Background
Serum bicarbonate is typically included in routine biochemical tests of renal function and is therefore likely to be measured regularly as part of the long term management of patients with type 2 diabetes. Clinically significant abnormalities of bicarbonate homoeostasis are rare in this setting. Indeed, guidelines recommend its measurement as an index of acid–base balance only when renal disease has developed [
1]. However, despite these considerations, some population-based studies and data from cohorts of patients with chronic kidney disease (CKD) have suggested a possible role for serum bicarbonate as an independent risk marker of mortality [
2‐
9] and heart failure (HF) [
2,
10] in addition to progressive renal disease [
2,
3,
11,
12]. Just one of these studies included only patients with diabetes and all had established CKD [
10]. This and other less selective studies of patients with CKD [
2,
4,
5] have suggested that an association between serum bicarbonate and mortality may be lost in diabetes.
There has been no investigation of the relationship of serum bicarbonate and key endpoints including death in unselected community-based patients with diabetes with renal function ranging from normal to end-stage renal disease. Such a study could provide evidence of the utility of a routinely-measured serum bicarbonate as a prognostic indicator that could prompt modification of therapy and thus potentially reduce morbidity and death. We have, therefore, assessed the associations between serum bicarbonate and all-cause mortality, coronary heart disease (CHD) and HF in well characterized representative community-dwelling patients with type 2 diabetes. We hypothesized that serum bicarbonate would have significant independent associations with mortality, CHD and HF in this cohort.
Discussion
In our community-based representative patients with type 2 diabetes, there was a significant independent inverse association between baseline serum bicarbonate and the risk of incident CHD in patients without a history CHD at baseline. Patients in the lowest quintiles of serum bicarbonate were at increased risk of death and incident HF, but adjustment for clinically relevant covariates in the most parsimonious models rendered both these associations non-significant. These results indicate that serum bicarbonate is a significant predictor of incident CHD in type 2 diabetes, but that it acts as a surrogate for other pathological processes underlying HF and all-cause mortality.
We observed an increased risk of all-cause mortality in those with serum bicarbonate ≤23 mmol/L compared to those in the highest quintile in unadjusted analysis during a mean of 12.9 years of follow-up, but this did not remain significant after adjustment for a large range of other explanatory variables. The only other study of the association between serum bicarbonate and mortality exclusively in diabetes involved combined data from patients with CKD who participated in the Reduction of End points in Non-insulin-dependent diabetes with the Angiotensin II Antagonist Losartan (RENAAL) trial or the Irbesartan Diabetic Nephropathy Trial (IDNT) [
10]. There was increased mortality during a median 2.8 years of follow-up in patients with serum bicarbonate in the lowest two quartiles that was no longer significant after full adjustment for other variables including eGFR. Similarly, significant bivariate associations between baseline bicarbonate and mortality in two other studies of CKD patients from the general population were also no longer present after adjustment [
2,
22]. Our observations are consistent with these studies in showing no independent association between serum bicarbonate and death.
Other studies in samples of generally healthy subjects and in CKD patients have, by contrast, demonstrated significant and independent associations between serum bicarbonate and all-cause mortality [
3‐
8]. In the Health ABC study of people aged 70–79 years [
7], a baseline arterialized venous blood serum bicarbonate <23.0 mmol/L was independently associated with increased mortality compared to 23.0–27.9 mmol/L. National Health and Nutrition Examination Survey III data from a cohort of 15,836 participants selected to represent the general US population also showed that a low serum bicarbonate (<22 mmol/L) was associated with mortality independently of age, gender, race, eGFR, albuminuria, cardiovascular disease, lung disease, diabetes, hypertension, smoking status, C-reactive protein and estimated protein intake normalized to body weight [
8]. In a Korean study of the health screening records from 31,590 participants, there was an independent increased mortality risk for those with a serum bicarbonate in the lowest compared to the highest quartile [
6]. In two clinic-based studies of non-dialysis CKD patients, there was also an independent increased mortality risk for those with low serum bicarbonate concentrations [
4,
5] in addition to increased mortality in the highest serum bicarbonate category (>32 mmol/L) in one [
5]. In a second analysis of the Chronic Renal Insufficiency Cohort (CRIC) study involving CKD patients not on hemodialysis, there was an independent association between a high updated mean serum bicarbonate and mortality, but no independent association for a low serum bicarbonate [
3].
