METHODS
Using Danish national registries3, we identified new users of SGLT2i with T2DM from January 2010 through March 2019 who had hgb measured within 3 months prior to SGLT2i initiation (hgbbaseline) and again after starting SGLT2i. New users of dipeptidyl-petidase-4 inhibitors (DPP4i) with T2DM during the same period with hgb measurements available were used as an active comparator group. Only individuals with hgb available at both time points were included. T2DM was defined as having filled a minimum of one prescription for either metformin (Anatomical Therapeutic Chemical [ATC]: A10BA02) or insulin (ATC: A10A) within 6 months prior to inclusion, or a diagnosis of T2DM by International Classification of Diseases 10th Revision ICD-10 code: E11. Individuals with type 1 diabetes (ICD-10: E10) were excluded. The last HbA1c measured within 6 months prior to initiating SGLT2i/DPP4i was retrieved along with the last creatinine measured within 2 years prior to exposure. Chronic kidney disease (CKD) was defined as a registered diagnosis or estimated glomerular filtration rate < 60 mL/min/1.73 m2 calculated using the 2021 CKD-EPI formula. Hypertension was defined as the use of ≥ 2 antihypertensive drugs within the 6 months prior to inclusion.4

Change in hgb (∆hgb) was calculated as the difference between hgbbaseline and hgbfollow-up and the mean ∆hgb was reported adjusted for hgbbaseline. A subgroup analysis was completed evaluating ∆hgb among participants with a hgbbaseline of < 12 or > 16 g/dL (outside normal range). The effect of SGLT2i vs. DPP4i on ∆hgb was assessed using multivariable linear regression adjusted for hgbbaseline, time between hgb measurements, age, sex, insulin use, cardiovascular disease, hypertension, HbA1c, cancer, and CKD.

In Denmark, ethics committee approval and informed consent are waived for registry-based studies.

RESULTS
We identified 7965 patients with T2DM who initiated an SGLT2i (mean hgbbaseline: 14.2 ± 1.5 g/dL) and 13,765 who initiated a DPP4i (mean hgbbaseline: 13.3 ± 1.9 g/dL) (Table 1). The adjusted mean ∆Hb for SGLT2i users was + 0.43 g/dL [95% CI: 0.43 to 0.44] compared with  − 0.02 g/dL [95% CI: − 0.03 to − 0.02] for DPP4i users (Fig. 1). Median time between hgbbaseline and hgbfollow-up was 93 days [IQR: 40, 205] in the DPP4i group and 105 days [IQR: 60, 203] in the SGLT2i group. In the fully adjusted model, SGLT2i use was associated with an increase in hgb of + 0.53 g/dL [95% CI: 0.50 to 0.56]; p < 0.001, compared with DPP4i use. In the SGLT2i group, 1755 (22.0%) had an unadjusted increase in their hgb > 2 g/dL vs. 1637 (11.9%) in the DPP4i group. In the SGLT2i group, those with a hgb increase > 2 g/dL were more often male, hypertensive, had lower HbA1c, and less cancer than those who did not experience > 2 g/dL increase. Among patients with hgbbaseline < 12 g/dL, the mean ∆hgb for SGLT2i was + 0.64 g/dL [95% CI: 0.54 to 0.73] and DPP4i for + 0.45 g/dL [95% CI: 0.40 to 0.49]. Among patients with hgbbaseline > 16 g/dL, the mean ∆hgb was + 0.12 g/dL [95% CI: 0.06 to 0.18] for SGLT2i and − 0.58 g/L [95% CI: − 0.64 to − 0.51] for DPP4i.

Table 1 Baseline Characteristics of Patients Receiving a DPP4i and SGLT2i
Full size table
Figure 1
figure 1
Mean adjusted change from baseline to follow-up in hemoglobin in the DPP4i group and the SGLT2i group. Mean change in hemoglobin from baseline to follow-up was adjusted for baseline hemoglobin, time between hemoglobin measurements, age, sex, insulin use, cardiovascular disease, hypertension, HbA1c, cancer, and CKD. Abbreviations: SGLT2i, sodium glucose co-transporter 2 inhibitors; DPP4i, dipeptidyl-petidase-4 inhibitors.

Full size image
DISCUSSION
In our nationwide study of adults with T2DM, we identified that new users of an SGLT2i had a significant increase in hemoglobin as compared with DPP4i. Data from randomized clinical trials (RCTs) suggest that SGLT2i increase hemoglobin through increased erythropoiesis and not hemoconcentraion.5 With the growing use of SGLT2i in diabetes, heart failure, and CKD, it is important for clinicians to be aware of the secondary erythropoiesis caused by SGT2i.2 Our results provide an estimate for the expected rise in hemoglobin following an initiation of an SGLT2i. This will allow clinicians to contextualize what they observe in clinical practice to delineate expected rises in hemoglobin from unexpected rises that may require additional workup. Moreover, the increase in hemoglobin has in clinical trials shown to be an independent predictor of cardio-renal benefits.6,7

CONCLUSION
Adults with T2DM who received an SGLT2i as part of routine care experienced a significant increase in hemoglobin level that was not observed in adults receiving a DPP4i. Further studies should be conducted evaluating the clinical significance of SGLTi-associated erythrocytosis.