The practice of clinical medicine is more often art than science and, as such, is prone to both subjective and personal interpretations in addition to analytic and objective evaluation of available evidence. The increasing difficulties of introducing evidence-based medicine into clinical practice testify both to the individualism of the medical profession and to the problems of relating conclusions derived from patient populations to decisions about the management of individuals. These observations help to explain the varying fortunes and acceptance of novel treatment interventions. The use of prophylactic cranial irradiation (PCI) to prevent or delay brain metastases in small-cell lung cancer, a disease with a high incidence of involvement at this metastatic site, is one such example.
In the late 1960s, prophylactic cranial irradiation was found to be effective in reducing relapse rates and improving survival for patients with acute lymphoblastic leukemia (ALL). Now, however, prophylactic cranial irradiation has all but disappeared from the treatment protocols for this disease.
The reasons for the stepwise, gradual withdrawal of prophylactic cranial irradiation from this setting were twofold: First, concerns that it might cause neurotoxicity in this curable population of children were confirmed; even low-dose prophylactic cranial irradiationwith associated reduced effectivenesswas found to lead to intellectual detriment on neuropsychometric testing. Second, these findings were accompanied by the development of new, effective alternatives to prophylactic cranial irradiation, in the form of modifications of chemotherapy regimens.
As is usual in pediatric oncology, this change of practice was achieved through a number of sequential and controlled steps and evaluated in large-scale clinical trials. This is an unusual example of evidence-based medicine in practice. Unfortunately, adult oncology has never been able to emulate the discipline of its pediatric cousin in complying with this type of clinical-evidence evaluation.
Prophylactic Cranial Irradiation in SCLC
The story of prophylactic cranial irradiation in small-cell lung cancer is a common example of the more intuitive and tortuous methods for determining clinical practice.
Rationale for Introducing PCIThe reasons for the introduction of prophylactic cranial irradiation in the treatment of small-cell lung cancer are similar to those for acute lymphoblastic leukemia: Brain metastases are a common and cumulatively increasing site of failure. Twenty percent of small-cell lung cancer patients present with central nervous system involvement, 50% develop symptomatic brain metastases by 2 years, and almost all have central nervous system involvement at postmortem.[2-4] Treatment of symptomatic brain metastases is unsatisfactory; only about half of patients achieve a useful palliation, and median survival is less than 3 months. The impact of brain metastases, on socioeconomic issues and on quality of life of patients, is significantly worse than the impact of failure at other metastatic sites, with patients spending prolonged time in hospitals and suffering loss of independence. In this setting, the prevention of brain metastases becomes desirable.
In a chemotherapy- and radiotherapy-sensitive disease like small-cell lung cancer, the reasons why brain metastases remain such a problem are not fully understood. We know that the brain is a site of preferential involvement for all types of lung cancer, and the close relationship between small-cell lung cancer growth and the number of neuropeptides may increase the likelihood of brain as a favored site for small-cell lung cancer metastases. A physical or pharmacologic blood-brain barrier limits access into brain for most water-soluble drugs, and small-volume tumors have not yet developed the tumor-associated vasculature that allows chemotherapy access to established and contrast-enhancing metastases. In light of all these problems, the effectiveness of prophylactic cranial irradiation in the management of brain involvement in acute lymphoblastic leukemia looked promising and suggested its application in small-cell lung cancer.
First Trials in SCLCThe first wave of trials of prophylactic cranial irradiation in small-cell lung cancer was completed in the 1970s and early 1980s.[11-20] The results of those studies confirmed a significant reduction in the incidence of brain metastases following prophylactic cranial irradiation (Table 1). This did not translate, however, into a demonstrable survival benefit. The patient populations had a variable distribution of prognostic factors, such as disease extent and response to induction chemotherapy. Because any advantage in response duration and survival attributable to prophylactic cranial irradiation was likely to be small, it could have been lost among the more powerful determinants. The ability to detect such small differences would have been further impaired by the small sample size, given that individual trials included between 30 and 250 patients, and only three studies had more than 100 randomized patients.[15,19,20]
Specter of NeurotoxicityThe feature that caused major concern was neurotoxicity. Disturbing numbers of long-term survivors of small-cell lung cancer were found to have variable degrees of dementia and radiologic abnormalities of the brain.[21-23] Most, but not all, had received prophylactic cranial irradiation as a part of their treatment. Given that all these reports were based on retrospective or recall evaluation of selected groups of patients, no formal relationship between the likely predisposing causes could be studied. As a well-recognized cause of central nervous system toxicity, prophylactic cranial irradiation was likely to play a role, but the assumption that it was the sole culprit could not be supported from the available evidence. Nevertheless, prophylactic cranial irradiation was banned from most multidisciplinary protocols and an active search for alternative strategies began.
