Background
Small cell lung cancer (SCLC) accounts for approximately 15% of lung cancer and is known for its rapid doubling time and potential for widespread metastases [
1,
2]. For the past several decades, the chemotherapeutic standard of care, cisplatin or carboplatin plus etoposide, has remained essentially unaltered for the treatment of SCLC [
3‐
5]. With the current combination treatment of chemoradiotherapy (CRT), the risk of thoracic recurrence decreases, and as a result, brain metastasis (BM) becomes one of the main types of relapse [
6]. Although patients initially respond well to CRT, the 2-year cumulative risk of developing BM is more than 50% and the median survival time after BM is only 4–5 months. Approximately 65% of patients have detectable BM on autopsy [
7,
8].
Because the blood brain barrier restricts the penetration of most chemotherapeutic agents into the brain, leaving the brain as a susceptible site for relapse, prophylactic cranial irradiation (PCI) has been used for patients with SCLC [
9]. A meta-analysis of studies that mainly included patients with limited stage SCLC (LS-SCLC) [
10] showed that PCI not only decreased the incidence of BM but also prolonged overall survival (OS). Since then, two more trials have focused on PCI in extensive stage SCLC (ES-SCLC) [
11,
12]. The European Organization for Research and Treatment of Cancer (EORTC) trial in 2007 found that PCI also decreased the risk of BM and prolonged OS in ES-SCLC [
11]. However, although the findings of this study led to changes in guidelines and clinical practice, the effect of PCI was still subject to debate, as the patients in the EORTC study did not undergo routine brain MRI. Another randomized trial in patients with ES-SCLC was performed and brain MRI was performed on every patient after initial CRT [
12]. The outcome showed that PCI had no benefit in prolonging OS in patients with ES-SCLC. Thus, in the latest version of NCCN guidelines (V3.2017), PCI has a lower recommendation level (category 2A) in ES-SCLC compared with LS-SCLC (category 1) [
13]. The necessity of PCI remains controversial.
The aim of the present study was to reassess the effect of PCI in SCLC by performing a systematic review of randomized controlled trials (RCTs) published in the literature over the past 30 years.
Discussion
Our present meta-analysis pooled 7 RCTs that evaluated the role of PCI in 2114 patients with SCLC. Interestingly, two RCTs that both studied PCI in patients with ES-SCLC reported opposite outcomes in terms of OS. In general, our meta-analysis revealed a positive role of PCI in improving survival and reducing the risk of BM. However, in subgroup analyses of OS, we found the pooled positive outcome was rather questionable.
Most of the randomized trials had showed a significant decrease of BM incidence; however, none of them individually could demonstrate a significant improvement in OS. Meert et al. [
23] revealed positive role of PCI in BM and OS in patients in CR after chemotherapy. More recently, the retrospective study by Patel et al. [
24] that involved almost 8000 patients supported the results, with a significant improvement in both overall and cause-specific survival in favor of PCI. However, these positive results were questioned by the most recently published Japanese trial [
12]. This recent study found PCI had no benefit in prolonging OS in patients with a confirmed absence of BM when patients received periodic MRI examination during follow-up (HR = 1.27; 95% CI, 0.96–1.68;
P = 0.094). According to our meta-analysis, the pooled HR showed a slight OS benefit with PCI (HR = 0.81). However, the heterogeneity was high (I
2 = 74.1%,
P = 0.001). Thus, we carried out subgroup analyses to explore the heterogeneity in OS according to three aspects: extent of disease, response of initial chemotherapy and the use of brain imaging after initial CRT. The benefit was consistent among subgroups defined according to the extent of disease (
P = 0.67
) and response of initial chemotherapy (
P = 0.76). However, the third subgroup, divided by the use of brain imaging after initial CRT, appeared to be discordant: the subgroup of brain imaging after initial CRT showed no OS improvement with PCI (HR = 0.94; 95% CI; 0.74–1.18) while the subgroup without brain imaging demonstrated an OS benefit (HR = 0.70; 95% CI: 0.57–0.85). This outcome was partly in concordance with the result of Takahashi’s group [
12]. However, it is difficult to explain why the group without brain imaging achieved favorable HR. We speculate that the favorable pooled HR may somewhat be due to the recruitment of asymptomatic patients. In other words, trials that did not perform brain imaging after initial CRT might have included a substantial number of patients who already had BM, and asymptomatic BM patients had a worse prognosis. According to Hochstenbag et al. [
25], asymptomatic BM was present in about 15% of patients with SCLC at diagnosis. Further, Manapov et al. [
26] revealed that 32.5% of patients with LS-SCLC suffered relapse with BM immediately before PCI. The presented treatment for pre-PCI patients with detected BM consisted of either whole-brain radiation alone (WBR) with 3.0 Gy fractions to a total dose of 30 Gy or WBR with 2.0 Gy fractions to a total dose of 40 Gy as a part of second-line CRT with topotecan. Therefore, compared with observation, this subset of asymptomatic BM patients could benefit from the commonly used PCI regimen of 2.5 Gy fractions to a total dose of 25 Gy. As a result, trials that did not perform brain imaging after initial CRT showed a favorable pooled HR of OS.
