Models support prophylactic cranial irradiation.

In this issue of the Journal of Clinical Oncology, Lee et al present a rather unique and useful statistical interpretation of the waxing question of prophylactic cranial irradiation (PCI) for patients with small-cell lung cancer (SCLC). The concept of PCI for SCLC has its origins in the comparison of the behavior of this disease with childhood leukemia, a disease that is essentially systemic in nature, widespread at diagnosis, and exquisitely chemosensitive, with a high rate of complete response. Unfortunately, SCLC is also characterized by frequent relapse, with the CNS— especially the brain—acting as a sanctuary site that frequently is involved at relapse. In fact, most clinical observations suggest an approximate 50% risk of developing intracranial metastases within 2 years in the absence of PCI. The widely accepted explanation for this observation is the presence of preexisting micrometastatic disease in the brain, which effectively remains unperturbed by systemic cytotoxic agents, the majority of which do not penetrate an intact blood-brain barrier (BBB) in sufficient quantities. Just as the disease is exquisitely sensitive to chemotherapy, it is similarly responsive to radiotherapy, which is not impeded by the BBB. Therefore, in this context, the use of cranial irradiation to eradicate residual microscopic disease is intuitively logical, although the concept of prophylaxis associated with this is actually a misnomer. A more detailed analysis of this subject raises several important issues: First, PCI is not harmless. Based on the total dose and fractionation as well as sequencing with chemotherapy, serious neurocognitive decline has been reported in some patients receiving PCI. Second, PCI is not universally effective in preventing clinical relapse. There might be a dose-response relationship, but this has not been identified clearly, and high doses are associated with greater neurocognitive decline. Third, many patients experience treatment failure both intraand extracranially at about the same time, and intracranial disease control therefore does not have a significant impact on survival. In such patients, there might be a clinical neurologic benefit from avoiding the overt manifestation of metastatic disease in the brain, but this has not been studied adequately. As a consequence of these issues, the use of PCI traditionally has been restricted to patients with SCLC who have obtained a complete response to therapy. The expectation is that those patients who have not achieved a complete response to therapy are destined to experience treatment failure extracranially, and should not be exposed to PCI. In addition, the sequencing of PCI after completion of systemic therapy would likely result in lower neurotoxicity. The exact dose and schedule for PCI still remains a matter of great debate, but relatively short schedules, employing low doses per fraction and also low total doses are common; one example is the use of 25 Gy in 10 fractions of 2.5 Gy each. Several clinical trials of PCI have demonstrated clearly a dramatic decrease in the development of brain metastasis in patients receiving PCI, and meta-analysis of the data demonstrate that control of intracranial disease actually results in a modest but real survival gain, the magnitude of which is comparable to that achieved with thoracic irradiation in SCLC. This survival benefit is not unexpected, given that the survival of patients who develop brain metastases is only about 3 months, and hence any retardation of this process should lengthen survival. Despite these data, many patients with SCLC do not receive PCI, primarily for issues related to concerns about neurotoxicity. It is in this context that the manuscript by Lee et al provides an interesting decision analysis tool. Their model varies survival rates and the rate of PCIassociated neurotoxicity in SCLC, and allows one to compute quality-adjusted life expectancy (QALE) based on these two variables. Clearly, such a tool permits a clinician to estimate these two outcomes for their patients to project an expected QALE, allowing more informed decision making. To illustrate the potential use of such a model, currently published survival data are a good starting point. The authors use 5-year survival rates of 26% and 22%, with and without PCI, respectively. Using these assumptions, an appropriate presumed risk level for PCI-associated neurotoxicity can now be assessed. The authors use two broad levels of neurotoxicity, low and moderate. As expected, when the rate of neurotoxicity is low, PCI clearly offers superior QALE, even when the degree of neurotoxicity is varied from mild to severe (QALEs of 4.31 v 3.7 and 4.1 v 3.7). Therefore, for a low rate of toxicity, the degree of toxicity does not overwhelm the QALE benefit. Even with a moderate incidence of neurotoxicity, PCI continues to offer superior QALE. Thus, current survival expectations would predict superior QALE for most scenarios for SCLC patients in complete response treated with PCI. What if survival were to improve considerably? For example, in a recent small randomized trial from China, 5-year survival for PCI-treated patients was 35% compared with 24% for the non-PCI group, and the 2-year incidence of brain metastases was 3.8% v 32%, respectively. This would clearly put more patients at risk for delayed PCI-associated neurotoxicity, thereby reducing the benefit of intracranial disease control. The authors have modeled this scenario as well. For example, if the survival in the PCI treated patients reaches 40%, PCI continues to provide better QALE than no PCI for mild neurotoxicity, but not for substantial degrees of JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 24 NUMBER 22 AUGUST 1 2006