Effects of SGLT2 inhibitors on haematocrit and haemoglobin levels and the associated cardiorenal benefits in T2DM patients: A meta‐analysis

Abstract To explore the effect and magnitude of effect of sodium‐glucose cotransporter‐2 (SGLT2) inhibitors on haematocrit and haemoglobin and the related cardiorenal benefits in patients with type 2 diabetes mellitus (T2DM), PubMed, Web of Science, CENTRAL and EMBASE were searched to identify eligible trials. Weighted mean differences (WMDs) with 95% confidence intervals (CIs) were calculated using a random‐effects model. Seventy‐eight studies were included in the meta‐analysis. SGLT2 inhibitors significantly increased haematocrit and haemoglobin levels compared with control (total WMD 2.27% [95% CI 2.08, 2.47] and 6.20 g/L [95% CI 5.68, 6.73], respectively). Except for dapagliflozin (p = 0.000), no notable dose‐dependent relationship was revealed for other SGLT2 inhibitors. The effect could be sustained or even slightly increased with long‐term therapy (coef. =0.009, 95% CI [0.005, 0.013], p = 0.000). In subgroup analyses, haematocrit elevation increased with higher body mass index (BMI). A greater haematocrit elevation could be observed in white patients or when compared with active controls. In conclusion, SGLT2 inhibitors increased haematocrit and haemoglobin levels in T2DM patients. Changes in haematocrit and haemoglobin seem to be surrogate markers of improvement in renal metabolic stress, and important mediators involved in cardiorenal protection.


| INTRODUC TI ON
Sodium-glucose cotransporter-2 (SGLT2) inhibitors exert glucoselowering effects by inhibiting SGLT2, which mediates glucose and sodium reuptake in the renal proximal tubule. 1 Many large clinical trials have shown promising benefits of SGLT2 inhibitors on cardiovascular and renal outcomes in participants with type 2 diabetes mellitus (T2DM) in addition to its safe and effective glucose-lowering effects. [2][3][4][5] The cardiorenal protective mechanisms of SGLT2 inhibitors have not been entirely clarified, and the possible mechanisms are multifactorial: improvements in various cardiovascular risk factors, such as hyperglycaemia, dyslipidaemia, hypertension, obesity, hyperuricaemia and albuminuria; reductions in oxidative stress, inflammation, apoptosis and mitochondrial dysfunction; correction of abnormal glomerular haemodynamics; and protection of cardiac structure and function by means of improving myocardial ischaemia, reducing ventricular load and improving myocardial metabolism. [6][7][8][9][10][11] | 541 TIAN eT Al.
Recently, a post hoc mediation analysis from the EMPA-REG OUTCOME trial prompted the idea that changes in haematocrit and haemoglobin mediated approximately half of the decrease in the risk of cardiovascular mortality associated with empagliflozin. 12 Coincidentally, markers of volume status and haematopoiesis exerted a strong mediating effect on heart failure and kidney protection, as revealed by two mediation analysis studies from the CANVAS Program. 13,14 Although there is sufficient evidence of associations between increased haematocrit and haemoglobin and SGLT2 inhibitor therapy amongst patients with T2DM, increases in haematocrit have not been consistently observed in several studies. 15 In addition, data on the impact of such changes in haematological parameters on cardiorenal protection in T2DM patients are still lacking. Hence, we undertook a meta-analysis to explore the effects of SGLT2 inhibitors on erythropoiesis parameters and the associated beneficial cardiorenal protection effects in T2DM patients.

| Search strategy
We carried out literature searches in PubMed, Web of Science, the Cochrane Central Register of Controlled Trials (CENTRAL) and EMBASE from database inception to March 8, 2021. MeSH terms and free-text terms associated with each gliflozin were used. The complete search strategy is presented in Table S1. Moreover, we manually scanned references from retrieved trials, relevant metaanalyses and reviews to search for additional reports.