Among the trials that enrolled patients with imaging proof of no BM, our results showed that these patients benefited little from PCI. Of note, though the pooled HR of this subset of patients was unfavorable, these patients actually had longer survival time. We checked the median survival time of the patients in the 7 trials. In three trials that only enrolled ES-SCLC patients, the median survival times (months) for the PCT group and not PCI group were as follows: 11.6 vs. 13.7, 9.6 vs. 7.9 and 6.7 vs. 5.4, respectively. Among these trials, only the Japanese trial performed brain MRI after initial CRT and before enrollment and reported longer median survival times (11.6 and 13.7) than the other two trials. The above findings were supported by recent retrospective studies of PCI in patients with ES-SCLC [
27,
28]. Most of the patients in these recent studies had brain imaging before PCI and the median survival time was similar to the outcome of Japanese trial. The longer median survival time in the trials with brain imaging again affirmed the assumption that a substantial part of patients were enrolled in trials without brain imaging. Therefore, to analyze the efficacy of PCI more accurately, standardized cranial MRI should be taken into consideration in future trials.
In addition to the efficacy of PCI, it is essential to discuss its toxicity. Usually, toxicity of PCI is defined as acute and long-term according to the 3-month cut-off point. Acute toxicity is generally manageable and consists of mostly alopecia, headache, fatigue, nausea and vomiting [
29,
30]. Long-term sequelae such as severe memory loss, intellectual impairment or even dementia and ataxia have been reported in several studies and attributed to PCI [
18‐
20,
29,
31‐
34]. A pooled analysis of the Radiation Therapy Oncology Group (RTOG) randomized trials 0212 and 0214 had shown that PCI was associated with a higher rate of decline in tested and self-reported cognitive functioning [
32,
35,
36]. In addition, many confounding factors, such as age and the toxicity of anticancer drugs, may add to the intolerance of neurotoxicity of PCI. Most recently, Farooqi et al. [
33] found that the risk of neurotoxicity and neurocognitive decline was greater in elderly patients and those with vascular comorbidities after PCI. In this meta-analysis, we were unable to pool the incidence of toxicity because of limited data. Three trials [
11,
18,
20] reported acute reactions and all three found more grade 3 or worse adverse events in the PCI group. Four trials reported late toxicity relating to PCI. Ohonoshi et al. found that late neurologic toxicity was infrequent; only one patient developed a mild deterioration among seven long-term disease-free survivors in the PC1 group. Arriagada et al. [
19] reported that the 2-year rates of abnormalities as indicated by CT scans of the brain were 21 and 27% (relative risk = 1.48;
P = 0.60), respectively. According to Gregor et al. [
20], the proportions of patients showing long-term impairment in each test were substantial but similar in the PCI and observation groups. No significant difference was found between the study groups in role functioning (
P = 0.17), cognitive functioning (
P = 0.07) or emotional functioning (
P = 0.18). An analysis of patterns of care in the USA [
37] reported a high adherence to guidelines; almost 98% of radiation oncologists recommended PCI for patients with ES-SCLC. Considering the increasing risk of toxicity together with the wide implementation of PCI, a critical re-evaluation of PCI is urgent to determine the appropriate management of SCLC patients. More clinical trials focusing on the analyses of late toxicity are needed in the future.
Several systematic reviews and meta-analyses [
10,
38‐
41] regarding the role of PCI for SCLC survival outcomes have been published. Compared with earlier meta-analyses [
23,
38,
39], we set more strict inclusion criteria and excluded trials that didn’t report PCI. Besides, we included several recent trials in our analysis. Two meta-analysis regarding this topic were published this year. Maeng et al. [
40] focused on the role of PCI in patients with ES-SCLC. The aim of their study was to perform a systematic review and meta-analysis to determine the role of PCI in patients with ES-SCLC who received PCI. Yang et al. [
41] analyzed the BM risk in p-stage I patients without PCI and they only pooled retrospective studies which are not as convincing as random trials. The present meta-analysis has several limitations that merit consideration. First, the significant statistical heterogeneity in the OS meta-analysis could not be fully explained. We were unable to explain why decreased BM come along with unfavorable OS outcome. Second, long-term neurotoxicity could not be addressed in this meta-analysis because neuropsychological evaluation was performed in only two of the trials [
12,
19]. Finally, we did not rule out the risk of bias in individual studies, as the number of included articles was less than 10.