The Colombian National Registry of Chronic Kidney Disease (NRCKD) is a database of all people with diagnosed diabetes, hypertension or chronic kidney disease who have been assisted by the Colombian Health System. The NRCKD is managed by the High-Cost Diseases Colombian Fund (“Fondo Colombiano de Enfermedades de Alto Costo” - CAC in Spanish) and has been operating since 2008 through a resolution from the Colombian Ministry of Health (19). Each new registry cycle starts on July 1 of a year, and ends on June 30 of next year. The NRCKD is a passive registry with a national scope because almost the totality of the population is affiliated to the national healthcare system (19), and insurers are mandated by law to report all eligible patients to the registry (19).

As previously described (16), for each new case entering the NRCKD an initial registration is completed; after which data are updated every year. Each data point registered in the database corresponds to the last measurement of that variable within the observation period. The NRCKD undergoes a data validity auditing process with several steps. The first step involves the use of an algorithm to identify mistakes in the reporting process. Then, an experienced team compares the reported information with clinical records by a well-established data monitoring process in a representative sample of cases stratified by hypertension, diabetes, and CKD status (20). If any inconsistency is identified, real data from clinical records are captured.

We retrospectively analyzed data on all adults with diabetes reported to the NRCKD between July 1^st, 2015 and June 30^th, 2019. For each year of the study, we excluded people aged <18 at the start of the study year. The presence of a diagnosis of diabetes or hypertension was analyzed as reported to the NRCKD (Y/N as defined by the treating physician).

Metabolic control goals were defined according to recommendations by the International Diabetes Federation, the American Diabetes Association, and the Latin American Diabetes Association - ALAD (21–23). Treatment goals were HbA1c <7% (<53 mmol/mol), systolic blood pressure (SBP) <130 mmHg, and LDL cholesterol (LDLc) <100 mg/dL. We also analyzed non-HDL cholesterol (non-HDLc) below 130 mg/dL as an exposure. The joint triple goal was HbA1c <7% (<53 mmol/mol) plus SBP <130 mmHg plus LDLc <100 mg/dL.

Data from the NRCKD were used to classify participants in terms of age, sex, race or ethnic group, and insurance status. The database also contains data on weight and height, BMI was classified as recommended by the World Health Organization (24). Plasma creatinine values were used to calculate the estimated glomerular filtration rate (eGFR) using the Modified Diet for Renal Disease (MDRD) equation, which has been found to be more accurate than other equations among patients with diabetes (25). Based on eGFR, CKD stages were defined as follows: stage 1: GFR ≥ 90 mL/min; stage 2: GFR: 60-<90 mL/min; stage 3: GFR: 30-<60 mL/min; stage 4: GFR 15-<30 mL/min and stage 5: GFR: <15 mL/min.

The Colombian health system has three health insurance types: Third-party payer (“régimen contributivo”), run by private insurers (“Empresas Promotoras de Salud” – EPS); state-run insurance (“régimen subsidiado”), run by a different type of insurer (mostly “Administradoras de Régimen Subsidiado - ARS) and a special/exceptional health system for the security forces and employees of some public universities (régimen especial/régimen de excepción) (26). We studied insurance type using these three categories.

For the effects of this study, we collapsed the NRCKD race categories “Raizal”, “Palenquero” and “Black, Mulatto, Afro-Colombian or Afro-descendant” into a single category called “black” and analyzed self-reported race as black vs. all others. We made this decision because very few individuals (<1% in any given year) identified themselves as belonging to one of the other race categories (indigenous or Roma).

For descriptive analyses of baseline clinical and demographic characteristics, quantitative variables are presented as means and standard deviations, categorical variables as absolute and relative frequencies. For all analyses, the main outcome was death from any cause, inferred from the variable “date of death”, reported by insurers to the NRCKD as the date registered in the death certificate if the patient died during the observation year, or as missing if the patient was still alive. This information was double-checked against the data warehouse of the social protection integrated information system (SISPRO) of the Colombian Ministry of Health and Social Protection. In the case of a discrepancy, the Colombian database of death certificates was consulted.

The association between metabolic control goals as explanatory variables and death from any cause as outcome was evaluated using multivariable logistic regression models. There was a set of potentially confounding variables adjusted for in all models, including sex, age, race, insurance type and BMI. Additionally, we adjusted for the variables representing goals other than the one being evaluated. Thus, in models to evaluate the association between HbA1c goal and mortality, we adjusted for the basic set of confounders, plus hypertension status and non-HDLc. When SBP was the main exposure, we adjusted for the basic confounders plus HbA1c and non-HDLc. When one of the plasma lipids was the exposure, we adjusted for basic confounders plus hypertension status and HbA1c. In models evaluating the triple metabolic goal, we only adjusted for the basic confounders.

We performed two types of analyses, a first group considering as independent variable baseline metabolic control, and a second group modelling the effect of sustained metabolic control throughout the complete study period. The set of confounders being adjusted for was identical in both groups of analyses. Additionally, we performed stratified analyses to explore how metabolic control was related to mortality in subgroups defined by race (black vs other), and BMI category (normal, overweight or obesity). Interactions were tested by the significance of the regression coefficient associated to a multiplicative interaction term between goal achievement status and the stratification variable. All associations are expressed as Odds Ratios (OR) with 95% confidence intervals. To explore the robustness of findings and sort out reverse causation, we performed a sensitivity analysis of metabolic control and mortality excluding participants with CKD stages 4 or 5 at baseline. All analyses were performed at a 5% significance level, and all reported significance tests are two-tailed. Statistical analyses were performed in Stata version 17 (StataCorp LP, College Station, Texas, USA).

