The Role of Prophylactic Cranial Irradiation for Non-small Cell Lung Cancer

Discussion

Improvements in RT techniques and systemic treatments have resulted in improved survival for patients with NSCLC, however, the risk of developing BM increases as survival is prolonged, with the brain as the most frequent site of initial treatment failure (5, 24). This review shows that PCI can improve disease-related outcomes without significantly compromising NF and QOL. Seven out of the 10 trials that analysed BM incidence showed a significant reduction in BM following PCI (11, 14-18, 22), with another reporting a borderline significant reduction (19). A challenge when comparing the published data on BM incidence was the lack of consistency in reporting time-specific endpoints. In a recent RCT, patients with stage IIIA–N2 NSCLC were assigned to observation or PCI. A significant reduction in the risk of BM following PCI at 5 years (20.3% vs. 49.9%; p<0.001) was reported, however, this trial closed prematurely due to a low accrual rate (18) and results should be interpreted with caution. This low accrual may be linked to the inclusion criteria which selected patients with stage IIIA/B post radical treatment with stable disease or partial/completed response. Treatment had to be completed with 16 weeks of study enrolment. These criteria may have inadvertently led to selecting patients with a low risk of BM, as 60% of BM occur in the first 6 months after treatment (4) and may have resulted in a small sample population. The RTOG 0214 and MDACC trials demonstrated that PCI significantly reduced the risk of BM for patients with stage III when compared with earlier stages of disease (14, 15). This is line with evidence that incidence of BM is more common in those with stage III (9) and node-positive N2-3 disease (5, 19, 24), highlighting that PCI has a role in this cohort with locally advanced disease. Histology has also emerged as a factor for patient selection; PCI was more effective at reducing BM in patients with the adenocarcinoma subtype (17) when compares to those with squamous cell subtype; this is to be expected as BM are more common in the adenocarcinoma subtype.

The RTOG 84-03 trial was the only RCT not to demonstrate a significant reduction in the incidence of BM but it did identify a delay in the progression of BM, favouring PCI (9% vs. 19%) (21). Based on the findings in this review it is clear that PCI can reduce the incidence of BM and it is reasonable to assume that this reduction may have a positive impact on patient QOL and morbidity.

According to the trials, there was no increase in OS for patients following PCI (13, 16, 18, 21, 23). The only RCT to demonstrate any positive benefit from PCI was that by Li et al., where DFS was significantly lengthened with PCI when compared with the control group and the median OS was marginally but not significantly improved in the PCI arm (18). The other RCT that investigated DFS showed no significant increase in DFS for similar patients with stage III disease (14). There was a trend towards better DFS with PCI at 3 months but this may have been due to more patients developing BM in the observation arm at 1 year. Both of these recent studies closed early due to poor accrual, with half as many patients being assessed in the trial by Li et al. versus the RTOG trial, thus making it very difficult to draw a firm conclusion to the question of survival benefit. DFS should be an end-point in further trials to aid with future patient decision-making.

Only the SWOG trial demonstrated a significantly reduced median survival with PCI, however, unlike the other trials, PCI was administered concurrently with thoracic RT and it used the highest PCI dose (37.5 Gy/15 fractions). The resultant toxicity may have been a factor in this negative survival outcome. There are points of concern in all six RTCs. In the four older trials, the thoracic doses used would not be considered appropriate by current-day standards; this may have resulted in poor locoregional and extracranial control; also, in some cases, the trial was not correctly powered for the study aims. Current evidence-based multimodality treatment may improve local control, prolonging survival, and this may allow for the survival benefit of PCI to become apparent, however, the two most recent RCTs closed early due to poor accrual, leaving the question of survival benefit for current patients with NSCLC unanswered.

By the nature of this disease, those patients who had their BM controlled with PCI are the most likely group to present with relapse at another site as the presence of brain micrometastases is strongly associated with disseminated disease. Therefore, the lack of OS benefit may be a result of uncontrolled disease in the thorax or extracranial metastases. This theory is supported by the MDACC trial (15), which found that 12 out of 14 control patients had BM as first site of relapse, while in the PCI-treated group, BM occurred only after failure at other sites, suggesting that if extracranial sites of relapse can be effectively managed then the development of BM could be delayed allowing the survival benefit of PCI to manifest itself. However, in the absence of an appropriately powered RCT, this cannot be proven.

