in their study were not adjusted for proteinuria, because this parameter was available in <50% of patients. Our findings highlight the importance of proteinuria assessment in DM-CKD patients as a tool to further refine CV risk above and beyond renal risk. We found that mortality and CV risk increase in parallel with proteinuria levels and independent of diabetes. Furthermore, the presence of moderate proteinuria (0.15–0.49 g/day) in DM-CKD patients is associated with a higher CV risk compared with non-diabetic CKD patients with similar proteinuria levels.

The increase in CV risk occurred despite better control of dyslipidaemia and more frequent use of statins. In a large population-based cohort, Tonelli et al. [12] reported that the coexistence of diabetes and proteinuric CKD heralds higher rates of myocardial infarction and death with respect to patients with either diabetes or CKD alone. To exclude a role for coronary heart disease history in the excess risk in our DM-CKD patients, we conducted a sensitivity analysis. Exclusion of these patients did not alter the original findings (Supplementary data, Table 1). The higher CV risk, independent of CV history, may be driven by hyperglycemic status that can promote atherogenesis and accelerate the progression of atherosclerosis through several mechanisms [41]. In addition, CKD-specific risk factors, such as low vitamin D levels and vascular calcification [41], may also play a role.

Our study is limited by the assessment of predictors only at baseline; nevertheless, the prolonged follow-up under nephrology care prior to the baseline visit reasonably excludes substantial changes of risk factors in the subsequent period. Furthermore, we cannot distinguish patients with spontaneous low proteinuria from those with a treatment-induced decline in proteinuria. On the other hand, a strength of the study is the large number of cardiorenal events that allowed adequate survival analyses. Indeed, over the 4-year period of follow-up, 42% of DM-CKD patients reached the hard endpoint of ESRD (164/693) or death (130/693). Additional strengths include the highly reliable evaluation of proteinuria by means of 24-h urine collection and detailed information on medication use.

In conclusion, this study provides novel information on CV and renal prognosis in DM-CKD patients referred to nephrology clinics. In the absence of proteinuria, DM-CKD patients are not exposed to increased cardiorenal risk when compared with non-diabetic CKD patients. In contrast, in proteinuric CKD patients, the risk of ESRD is primarily driven by the proteinuria level independent of diabetic status. The independent effect of diabetic status per se only emerges in the presence of moderate proteinuria (0.15–0.49 g/day), where diabetes heralds a higher risk of CV morbidity and mortality.

ACKNOWLEDGEMENTS

This work was endorsed by the Italian Society of Nephrology (Gruppo di Studio sul Trattamento Conservativo della Malattia Renale Cronica) without any financial support.

AUTHORS’ CONTRIBUTIONS

R.M., M.P., P.C., G.C. and L.D.N contributed to the study concept and design. M.P., P.C., S.B. and C.G. undertook statistical analysis. R.M., F.B.G., M.P., P.C., G.C. and L.D.N contributed to the interpretation of data. R.M., F.B.G., F.C.S., V.B., G.C. and L.D.N contributed to manuscript writing. All authors participated in revising the manuscript and providing intellectual content of critical importance. The corresponding author (R.M.) has full access to all the data in the study and had final responsibility for the decision to submit for publication.

CONFLICT OF INTEREST STATEMENT

F.C.S. has received lecture fees from Eli Lilly, Novo Nordisk, Sanofi-Aventis, Merck, Roche Diagnostics and Janssen. L.D.N. has received consulting fees from AstraZeneca and Janssen. The remaining authors declare no conflicts of interests.

(See related article by Chan and Tang. Proteinuria reaffirmed as a risk modifier in diabetic chronic kidney disease. Nephrol Dial Transplant 2018; 33: 1873–1874)

REFERENCES