Back

Cost Comparison of Fixed Dose Combination vs. Monotherapy Antihypertensive Therapies: A Medicare Part D Analysis (2013 to 2022)

Sami, N.; Rana, D.; Milshteyn, A.; Novin, N.; Dintakurti, A.

2025-12-02 health economics
10.64898/2025.12.01.25341382 medRxiv
Show abstract

Fixed-dose combination (FDC) antihypertensive therapies are widely used to improve adherence and simplify hypertension management, but their long-term cost patterns compared with prescribing individual component drugs remain unclear. We analyzed national Medicare Part D Prescription Drug Event data from 2013 to 2022 to compare the average cost per claim and per beneficiary for six two-drug FDC antihypertensive therapies and the summed costs of their individual components. Analyses were limited to years with complete data for both the combination and its components, and ten-year averages were calculated. Three of the six FDC therapies were less expensive than prescribing their components separately. The greatest savings were seen with valsartan/hydrochlorothiazide ($1 per claim, $174 per beneficiary), lisinopril/hydrochlorothiazide ($4 per claim, $20 per beneficiary), and atenolol/chlorthalidone ($12 per claim, $44 per beneficiary). Amlodipine/valsartan ($63 per claim, $115 per beneficiary), telmisartan/amlodipine ($105 per claim, $386 per beneficiary), and irbesartan/hydrochlorothiazide ($3 per claim, $12 per beneficiary) were more expensive. This national Medicare analysis found that several fixed-dose combination antihypertensive therapies cost less than prescribing their individual components, while others remain more expensive. These results highlight clear differences in drug pricing that may help guide value-based prescribing and formulary decisions in hypertension management.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.