Skip to main content
. Author manuscript; available in PMC: 2020 Mar 5.
Published in final edited form as: Pharmacoepidemiol Drug Saf. 2019 May 27;28(7):934–941. doi: 10.1002/pds.4767

Table 1.

Hazard ratios and 95% confidence intervals for myocardial infarctions for the non-steroidal anti-inflammatory drugs example.

PS Adjustment Diclofenac vs Naproxen Ibuprofen vs Naproxen Diclofenac vs Ibuprofen Expected for third contrast
- Unadjusted 0.899 [0.635, 1.274] 1.022 [0.833, 1.253] 0.880 [0.620, 1.248] 0.880
Multi IPTW 0.901 [0.626, 1.295] 0.940 [0.765, 1.155] 0.958 [0.665, 1.380] 0.958
Multi MW 0.886 [0.623, 1.260] 0.877 [0.700, 1.099] 1.010 [0.705, 1.447] 1.010
Multi OW 0.904 [0.636, 1.286] 0.899 [0.724, 1.115] 1.006 [0.704, 1.439] 1.006
Pair IPTW 0.906 [0.634, 1.294] 0.947 [0.771, 1.164] 0.943 [0.652, 1.363] 0.957
Pair MW 0.883 [0.621, 1.256] 0.944 [0.767, 1.160] 1.010 [0.705, 1.446] 0.936§
Pair OW 0.894 [0.629, 1.271] 0.946 [0.770, 1.163] 1.006 [0.704, 1.438] 0.945§

Abbreviations: PS: propensity score; Multi: multinomial generalized propensity score; IPTW: inverse probability of treatment weights; MW: matching weights; OW: overlap weights; Pair: pairwise propensity score.

Direct comparison results for diclofenac vs ibuprofen.

Indirect comparison results for diclofenac vs ibuprofen based on the first two contrast. That is, (hazard ratio for diclofenac vs naproxen) / (hazard ratio for ibuprofen vs naproxen).

§

Non-transitive results for the third contrast. The direct comparison estimates and the indirect comparison estimates are on the opposite side of 1.0 although the difference is small.