• Top
• Objective and...
• Results
• Conclusion

Objective and methods

Inter-trial heterogeneity of treatment effect can make the interpretation of individual patient data based meta-analysis difficult. In survival analysis, there may be heterogeneity of baseline hazard between trials, heterogeneity of treatment effect, or both. We present a class of random effect survival models that allows a random treatment effect, a random trial effect, and/or multilevel modeling of treatment effects in an effort to take into account these various components of heterogeneity.

Results

This class of models has been applied to the Meta-Analysis of Chemotherapy in Head and Neck Cancers (MACH-NC), a pooling of 65 trials comparing chemotherapy versus no chemotherapy involving 10,850 patients. Chemotherapy was given either before (neoadjuvant), at the same time (concomitant), or after (adjuvant) loco-regional treatment. Strong between-study heterogeneity is observed in the concomitant group. The use of random effect models confirmed the main results of the fixed effect model: a) an overall benefit of chemotherapy ; b) variation of this effect according to the timing of chemotherapy (with the highest benefit for concomitant and the lowest for adjuvant). Moreover, these models allowed a better understanding of the results by detecting trials with a very low baseline risk, which all happen to concern adjuvant chemotherapy.

Conclusion

This family of random effect models is a useful technique, allowing for a more complete understanding of the data in the presence of heterogeneity.

Have something to say? Post a comment on this article!