By contrast, three studies of community-based participants and two carried out in clinic-based samples of non-dialysis dependent CKD patients found independent associations between serum bicarbonate and mortality. The apparent discrepancy between studies based on the diabetes status of the participants may relate to the availability of covariates. Factors such as exercise and comorbidity were significantly associated with mortality in the most parsimonious regression model in our study but were typically not available for use in multivariable mortality models in general population studies. Alternatively, it may be that the pathophysiological changes associated with diabetes swamp a potential contribution from serum bicarbonate. This would be consistent with the observation in several studies that, while serum bicarbonate was associated with death in the sample as a whole, there was no such association in the sub-groups of patients with diabetes [
4,
5]. This finding, the RENAAL/IDNT study data [
10] and our own observations all suggest that a low serum bicarbonate is not an independent risk factor for all-cause mortality in type 2 diabetes.
In contrast to mortality, we found a significant independent inverse association between serum bicarbonate and first incident CHD event, with a linear 5 % reduction in risk per 1 mmol/L increase in serum bicarbonate. For an FDS patient with serum bicarbonate of 22 mmol/L, the lower limit of the laboratory reference interval, the relative risk of CHD was approximately 25 % greater that of a patient with a serum bicarbonate at the mean concentration in the total sample of 27 mmol/L. Three reports from two other studies have also assessed this association, specifically the combined RENAAL/IDNT study of patients with diabetic nephropathy [
10] and the CRIC study [
2,
3], but there was no independent association between serum bicarbonate and atherosclerotic events in either study after median durations of follow-up of 2.8, 3.9 and 6 years, respectively. The cardiovascular events in these studies included not only myocardial infarction but also stroke and, in the CRIC cohort, peripheral vascular disease. This heterogeneous endpoint may have masked a specific association between serum bicarbonate and CHD. Furthermore, prior cardiovascular disease was not included as a covariate in the RENAAL/IDNT study [
10] and in the earlier CRIC analysis [
2]. The longer follow-up in the present study (with a resultant higher proportion of patients experiencing CHD than in the RENAAL/IDNT and CRIC studies), the more specific end-point, and our exclusion of patients with prevalent CHD may explain the greater sensitivity of the present study to detect an association between serum bicarbonate and incident CHD risk.
We observed a significant increase in risk of HF in our patients with a serum bicarbonate in the lowest quintile, and a significant inverse association between incident HF and serum bicarbonate when analysed as a continuous variable. However these associations were lost in the fully adjusted model. Of note, however, the lowest risk appeared to be in the 4th rather than the 5th quintile, suggesting the possibility of a U-shaped relationship, although this difference was not significant. In contrast to our observations, two studies have found significant positive associations between measures of serum bicarbonate and incident HF [
3,
9]. One study found a significant positive association between updated mean bicarbonate and incident HF in patients with CKD, and those with mean bicarbonate >26 mmol/L had an independent 66 % increased risk compared to those with mean bicarbonate of 22–26 mmol/L [
3]. Similarly, a second study found that those in the highest quartile of bicarbonate had an increased risk of HF hospitalization than those in the 3rd quartile after adjustment, but not in the unadjusted model [
10]. In neither study was a low serum bicarbonate associated with incident HF. It may be that there is a difference between the patients in these studies that modulates the association between bicarbonate and HF. For example, the patients in the previous studies [
3,
10] had relatively poor renal function compared to those in the present study and the adjustments in previous studies included many variables associated with serum bicarbonate, but they did not include recent exercise and digoxin use which may reflect increased HF risk and which we found to be significant covariates in the most parsimonious model. As noted above, the analysis from the RENAAL/IDNT study did not include adjustments for prevalent cardiovascular and cerebrovascular disease [
10] while the use of updated mean serum bicarbonate may not be directly comparable to baseline serum bicarbonate in an ageing patient [
3]. We conclude that serum bicarbonate is not a significant risk factor for HF in community-dwelling patients with type 2 diabetes.
