This is an analytical study was carried out with two cohorts of patients diagnosed with CKD and four years of follow-up, evaluated in a previous study11 to determine the effectiveness of a renal protection program (RPP).

The population evaluated in the study belongs to two health service provider companies (EPS) in the department of Antioquia in Colombia, specifically in the city of Medellín and its metropolitan area and municipalities in eastern Antioquia. Data collection was carried out from both EPS between January 2007 and June 2009. The data recorded in the electronic clinical record from April 2004 (the beginning of the RPP) to June 2007 were retrospectively analyzed and prospective analysis continued until April 30, 2008, for a total of four years of follow-up.

Each EPS, from which the data for the study was taken, cared for their patients with a renal protection program (RPP) or with conventional treatment (TC). The PPR is an interdisciplinary and specialized program for the management of patients –with a preventive approach since the initial stages–; the TC consists of first level of management which is less preventive and more therapeutically oriented.

The patient inclusion criteria were: being affiliated to the EPS during the four years in which the study was conducted and meeting the CKD diagnostic criteria established by the National Kidney Foundation in the Kidney Disease Outcomes Quality Initiative.9,12 Patients under 15 years of age and those registered at both institutions at any time during the study were excluded. Follow-up began when they were diagnosed with CKD or included in the PPR. This study was approved by the ethics committee of the institutions involved and all the patients in the prospective phase signed the informed consent for participation in the study.

Diagnostic criteria of CKD: a GFR<60mL/min/1.73m2 (calculated using the Cockroft–Gault formula), renal abnormality on ultrasound, proteinuria >150mg/day, abnormal urinary sediment, and if diabetics a microalbuminuria >30mg/g or 30mg/24h.

Exposure variables: diagnosis of T2DM by clinical history and stages of chronic kidney disease defined according to the 2012 KDIGO guideline13 (Table 1).There ara five stages defined, one is the most moderate and five the most serious.

Dependent variables: all-cause mortality, CKD stage progression.

Descriptive variables: values of blood pressure at diagnosis, fasting glycemia, glycated hemoglobin, serum creatinine, 24-h proteinuria, HDL and LDL cholesterol, hemoglobin, electrolytes (potassium, calcium, phosphate), albumin, and parathyroid hormone.

Confounding variables: according to the reviewed publications,1,14 the following variables were assumed as potential confounding variables: age, sex, CKD stage at the time of diagnosis, arterial hypertension (AHT), dyslipidemia, EPS, nephrotoxic, nephroprotective, and Cardiovascular disease (defined as a history of acute myocardial infarction, coronary heart disease, or cerebrovascular accident).

The information was taken manually from the databases of each EPS and from the medical records, obtaining the information progressively during the study period.

To avoid measurement bias, the team dedicated to data collection was trained, the process for applying the tools was standardized, and a pilot test was carried out. In addition, each clinical history was entered into the database by two different professionals. During the study, an effort was made to reduce the risk of patient loss. With this purpose, the new telephone numbers of the patients were searched in the records and in other documents of the EPS. Likewise, telephone numbers of friends and relatives of the patients were sought by the team in charge of the programs. All patients were contacted by telephone to verify their health status after completing the study. Information could be obtained from 80% of them, without any difference between both EPS. Censored cases were those patients in which death was not reported as an outcome, the lost to follow-up, and patients who withdrew from the study. The censoring time was defined as the last date reporting that the patient was alive.

The number of participants was defined after counting the number of patients that met the CKD criteria according to KDOQI and had an electronic medical record in the provider companies. No formal sample size calculation was made in the original study, since the entire population that met the inclusion criteria was included.

The GFR was taken into account as a variable of progression in each stage. For the statistical analysis, stages 1 and 2 were assumed as the control group and stages 3a, 3b, 4 and 5 as the exposure group. Age was divided into two groups: under 65 years and over 65 years. This cut-off point was defined based on the fact that it was the mean of the total population studied and the threshold declared by the World Health Organization (WHO) to differentiate the old from the adult population.

For descriptive statistics it was used the mean, median or frequency distributions with their corresponding measure of dispersion, depending on the type of data. For the bivariate analysis, the X2test was used to assess the risk of mortality in stages 3 (a and b), 4, 5 of CKD and in diabetic patients, as well as the association between T2DM and CKD progression. The Kaplan–Meier survival curve analysis, was used to assess the probability of death in CKD stages 3, 4 and 5 in diabetic patients vs. not diabetics. Subsequently, stages 3a–3b and 4–5 subgroups were analyzed separately in the presence or absence of T2DM. Cox proportional hazards models were performed in patients with and without T2DM, adjusting for covariates. The ratio of incidence rate was estimated with a 95% confidence interval.