The realization that chemotherapy alone will not prevent central nervous system disease[25,26] led to a new wave of trials involving more than 1,000 small-cell lung cancer patients[27-30]. Most of these patients had limited disease and good responses to induction chemotherapy. In addition, in most cases, prophylactic cranial irradiation was delivered at the time of remission, thus avoiding postradiation chemotherapy, which is known to potentiate neurotoxicity. Prophylactic cranial irradiation doses were between 24 and 36 Gy, given in 2- to 2.5-Gy fractions. Of these modern studies, three large ones (approximately 300 patients each) have been completed and reported in the last 18 months.[27-30] Two of these, CPH and UK02, incorporated prospective neurologic and neurofunctional assessment in their designs.[27,29]
The two French collaborative trials, PCI 85 and PCI 88, ran parallel.[27,28] Both were randomized trials of prophylactic cranial irradiation and controls, but PCI 85 required a fixed dose of 24 Gy in eight fractions, whereas PCI 88 allowed institutional choice of radiation schedules, but 76% of patients received 24 Gy in 8 fractions (Table 2). As reported at the 1995 European Cancer Conference (ECCO), the radiation arm had a highly significant reduction in brain metastases (overall, from 59% to 40%, P < .0001; isolated, from 57% to 39%, P < .0001). Both trials have shown a trend for survival benefit of PCI, which did not reach statistical significance (RR, 0.85; P = .1).
The PCI 85 trial included a prospective neurologic evaluation that showed a low and clinically insignificant rate of radiologic abnormalities on computerized tomography, but no evidence of dementia or serious central nervous system morbidity.
The UK02 trial was initially designed as a three-arm randomized study to compare two dose levels of prophylactic cranial irradiation (24 Gy and 36 Gy in 2-Gy fractions) vs a control arm (no prophylactic cranial irradiation). In the first 3 years, only 100 patients were randomized and the trial was relaunched as a two-arm study, allowing investigators at the participating institutions a choice of prophylactic cranial irradiation schedules. A total of 40% of patients received 30 Gy in 10 fractions, but schedules from 8 Gy in single fractions to 36 Gy in 18 fractions were used. It also incorporated prospective neuropsychometric assessment for all patients recruited at three of the participating institutions. Intake picked up and the trial closed in May 1995, exceeding its target with 314 randomized patients.
The eligibility criteria, which remained the same throughout, comprised limited-disease patients who achieved remission following induction chemotherapy. Patients were randomized within 4 weeks of response assessment. No planned postprophylactic cranial irradiation or concurrent chemotherapy was allowed, although concurrent chest irradiation could be performed in the patients randomized to receive prophylactic cranial irradiation.
ResultsThe trial confirmed the effectiveness of prophylactic cranial irradiation in reducing the rate of brain metastases. At 2 years follow-up, 52% of the control group vs 29% of prophylactic cranial irradiation-treated patients failed in the brain (Figure 1). [Hazard ratio 0.41(95% CI, 0.27 to 0.63), P = .0002].
Interestingly, among the first 100 patients randomized, the advantage was seen only at the higher prophylactic cranial irradiation dose level (36 Gy in 18 fractions): hazard ratio 0.16 (95% CI, to 0.07 to 0.36). Patients receiving lower dose level (24 Gy in 12 fractions) behaved like the control patients. [Hazard ratio 0.71 (95% CI, 0.36 to 1.43)]. The French trials also used 24 Gy, but in 8 rather than 12 fractions; [27,28] it is possible that low-dose prophylactic cranial irradiation schedules require increased fraction size.
Dose-Response RelationshipThe relationship between risk of brain relapse and prophylactic cranial irradiation schedule in the UK02 trial can be seen in Figure 2. The radiation schedules have been converted into a biologically equivalent dose at 2 Gy (BED2), using a linear quadratic model and coefficient for acutely reacting tissues or tumor. The confidence intervals on some of the schedules are large because only a few patients received them, but for the randomized comparisons and the majority of patients who were treated with 30 Gy in 10 fractions, the relationship appears linear.
This demonstration of a dose-response relationship in small-cell lung cancer is a useful contribution to the overall evidence of radioresponsiveness in this disease, and it is the first time a relationship between radiation dose and local control could be seen in the setting of brain metastases. It now needs to be confirmed in a larger randomized trial, which also should try to determine whether the size of the total radiation dose can be at least partially compensated by increasing the individual fraction size. This approach may be considered risky, as large fraction sizes have been thought to predispose to increased levels of central nervous system toxicity.