| Inclusion and exclusion criteria
Trials fulfilling the following inclusion criteria were included: (1) randomized controlled trials (RCTs) conducted with participants with T2DM comparing SGLT2 inhibitors, either as monotherapy or as an add-on to other hypoglycaemic drugs or insulin, with placebo, active control or standard treatment; (2) mean (SD) changes from baseline in levels of erythropoiesis parameters reported for every group or other data allowing for the calculation of the above variables. The primary outcomes were mean (SD) changes in haematocrit and haemoglobin levels, and the secondary outcomes included mean (SD) changes in erythrocytes, reticulocytes and erythropoietin (EPO) levels from baseline. We excluded observational studies, pooled analyses, noncontrolled or nonrandomized trials, articles enrolling nondiabetic or patients with type 1 diabetes mellitus (T1DM) or articles not reporting the outcomes of interest.

| Data extraction and quality assessment
Data were extracted from the full-text and supplementary information of eligible publications according to a prespecified electronic data collection form. For each study, the following data were carefully extracted: first author, publication year, registration number, study design, sample size, baseline characteristics, types and dosages of SGLT2 inhibitors and control compound, treatment duration and changes (SD) in erythropoiesis parameters from baseline. The data from different study periods for the same subjects were presented together.

The quality of RCTs was assessed by the Cochrane Collaboration
Risk-of-Bias Tool consisting of five aspects: random sequence generation, allocation concealment, blinding, incomplete outcome data and selective reporting. Two authors (QT and KYG) independently extracted data and evaluated the quality of every RCT. If there were any divergences, an agreement was reached after discussion; otherwise, another author (LY) was consulted to resolve the conflict.

| Statistical analysis
Data available at the first observation point were used, and a randomeffects model was applied to calculate weighted mean differences (WMDs) with 95% confidence intervals (CIs) between individual SGLT2 inhibitor and control groups. If SDs were not reported, we used relevant guidelines to calculate SDs. 16 For trials providing median and range values, we estimated the mean and SD according to the appropriate formulas. 17 If necessary, mean (SD) changes from baseline were imputed from baseline and endpoint values. Statistical heterogeneity was quantified using the I 2 statistic (significant for I 2 > 50%). 18 Predefined subgroup analyses were carried out for various types and dosages of SGLT2 inhibitors. Sensitivity analysis was conducted utilizing the leave-one-out method. Meta-regression analyses were conducted to assess any possible dose-and study period-dependency between each SGLT2 inhibitor and changes in the levels of haematopoietic parameters. To analyse the treatment duration, data from multiple intervention arms of one observation were collated to a single group, and different comparator data were also merged when needed.
Additional subgroup analyses were employed to assess whether the effect size was related to the type of comparator or baseline characteristics. Moreover, studies with insulin as background therapy were analysed separately. Funnel plots and Egger's test were used to explore publication bias. All statistical analyses were performed with STATA 11.0 (Stata Corporation, TX, USA) and were reported according to the Preferred Reporting Items for Systematic Reviews and Metaanalysis (PRISMA) guidelines. The PRISMA checklist for meta-analysis is presented in Table S3. P < 0.05 indicated statistical significance.

| Subgroup analyses and sensitivity analysis
Several prespecified subgroup analyses were performed to analyse heterogeneity of the effect of SGLT2 inhibitors on haematocrit. When stratified by race, comparator type, baseline duration of T2DM or BMI, there was a statistically significant difference in the increase in haematocrit levels between the therapy and control groups ( Figure 5). No significant differences were found for age, sex, baseline HbA1c, eGFR or haematocrit. The effect of SGLT2 inhibitors combined with insulin therapy on haematocrit was also significant in seven RCTs (total WMD 2.55% [95% CI 1.82, 3.28] p = 0.000, Figure S4). In addition, we evaluated the robustness of the analysis results by means of a leave-one-out sensitivity analysis (data not presented).

| Meta-regression
A meta-regression was conducted to assess whether the increase in haematocrit was dependent on the dose or duration of SGLT2 inhibitor therapy. Except for dapagliflozin, which exhibited a dosedependent relationship with haematocrit level (p = 0.000) ( Figure   S5), no notable relationship was found between the haematocritincreasing effect of individual SGLT2 inhibitors and the various doses (p > 0.05). Furthermore, irrespective of the type of SGLT2 inhibitor, meta-regression was conducted to reveal the association between therapy effect and duration (coef. = 0.009, 95% CI [0.005, 0.013], p = 0.000; Figure S6), showing that the increase in the mean change in haematocrit could be sustained or even slightly increased with long-term therapy.