To detect the presence of the interaction, a stratified analysis of mortality in CKD stages was first performed using the Mantel–Haenszel technique, and confirmed with the Breslow–Day test. The interaction was evaluated as additive and as multiplicativel. The additive scale assesses whether the combined effect of both variables is different from the algebraic sum of the individual absolute risks, while the multiplicative scale assesses whether the effect is different from the product of the individual effects.15 The additive scale has proven to be important in public health analyzes and in the evaluation of biological interaction, but the general recommendation is to evaluate always both scales.16 Subsequently, two Cox proportional hazards models were performed to assess the risk of mortality in patients with CKD, adjusting for covariates and the CKD–DMT2 stage interaction. Both models were evaluated using the likelihood ratio test to determine the difference between the complete and nested models. The SPSS version 22 and STATA version 14 were used for the statistical analysis.

A total of 9887 medical records were reviewed, 56% from the PPR program and 44% from the TC program; as many as 5663 met the inclusion criteria and 3832 had at least two years of follow-up of, 8.9% withdrew and 5.5% died.

The mean age was 67 years, with no significant differences in the distribution by sex. A 35.3% of the patients (n=1998) had a diagnosis of T2DM. More than 50% of the patients with T2DM and CKD stages 3, 4, and 5 were 65 years of age or older. A 57.6% of the population had stage 3 CKD (a or b) and 1.8% had stage 5 CKD (Table 2 and Appendix B Supplementary Table 1).

The median weight in the T2DM group was 68kg and in the group without T2DM it was 60kg (p<0.001). The median 24-h proteinuria was 265mg/day and 190mg/day in the T2DM and non-DMT2 groups respectively (p<0.001). Regarding the lipid profile, the median value of triglycerides and HDL was 170mg/dL and 41mg/dL in the patients with T2DM and 154mg/dL and 43.3mg/dL in the control group respectively (p<0.005) (Appendix B, Supplementary Table 2).

In the group of patients with CKD stages 3, 4 and 5, there was a higher prevalence of comorbidities such as hypertension, dyslipidemia and cardiovascular disease, as well as more frequent use of nephroprotectors, all with significant differences (p<0.01). No significant differences were found regarding the prevalence of T2DM in both groups (Appendix B Supplementary Table 1).

The analysis found that both the stage of CKD (RR of CKD stages 3, 4, 5 vs. 1, 2=3.6; 95% CI=2.6–4.8) and diabetes (RR=1, 7; 95% CI=1.35– 2.14) behave as significant risk factors for mortality.

In the Kaplan–Meier curves, comparing T2DM patients vs. those without T2DM, all in CKD stages 3–4–5, it is observed that the probability of survival is higher in the group without T2DM showing statistically significant differences (Log-rank p=0.0076) (Fig. 1). A subgroup analysis of patients with and without diabetes, no significant differences were observed between CKD stages 3 or 4–5 (Log-Rank test p=0.70 and p=0.95, Figs. 2 and 3, respectively).

Fig. 1.

Survival analysis in CKD stages 3–4–5 in patients with and without T2DM.

Kaplan–Meier curve that evaluates the probability of survival in patients with CKD stages 3–4–5, with and without T2DM.

CKD: chronic kidney disease; T2DM: type 2 diabetes mellitus.

(0.15MB).

Fig. 2.

Survival analysis according to CKD stage in patients without T2DM.

Kaplan–Meier curve that evaluates the probability of survival in patients with CKD without T2DM. CKD: chronic kidney disease; T2DM: type 2 diabetes mellitus.

(0.15MB).

Fig. 3.

Survival analysis according to CKD stage in patients with T2DM.

Kaplan–Meier curve that evaluates probability of survival in CKD patients with T2DM.

CKD: chronic kidney disease; T2DM: type 2 diabetes mellitus.

(0.15MB).

The association of T2DM with mortality between the different CKD stage strata was evaluated using the Mantel–Haenszel technique, statistically significant differences were found in stages 3–4–5: OR 1.49 (95% CI: 1.17; 1. 91) and in stages 1–2: OR 5.47 (95% CI: 2.42; 12.39). The Breslow–Day statistical criterion shows a p value of 0.002, confirming that the stage behaves as an interaction variable in the association of T2DM with mortality.

Analysis of the interaction in the additive and multiplicative scales showed that it was significant in both scales: the joint effect observed is less than expected, which accounts for a negative or antagonistic interaction. Heterogeneity is also observed in the strata of both scales, which indicates that it is a quantitative interaction, that is, the association with the outcome of death is in the same direction, although with different magnitude (Table 3).

In the multivariable analysis, the T2DM–CKD stage interaction was also evaluated by means of an additional term, adjusting for covariates, finding that it was statistically significant (p=0.005). This was confirmed by the difference in the likelihood ratio test between the complete and nested models (p=0.0028) (Supplementary Tables 3 and 4).

In the multivariable analysis, it was found that the CKD stage and the covariates age, hypertension, dyslipidemia, and EPS are associated with mortality in patients without T2DM (Table 4). By contrast, in patients with T2DM, only age, dyslipidemia, and the use of nephroprotectors were found to be significantly associated with mortality (Table 5).