Sixty-three studies were included in the review (number of patients not reported): six randomised controlled trials, eight prospective cohort studies, and 49 retrospective observational studies.
Recall rates were reported in thirty-nine studies (median 52.7%, range 11 to 100%), and sample sizes ranged from 22 to 2,921 teeth or 38 to 2,921 roots. Fewer than a third of the studies using radiographic assessment of outcome employed at least two observers to carry out the assessment, and observers were calibrated prior to evaluation of radiographs in eight studies and intra/inter observer reliability tests were carried out in nine studies.
Outcome measure used: Pooled success rates (combining the two examination methods: radiographic examination alone and radiographic examination in combination with clinical findings) were 74.7% (95% confidence interval (CI): 69.8 to 79.5; 40 studies) using strict radiographic criteria, and 85.2% (95% CI: 82.2 to 88.3; 36 studies) using loose radiographic criteria. Meta-regression analysis indicated that success rates based on strict radiographic criteria were significantly lower than success rates based on loose radiographic criteria (10.5%, 95 CI: 4.4 to 16.7; p=0.001). Radiographic criteria were found to be a substantial source of statistical heterogeneity.
Duration after treatment: Pooled success rates increased with longer follow with strict radiographic criteria, ranging from 29.6% (95% CI: 14.2 to 73.3; two studies) at six months, to 67.7% (95% CI: 39.0 to 96.4) at 12 months, to 85.4% (95% CI: 80.3 to 90.6; eight studies) at greater than 48 months. No obvious trend in success rate was found by duration of follow-up when loose radiographic criteria were used.
Year of publication: No obvious trend in success rate was found by decade of publication.
Geographic location: North American countries more frequently reported success rates based on loose radiographic criteria and Scandinavian countries more frequently reported success rates based on strict radiographic criteria. Pooled success rates using loose criteria were 70.3% for Scandinavian countries, 88.1% for North America and 84.5% for 'other' countries. Pooled estimate using strict criteria for Scandinavian countries was 80.5%.
Qualification of operators: One study directly compared outcome of treatment by qualification of operators. This study found no statistically significant difference in success rates for operator based on qualification. Weighted pooled estimates for success (strict/loose) were: 74.8% (95% CI: 67.0 to 82.7; 14 studies) and 83.3% (95% CI: 75.8 to 90.9; 11 studies) for undergraduates; 65.7% (95% CI: 56.3 to 75.1; six studies) and 86.2% (95% CI: 82.9 to 89.5; five studies) for general dental practitioners; 77.2% (95% CI: 64.5 to 89.8; four studies) and 93.1% (95% CI: 91.5 to 94.7; two studies) for postgraduate students; and 84.8% (95% CI: 80.1 to 89.4; 11 studies) and 87.6% (95% CI: 83.9 to 91.3; 17 studies) for specialists.
Meta-regression indicated that none of the study characteristics investigated were found to have a significant effect on the success rates reported (separately on loose or strict radiographic criteria) or account for any of the statistical heterogeneity found.