OBJECTIVES: We explored whether the selection of a willingness-to-pay (WTP) threshold in cost-utility analysis (CUA) was endogenous to the economic model output in terms of the incremental cost-effectiveness ratio (ICER). This is based on a concern that CUAs are using higher WTP thresholds to compensate for technology of relatively lower value.

METHODS: We conducted a systematic review of U.S. economic literature between 2000-2017, including studies with explicit WTP thresholds and ICERs from U.S. perspectives. We classified the ratio of studies hypothesized to analyze cost-effective health technologies at low WTP thresholds (i.e. $50,000/QALY) vs. higher thresholds (i.e. $100,000/QALY or $150,000/QALY) relative to reported ICERs. We analyzed whether technologies that were cost-effective at high WTP thresholds would still be cost-effective at lower thresholds using a chi-square test.

RESULTS: Among 317 reported ICERs: 185 (58.4%) were <$50,000/QALY; 53 (16.7%) were between $50,000/QALY and $100,000/QALY; 20 (6.3%) were between $100,000/QALY and $150,000/QALY; and 59 (18.6%) were > $150,000/QALY. Although most studies referenced a single WTP threshold (75.1%), 13.6% and 11.4 % of the studies reported multiple WTP thresholds and did not reference WTP threshold, respectively. Among those articles referencing a single WTP threshold, 194 reported that both the ICER and referenced WTP threshold were less than or equal to $150,000/QALY. We dichotomized the ICERs and the WTP thresholds into two categories, either between $50,000/QALY–$100,000/QALY or between $100,000/QALY–$150,000/QALY. This approached illustrated that WTP thresholds were not independent of the ICER results (P<0.00001).

CONCLUSIONS: WTP thresholds represent the hypothesis test for CUA. We shoe an association between a large proportion of CUAs with higher WTP thresholds having greater ICERs, that would otherwise not have been cost-effective at lower standard WTP thresholds. Researchers may select a WTP threshold after the ICER is calculated to infer value that suits their hypothesis.