Background
Methods
Research protocol
Data source
Inclusion criteria
Data collection
Data analysis
Results
Basic characteristics
Basic characteristics | No. of systematic reviews (N = 640) |
---|---|
Year | |
2008–2011 | 184 (28.75%) |
2015–2020 | 456 (71.25%) |
Region of corresponding author | |
Africa | 15 (2.34%) |
America (North and South) | 193 (30.16%) |
Asia | 193 (30.16%) |
European | 219 (34.22%) |
Oceania | 20 (3.13%) |
Intervention | |
Drug | 483 (75.47%) |
Surgery or device | 108 (16.88%) |
Others (e.g., complementary medicine) | 49 (7.66%) |
Type of meta-analysis | |
Pairwise meta-analysis | 585 (91.41%) |
Network meta-analysis | 17 (2.66%) |
Both pairwise and network meta-analyses | 32 (5.0%) |
Association | 2 (0.31%) |
Individual participant data | 4 (0.63%) |
Protocol | |
Yes | 166 (25.94%) |
No | 474 (74.06%) |
Type of study for meta-analysis | |
Randomized controlled trial (RCT) | 460 (71.88%) |
Non-randomized study of intervention (NRSI) | 89 (13.91%) |
Both RCT and NRSI | 87 (13.59%) |
Not reported | 4 (0.63%) |
Effect estimator | |
Odds ratio, including Peto’s OR | 266 (41.56%) |
Risk ratio or relative risk | 316 (49.38%) |
Hazard ratio | 9 (1.41%) |
Risk difference | 15 (2.34%) |
Others (e.g., mixed use, Chi2, coef.) | 34 (5.31%) |
Systematic review with zero-events studies in meta-analysis | |
Yes | 406 (63.45%) |
No | 88 (13.75%) |
No data available for judgment | 146 (22.80%) |
Power analysis | |
Yes | 16 (2.50%) |
No | 624 (97.5%) |
Methodological information for dealing with zero-events studies | Yes (%, exact CI) | No (%, exact CI) | NC or NA (%, exact CI) | Total |
---|---|---|---|---|
1) In the systematic review, did any meta-analysis contain zero-events studies? | 406 (63.44, 59.57 to 67.18) | 88 (13.75, 11.18 to 16.66) | 146 (22.81, 19.62 to 26.26) | 640 |
2) If zero-events studies were involved, what were the types of zero-events studies? | 406 | |||
• Single-arm-zero-events only | 174 (42.86, 37.99 to 47.83) | ‐ | ‐ | |
• Double-arm-zero-events only | 7 (1.72, 0.70 to 3.52) | ‐ | ‐ | |
• Both single and double | 208 (51.23, 46.25 to 56.19) | ‐ | ‐ | |
• At least containing single-arm-zero-events study | 7 (1.72, 0.70 to 3.52) | ‐ | ‐ | |
• At least containing double-arm-zero-events study | 8 (1.97, 0.85 to 3.85) | ‐ | ‐ | |
• Cannot be judged | 2 (0.49, 0.06 to 1.77) | ‐ | ‐ | |
3) If zero-events studies were involved, whether the authors specified methods to deal with zero-events studies? | 131 (32.27, 27.74 to 37.05) | 275 (67.73, 62.95 to 72.26) | 0 (0.00, 0.00 to 0.90) | 406 |
4) If the authors specified methods, whether they clarified reasons for selecting the method for zero-events studies? | 44 (33.59, 25.58 to 42.36) | 87 (66.41, 57.64 to 74.42) | 0 (0.00, 0.00 to 2.78) | 131 |
5) If zero-events studies were involved, how zero-events studies were dealt with by the authors? | See detailed methods in Fig. 2 | |||
• Discarding single-arm-zero-events studies (174 + 208 + 7 = 389) | 3 (0.77, 0.16 to 2.24) | 384 (98.71, 97.03 to 99.58) | 2 (0.51, 0.06 to 1.85) | 389 |
• Discarding double-arm-zero-events studies (7 + 208 + 8 = 223) | 170 (76.23, 70.09 to 81.66) | 50 (22.42, 17.12 to 28.47) | 3 (1.35, 0.28 to 3.85) | 223 |
6) Whether a sensitivity analysis was employed through at least one different synthesis method for dealing with zero-events studies? | 53 (13.05, 9.93 to 16.73) | 353 (86.95, 83.27 to 90.07) | 0 (0.00, 0.00 to 0.90) | 406 |
7) For those failed to synthesize studies with zero-events, did the authors discuss about the potential impact of such studies (e.g., discarding them) on the results? | 22 (12.94, 8.29 to 18.94) | 148 (87.06, 81.06 to 91.71) | 0 (0.00, 0.00 to 2.15) | 170 |
Dealing with zero-events studies in meta-analysis: in the past
Dealing with zero-events studies in meta-analysis: at present
Comparing the present to the past
Additional analysis
Discussion
Main findings
Implications
How to make full use of zero-events studies in meta-analysis?
-
Second, in case that the total event count of each arm of a meta-analysis is at least 10, the one-stage methods, e.g., the mixed Poisson regression models, the beta-binomial model, the generalized linear mixed models, and the generalized estimating equations, could be considered as the primary option [5, 6, 11, 16, 18, 19, 21, 43].
-
Third, when the total event count of a meta-analysis is < 10 or one-stage methods cannot achieve convergence, two-stage methods such as Peto’s OR, Mantel-Haenszel OR, and Mantel-Haenszel RD could be considered, while Peto’s OR and Mantel-Haenszel OR can only be considered when there are no studies with no events in both arms [6, 26, 38, 44].