Postoperative radiotherapy of non-small-cell lung cancer: A flaw in the PORT meta-analysis

The PORT meta-analysis was carried out with the declared objective of "...a clear decision on whether postoperative radiotherapy is beneficial or not in the treatment of NSCLC." Data on 2128 patients from 9 randomized trials were analyzed using overall survival, disease-free survival, and other endpoints. The results showed a significant adverse effect of postoperative radiotherapy on survival (hazard ratio 1.21 with 95% CI 1.08 - 1.34); there was no significant indication of heterogeneity among the trials (p=0.11 by the chi(2) heterogeneity test). In particular, the trialists saw no reason to suppose that the results were influenced by radiotherapy dose and therefore no indication that any one of the individual schedules used was any more or less detrimental than the others. However, a simple graph of the odds of death vs. total dose is strong evidence in favor of the opposite conclusion (the slope is significantly positive, p<0.001). The presence of such pronounced heterogeneity nullifies the assumption underlying the chi(2) test, that we are dealing with normally distributed fluctuations due to random errors. Despite this, and the many clinical flaws that have been pointed out, this meta-analysis (and the long controversy that had preceded it) was the basis of the decision to stop treating resected lung cancer patients postoperatively with radiotherapy in the UK.