Table 2.
Reporting of propensity score analysis in included studies
| Variables | Cancer studies* | Cancer surgical studies |
|---|---|---|
| No. of studies/total No. (%) | No. of studies/total No. (%) | |
| Variables used to estimate the PS | ||
| Yes | 29/33 (87.9) | 279/306 (91.2) |
| No/unclear | 4/33 (12.1) | 27/306 (8.8) |
| Inclusion of non baseline variables† | ||
| Yes | 1/29 (3.4) | 26/279 (9.3) |
| No | 28/29 (96.6) | 253/279 (90.7) |
| Comparability of baseline characteristics‡ | ||
| Yes | 25/32 (78.1) | 232/275 (84.4) |
| No | 7/32 (21.9) | 43/275 (15.6) |
| Matching algorithm§ | ||
| Yes | 17/21 (81.0) | 140/219 (63.9) |
| No | 4/21 (19.0) | 79/219 (36.1) |
| Distance metric§ | ||
| Greedy nearest neighbor matching | 4/21 (19.0) | 46/219 (21.0) |
| Greedy matching within specified caliper distances | 8/21 (38.1) | 49/219 (22.4) |
| Greedy matching by digit | 4/21 (19.0) | 19/219 (8.7) |
| Greedy matching without distance metric specified | 0 | 18/219 (8.2) |
| Optimal matching | 1/21 (4.8) | 8/219 (3.7) |
| Not reported | 4/21 (19.0) | 79/219 (36.1) |
| Matching ratio§ | ||
| Yes | 21/21 (100) | 211/219 (96.3) |
| No | 0 | 8/219 (3.7) |
| Use of replacement§ | ||
| With replacement | 1/21 (4.8) | 4/219 (1.8) |
| Without replacement | 5/21 (23.8) | 27/219 (12.3) |
| Not reported | 15/21 (71.4) | 188/219 (85.8) |
| Method to assess comparability of baseline characteristics between matched groups§ | ||
| Standardized difference | 6/21 (28.6) | 31/219 (14.2) |
| C-statistic | 0 | 17/219 (7.8) |
| Absolute difference | 1/21 (4.8) | 1/219 (0.5) |
| Paired test‖ | 2/21 (9.5) | 20/219 (9.1) |
| Independent sample test¶ | 3/21 (14.3) | 31/219 (14.2) |
| Regression | 0 | 2/219 (1.0) |
| Assessed but method not reported | 4/21 (19.0) | 93/219 (42.5) |
| Not assessed | 5/21 (23.8) | 24/219 (11.0) |
| Imbalanced baseline characteristics# | ||
| Yes | 2/25 (8.0) | 31/232 (13.4) |
| No | 23/25 (92.0) | 201/232 (86.6) |
Cancer studies in top medical and cancer journals in 2014 and 2015. Nine studies were included in both cancer studies and cancer surgical studies. PS = propensity score.
The reporting of whether non baseline variables were included was not evaluable if the answer of reporting of variables used to estimate the PS was “no/unclear”.
The reporting of comparability of baseline characteristics in PS analyses was evaluated in studies utilizing matching, weighting, or stratification.
The reporting was evaluated in studies utilizing matching.
The statistical test for paired or matched sample, eg, paired t tests, Wilcoxon signed rank test, and McNemar’s test.
The statistical test for independent sample, eg, unpaired t tests, Wilcoxon rank sum test, chi-square test, and Fisher’s exact test.
The reporting of whether baseline characteristics were imbalanced was evaluated in studies reporting comparability of baseline characteristics.