| Publication bias
No significant publication bias was observed. A symmetrical funnel plot revealed no potential publication bias for the comparison of haematocrit levels between the intervention and control groups, which was confirmed by Egger's test (p = 0.505) ( Figure S7).

| DISCUSS ION
We demonstrated the effects of SGLT2 inhibitors on erythropoiesis parameters through a meta-analysis including 78 RCTs. SGLT2 It is acknowledged that insulin therapy has a persistent antinatriuretic effect leading to elevated plasma volume. 19,20 Our metaanalysis revealed that the effect of SGLT2 inhibitors combined with insulin on haematocrit was also significant in the subgroup analysis of seven RCTs (total WMD 2.55% [95% CI 1.82, 3.28]). In subgroup analyses, it was interesting that haematocrit elevation increased with higher BMI at baseline. So, to further investigate whether there was a lower haematocrit baselines in obese subjects, we correlated the mean of BMI and haematocrit baselines of each RCT and showed no statistically significant difference in the association between BMI and haematocrit (r = −0.224, p = 0.091).
Owing to limited published data and only three RCTs with the mean of BMI baseline less than 25 kg/m 2 , we should be cautious about this result. Further studies are needed to investigate the phenomenon and clarify the underlying mechanisms.
Recently, a meta-analysis on a similar topic was published, but the subjects studied were limited to T2DM patients with chronic kidney disease, and only four articles were included. 21  Given that SGLT2 inhibitors induced glycosuria and natriuresis, increased haematocrit and haemoglobin levels were previously thought to result from haemoconcentration. However, this result F I G U R E 3 Meta-analysis of WMD and 95% CI of changes in haematocrit (%) level for SGLT2 inhibitors, stratified by drug. The WMD of each dose was a combined result of multiple observations. WMDs are from a random-effects model analysis.   26 The inhibition of hepcidin associated with enhanced haematopoiesis maybe contribute to this longterm effect. Further work is needed to explore other mechanisms of this long-term effect.
The mechanisms by which SGLT2 inhibitors augment erythropoiesis remain unclear. EPO is an integral erythropoietic hormone mainly arising from renal erythropoietin -producing cells (REPs).
Hypoxia-inducible factor (HIF) regulates EPO synthesis and secretion in a hypoxia-inducible manner. 27 EPO levels are usually low in diabetic patients, even in the presence of normal renal function. 28 Proinflammatory molecules such as TNFα, IL-1 and IL-6, which are produced by stressed renal tubular epithelial cells, stimulate this conversion of REPs from hypoxia-responsive cells to fibrogenic myofibroblasts that have decreased sensitivity to hypoxic response and produce inflammatory cytokines and fibrotic molecules instead of EPO. 29,30 In addition, inflammatory and fibrotic signals inhibit HIF through overactivation of prolyl hydroxylase domain (PHD) enzymes, even in a pathological hypoxic environment, which further impairs EPO synthesis and secretion capacity. 31  liver can also play a role in hepcidin downregulation. In sum, SGLT2 inhibitors may reduce glucose reabsorption in the proximal tubules, improve renal cortical hypoxia, diminish glucotoxicity, alleviate metabolic stress in renal proximal tubules and adjacent interstitium, promote the recovery of myofibroblasts to REPs, restore HIF activity, upregulate EPO synthesis and secretion, increase the absorption and utilization of iron by inhibiting the production of hepcidin and regulating other iron-regulated proteins, 35 and finally augment haematopoiesis. 36,37 The increase in haematopoietic parameters might have sig- Hyperactivity of the sympathetic nervous system can be lowered by SGLT2 inhibitors, which partially explains the cardiovascular benefit.
Accordingly, the kidney also plays a significant role in cardiovascular protection. 41 Furthermore, elevated haemoglobin is beneficial for the improvement in tissue oxygenation in a damaged cellular environment, thereby exerting cardiorenal protection effects to some extent.
Several limitations of the meta-analysis should be noted. First, significant heterogeneity existed and was not well explained by heterogeneity analyses. Second, almost none of the included trials were designed to assess the effects of SGLT2 inhibitors on increases in erythropoiesis parameters, and a majority of the analysed trials did