Affordable Access

Meta-analysis of chemotherapy in head and neck cancer: reply to the letter from JP Pignon et al

British Journal of Cancer
Nature Publishing Group
Publication Date
  • Letter To The Editor
  • Medicine


British Journal of Cancer (1995) 72, 1063 © 1995 Stockton Press All rights reserved 0007-0920/95 $12.00 LETTER TO THE EDITOR Meta-analysis of chemotherapy in head and neck cancer: reply to the letter from JP Pignon et al Sir - I would like to thank Dr Pignon and colleagues at the MACH-NC for their helpful and encouraging response to the recently published meta-analysis of chemotherapy trials in head and neck cancer (Munro, 1995), and also for providing me with a copy of their protocol. I am also grateful for being given the opportunity to reply to some of the points they raise in their letter. I confined my attention to trials published before August 1993; the MACH-NC group include trials still accruing patients in December 1993. This has an obvious impact on the number of trials that would be considered eligible for analysis. The lag between accrual and publication could easily explain why the MACH-NC group was able to identify 20-25 more trials than were included in my study. The comparison of patient numbers between the two overviews is shown in Table I. Trial identification and bias are inevitable problems in any meta-analysis. The sensitivity analyses in the original publica- tion show that failure to include the trials identified by the MACH-NC group would be unlikely to affect the overall conclusion. There are interesting parallels with the recent meta-analyses on the role of thoracic irradiation in small-cell lung cancer. The conclusions from the literature-based analysis of Warde and Payne (1992) were in broad agreement with those from the per-patient analysis of Pignon et al. (1992) (Table II). The information content of a clinical trial will depend upon the number of comparisons made. A three-arm trial makes two comparisons; if there is a control arm and two experimental arms then the control group is, effectively, used twice. In assembling data for a meta-analysis it seems reasonable to count such a trial as two comparisons rather than as a single trial. The point on survival t

There are no comments yet on this publication. Be the first to share your thoughts.


Seen <100 times