Table 3.
Total number of participants | Intervention (n recruited/N invited) | Control (n recruited/N invited) | Baseline (control) recruitment rate | Effect estimate (95% CI) | |
---|---|---|---|---|---|
Bauer, 2004 [27] (Sending $10 or $2 with invitations to return DNA sample (in mouthwash). Comparator was no money. People responding were a subgroup of a smoking cessation trial population). |
300 | 77/200 | 34/100 | 34% | 5% (−7% to 16%) |
Kenyon, 2005 [28] (Sending £5 voucher with invitations to return trial follow-up questionnaire. Comparator was no money. People responding were taking part in a trial to improve neonatal outcomes). |
722 | 156/369 | 108/353 | 31% | 12% (5%–19%) |
Gates, 2009 [29] (Sending £5 voucher with invitations to return trial follow-up questionnaire. Comparator was no money. People responding were taking part in a trial to improve neck injury outcomes). |
2144 | 560/1070 | 493/1074 | 46% | 6% (2%–11%) |
Cumulative results (Bauer + Kenyon + Gates) |
3166 | 793/1639 | 635/1527 | 37% (mean) | 8% (4%–11%) |
The GRADE rating of the certainty in the evidence is moderate
1. Only one of the three trials is scored as low risk of bias on the Cochrane Risk of bias tool; one was uncertain, the other high risk of bias. We considered this a serious limitation and downgraded 1 level
2. The results have some inconsistency in confidence intervals but not the direction of effect and on balance we decided not to downgrade
3. The outcome was direct
4. The results showed signs of imprecision but just for the smallest trial; the confidence intervals of the two larger trials are not too large and wholly on the side of benefit. We did not downgrade
5. There are too few trials for an assessment of publication bias and we have assumed that there is none