Eleven RCTs (n=2,228: HDT/ASCT n=1,113; control n=1,115) were included in the review.
When all studies were included in the meta-analysis there was no statistically significant difference in overall mortality between patients receiving HDT/ASCT and those receiving control therapies (OR 1.02, 95% CI 0.86, 1.21). However, the studies included in the analysis showed highly statistically significant heterogeneity (chi-squared 33, P=0.0003). The meta-regression analysis (to investigate this heterogeneity) revealed the use of shortened induction therapy to be significant only when compared with full course induction and up front HDT/ASCT.
The following subgroup analyses found a statistically significant difference between HDT/ASCT and control therapy.
High-dose sequential chemotherapy before a myeloablative regime followed by haematopoietic stem cell support showed a statistically significant advantage over control therapy (2 studies, n=228; OR 0.49, 95% CI 0.28, 0.85, P=0.01).
Studies with a drop-out rate of less than 25% (5 studies, n=537) showed a statistically significant advantage for HDT/ASCT over control therapies (OR 0.63, 95% CI 0.44, 0.90, P=0.01), whereas those with a drop-out rate of more than 25% (6 studies, n=1,691) showed a non significant advantage of control therapies over HDT/ASCT (OR 1.18, 95% CI 0.97, 1.44, P=0.095). There was statistically significant heterogeneity in this analysis (chi-squared 9.1, P=0.0026).
When bulky disease was present in less than 50% of the participants (4 studies, n=838), there was a statistically significant advantage of control therapies over HDT/ASCT (OR 1.31, 95% CI 1.01, 1.70). There was statistically significant heterogeneity in this analysis (chi-squared 4.06, P=0.044).
Overall mortality was statistically significantly improved when a full course of induction therapy was given before HDT/ASCT in comparison with control therapy (OR 0.55, 95% CI 0.35, 0.87, P=0.01). There was statistically significant heterogeneity in this analysis (chi-squared 5.08, P=0.024).
The authors reported that a funnel plot showed no evidence of publication bias.