This research was based on anonymized secondary data sources and did not include any private information that could make any subject identifiable. To protect privacy, data were anonymized through the use of a database-specific individual ID. Because the study involved only secondary retrospective analyses of an anonymized database, it did not qualify as human subjects research as was exempted from IRB review. Confidentiality was guaranteed throughout the information processing (reporting, managing, and analysis).

There were 107 839 deaths over the entire follow-up (total cumulative mortality 7.97%). The variable whose baseline control was most strongly associated with lower mortality was a SBP<130 mmHg, an association that persisted and actually became stronger after multivariable adjustment (OR 0.72, 95% CI 0.71-0.74)). Glycemic control with HbA1c<7% at baseline was associated with only 3% lower odds of mortality in univariate analysis, but after adjustment for confounders the negative association grew stronger (18%, 95% CI 17-19%). Patients who achieved the joint triple goal at baseline had significantly lower mortality (OR 0.85, 95% CI 0.83-0.87).

The effect of sustaining basic diabetes treatment goals largely exceeded that of achieving them only at baseline. In models adjusted by sex, age, race, insurance type and baseline BMI, sustained SBP and HbA1c control were associated with 58% lower and 75% lower odds of death, respectively ( Figure 1 ). Sustained control of LDLc under 100 mg/dL also showed a strong negative association with mortality (OR 0.28 [0.27-0.29]). When we used the stricter 70 mg/dL cLDL goal, sustained LDLc control was associated with 75% lower odds of all-cause mortality (OR 0.25, 95% CI 0.23-0.27). When the three metabolic goals were achieved and sustained, the decrease in the odds of mortality was 81% (79-82%) ( Figure 1 ). Unfortunately, only 5.4% of study participants both achieved and sustained this degree of metabolic control. When a non-HDLc <130 mmHg substituted LDLc<70 mg/dL in the triple goal, the results were quite consistent, maintenance of this joint goal was associated with an adjusted OR for mortality of 0.19 (95% CI 0.17-0.21). Sustained control of albuminuria (urinary albumin excretion rate<30mg/g creatinine or <20 mg/L), was associated with a multivariable-adjusted OR for total mortality of 0.26 (95% CI: 0.24-0.27).

The positive effect of SBP control did not differ by between patients of black race and those of other races (p-value for the interaction 0.32). By contrast, the beneficial effect of sustained glycemic control was significantly smaller for patients of black race (OR 0.39, 95% CI 0.34-0.45 for black race; OR 0.24, 95% CI 0.25-0.25 for other races, p-value for interaction <0.001) ( Table 2 ). However, the largest difference according to race existed in the relationship of sustained lipid control and mortality, the impact being much larger for patients of non-black race. The reduction in the OR for sustained control of non-HDLc was 30% for black patients and 73% for patients of other races (p-value for interaction<0.001). In the case of sustained LDLc control, the figures were 59% lower odds for black patients, and 73% lower odds for patients of other races (p-value for interaction <0.001). In consequence, the reduction in the odds of death with sustained achievement of the triple goal was 69% for black patients, and 82% for patients of other races ( Table 2 ).

There was no significant difference in the association between sustained SBP control and mortality for patients with a BMI in the normal, overweight or obesity range (p-value for interaction 0.54, Table 3 ). Conversely, sustained glycemic control was most negatively associated with mortality among patients with obesity (p-value for interaction <0.001). While sustained control of LDLc had a similar impact in all BMI categories, non-HDLc control also had a greater impact in the obesity category (p-value for interaction 0.001). Remarkably, the effect of sustained joint metabolic control on mortality was similar across BMI categories, with no significant interaction ( Table 3 ).

Given that individuals with advanced CKD may experience reductions in blood pressure, glycemic levels and/or plasma lipids towards the end of life (leading to the so-called reverse causation problem), we performed a sensitivity analysis of sustained goal achievement and mortality excluding participants at CKD stages 4 or 5 at baseline. The results were generally consistent with those for the complete study sample. The OR for mortality according to sustained control of each variable were 0.50 (0.49-0.51) for SBP, 0.25 (0.24-0.26) for HbA1c, 0.27 (0.26-0.29) for LDLc, and 0.21 (0.19-0.23) for the triple goal ( Table 4 ).

Our study is based on a large, nationwide, centrally administered database of compulsorily reported and constantly audited data. In addition, the primary outcome of death from any cause is verified against official government sources. Limitations of the study include its relatively short follow-up for metabolic control to impact mortality. Also, we could not differentiate between patients with type 1 or type 2 diabetes, and metabolic control may have a numerically different impact in these two patient populations. In addition, the NRCKD database does not collect data on current medications, except for angiotensin-converting enzyme inhibitors and angiotensin receptor blockers. Arguably, however, most of the effects of antidiabetic, antihypertensive or lipid lowering drugs should be manifest in the values of SBP, HbA1c or blood lipids, which we did analyze. As in most registry-based studies of chronic diseases, the exact date of start of diabetes was not known, so we could not include diabetes duration as a covariate in our models.

In conclusion, our results showed that successful control of fundamental variables has a strong negative association with mortality in diabetes. Despite the existence and availability of the lifestyle and pharmacological means to attain these goals, they are still not being achieved by the overwhelming majority of patients. Widespread pursuit of tight metabolic control may yield large benefits in terms of mortality in diabetes.