We examined the independent cross-sectional associates of serum bicarbonate to shed light on factors which may be influential and which may need to be considered as covariates in the prospective analyses. Serum chloride was negatively associated, consistent with known mechanisms that maintain electrochemical neutrality and systemic pH. Diuretic therapy was positively associated with serum bicarbonate, probably explained by mild potassium loss and a tendency towards metabolic alkalosis in these patients. The negative association between serum bicarbonate with serum triglycerides may be due to associated higher plasma fatty acid or ketone body concentrations in those with higher serum triglycerides, tending to cause a mild metabolic acidosis. Interestingly, although eGFR category was significantly associated with quintile of serum bicarbonate in univariate analysis (see Table
1), it was not associated in the multivariate model. This may reflect a U-shaped relationship between lower bicarbonate category and eGFR, with higher proportions of patients in lower bicarbonate quintiles in both the highest and lowest eGFR categories.
The question arises as to why a low serum bicarbonate appears to be an independent risk factor for CHD in type 2 diabetes. The independent associations between both serum triglycerides and insulin therapy and serum bicarbonate in our cohort suggest a possible relationship between bicarbonate and insulin resistance. Supportive evidence for this possibility comes from two sources. Insulin sensitivity in clamp studies was independently associated with serum bicarbonate and apolipoprotein A1: apolipoprotein B ratio in a sample of CKD patients without diabetes [
23]. Experimental metabolic acidosis in healthy humans in vivo and acidification of the medium in which adipocytes were cultured in vitro leads to reductions in serum adiponectin and adiponectin mRNA production, respectively [
24]. Therefore, metabolic acidosis may worsen insulin resistance and thereby increase CHD risk. An alternative explanation may be that the low serum bicarbonate associated with CHD reflects a compensated respiratory alkalosis due to impaired lung function. Type 2 diabetes is associated with a 10–20 % reduction in measures of lung function that was associated with prevalent CHD in a sub-group of FDS patients [
25] while mild respiratory alkalosis has been associated with mortality in a sample of healthy older people [
7]. Nevertheless, in the absence of a measure of systemic pH we are unable to say whether relative acidosis or alkalosis is associated with CHD in our cohort. Further studies are required to determine how serum bicarbonate is associated with CHD in diabetes.
A recent systematic review of the few trials of oral bicarbonate therapy in patients with CKD confirmed preservation of renal function and reduced incidence of progression to renal replacement therapy in bicarbonate-treated patients, and recommended that large, well-controlled trials of bicarbonate treatment should be carried out [
26]. Although some guidelines for CKD management include oral alkalinization therapy for selected patients, current US and Australian guidelines for prevention and management of CKD in type 2 diabetes do not recommend bicarbonate therapy [
1,
27]. Our observation of an independent association between serum bicarbonate and CHD risk adds impetus to calls for high quality trials of bicarbonate therapy with mortality and cardiovascular end-points, as well as renal outcomes [
28,
29]. Our data suggest that enrolling those with bicarbonate concentrations ≤25 mmol/L would include the majority of patients at increased risk. Until this evidence becomes available, patients with diabetes should be encouraged to follow diets rich in fruit and vegetables as a means of providing alkalinization therapy that is safe and as effective as oral sodium bicarbonate in preserving renal function [
30]. For patients with diabetes who have a serum bicarbonate <22 mmol/L, measurement of venous pH to assess acidosis status appears prudent before more definitive alkalinization therapy is suggested.
Limitations of the present study
Serum bicarbonate concentrations were reported in whole numbers (as is usual laboratory practice) which limited the ability to define exact quintile boundaries. Our analyses were based on a single serum bicarbonate measurement and so the contribution of changes in serum bicarbonate to outcome were not captured. We did not have simultaneous pH measurements to allow us to establish acid–base status. Although we had a wide range of plausible confounding factors available to use as covariates in the multivariable analysis it is possible that residual confounding from some unmeasured factor remained. Our cohort did not have sufficient cases of incident end-stage renal disease to allow a meaningful assessment of this outcome. The strengths of the present study include the long duration of follow-up, the range of covariates available for analysis, and the near complete ascertainment of mortality and other endpoints from active ascertainment and linkage to the WADLS.
Authors’ contributions
SAPC was responsible for all biochemical analyses and produced the first draft of the manuscript. WAD performed all statistical analyses, and reviewed and edited the manuscript. KEP assisted with data collection, and reviewed and edited the manuscript. TMED is principal investigator of the FDS, conceived the present sub-study and produced the final version of the manuscript. All authors read and approved the final manuscript.