Author reply

Abstract

Wronski and Lederman suggest that the inclusion of patients with active systemic disease could have impacted on the results of our trial. Approximately 45% of the patients in our trial had extracranial metastases compared with 37.5% and 31.7%, respectively, in the 2 other published randomized trials evaluating the addition of surgery to radiotherapy in patients with a single brain metastasis.1, 2 In our trial, the two treatment groups were reasonably well-balanced in terms of the presence of extracranial metastases: 49% in the radiation alone group and 41% in the surgery plus radiation group. Hence, there is no bias in the comparison of the two groups. Furthermore, there was no statistically significant difference in survival between radiation and surgery plus radiation in patients with localized disease and in patients with extracranial metastases. We do agree that based on the results of our trial and the other two randomized trials, surgery is not warranted for patients with widespread systemic metastases. Wronski and Lederman suggest that in imbalance in histology between treatment groups could have biased the results in favor of radiation therapy. They argue that 10 patients (24%) in the surgical group had colon carcinoma, which is relatively radioresistant and that 10 patients (23%) in the radiation group had breast carcinoma, which is relatively radiosensitive. In fact, as shown in Table 1 in our article, eight patients in the radiation group had breast cancer. The most important baseline characteristic influencing survival in these patients is the presence or absence of systemic metastases, not individual histology. In fact, in our trial 21 patients in the radiation group had extracranial metastases compared with 17 patients in the surgery plus radiation group. In the analysis of a randomized trial it is usual for the length of survival to commence with the date of randomization. It is obvious that the addition of 2 weeks of radiation treatment to the radiation arm does not impact in a meaningful way on the comparison of median survivals between the two groups. It is the convention of most journals for an intention-to-treat analysis to be performed. Patients are analyzed according to what treatment arm they are randomized to, not by what treatment they actually receive. The inclusion of the ten radiation patients who received surgery in the surgery arm is inappropriate. They may have been patients with a worse prognosis and their inclusion in the surgery arm could potentially bias the results against surgery. We did consider as a worst case scenario, the occurrence of a surgical intervention as an outcome event, i.e., a failure of treatment. This analysis did not change the results of the trial with the cumulative mortality curves of the two treatment groups being superimposable (P = 0.82). The centers participating in our study are well established oncology centers with experienced neurosurgical units. Seventy-four of the 84 patients were contributed by 3 centers. There was no imbalance within centers between treatment arms. The strongest data supporting the efficacy of any treatment comes from a randomized trial.3 There are three published randomized trials evaluating radiation versus radiation plus surgery in patients with single brain metastases.1, 2, 4 We are surprised that Wronski and Lederman base much of their argument concerning potential biases on nonrandomized descriptive series that in fact are far more open to bias. Mark Levine M.D. M.Sc. F.R.C.P.(C)*