As with all RT treatments, there is a risk of toxicity. PCI can be accompanied by early and late side-effects (25) which may have an impact on QOL. The potential negative effects on NF and QOL following PCI treatment are one argument against its use. However, BM can also cause a decline in NF and QOL, and lead to poorer prognosis (10-12, 26). Two trials reported no late complications from PCI (15, 21); another reported that late toxicity to the brain was acceptable (11) and a fourth stated the risk of NF decline was low (18). Testosterone level as an indicator of pituitary function were tested following PCI by one trial. This study was the first estimation of endocrine changes after PCI in the literature and no significant changes were found (12). Only two RCTs (RTOG 0214 and that by Li et al.) had set QOL and toxicity as study endpoints (18, 20). The RTOG 0214 reported no decrease in QOL at 6 or 12 months from baseline for patients treated with PCI (p>0.5) (20). This finding was supported by the other RCT using different QOL assessment tools (18). It must be noted that both of these studies closed prior to their targeted accrual so their findings may not be fully validated. By comparison, QOL after PCI in small cell lung cancer also showed no significant difference in QOL for the PCI-treated and observation arms (27, 28).

Early delayed symptoms present typically 6 months or earlier after PCI and include somnolence syndrome, short-term memory loss and acute leukoencephalopathy (29). Late delayed complications are of the greatest concern as they have the most impact on NF (29). PCI may cause toxicities that can lead to a decline in NF. Several studies have reported intellectual/NF impairment and abnormalities on brain computed tomography (11, 12, 16, 20) which were conceivably related to PCI but may also have been related to the other treatment modalities used.

The RTOG 0214 found a trend for greater decline in patient-reported cognitive functioning and a significant decline in memory (20). They also reported significantly worse NF for the PCI-treated arm at 3 months but not at the trial end point of 1 year. Other retrospective/nonrandomised trials provide mixed results, with two studies demonstrating that NF impairment was not exclusive to the PCI arm but also identified in the observation arm, suggesting that these changes are not a direct result of PCI alone (11, 16).

Three studies used MRI to evaluate physical changes to the brain following PCI and revealed white matter abnormalities of higher grades in patients who received PCI (11, 12, 16). (16). Such damage to white matter is correlated with neurocognitive dysfunction (30), however, based on the RCTs in this review, this neurocognitive dysfunction only manifested as a significant memory loss in one study (20). The literature suggests NF decline is not exclusive to patients treated with PCI but is seen in all patients post multimodality treatment for NSCLC and that some of the initial NF changes are reversible over time.

With a shortened life-expectancy, QOL is an important consideration in decision-making; PCI has the potential to reduce QOL through toxicity following treatment. Li et al. reported acute toxicity rates ranging from insomnia in 2% of patients to 27% of patients reporting headache (18). PCI also results in alopecia. Alopecia can be psychologically damaging and lead to personal and social problems (31). Patients who suffered from alopecia after PCI had a median hair regrowth of 2.5 months (12). Interestingly, the main reason for patient refusal to participate in the Gawkowska-Suwińska et al. study was fear of alopecia. A study of PCI for small cell lung cancer commented that decline in QOL was specifically linked to hair-loss and fatigue (32); therefore if PCI is implemented consideration of hair-loss and fatigue management should be areas of focus. Other late toxicities recorded include mild headache or slight lethargy (22.2%), moderate headache or great lethargy (11.1%) and severe headaches (2.5%) (18). Proposed solutions to alleviate certain side-effects and further increase the tolerance of PCI include RT techniques that spare the inner ear, pituitary gland and hippocampus (12).

In the absence of a definitive answer from the trials included in this review, patient choice and shared decision-making play an important role. A discrete choice experiment by Lehman et al. found that 50% of patients with NSCLC, when making hypothetical choices regarding PCI, would accept PCI for no survival advantage long as there was no change in memory or ability to self-care (33). They also found that 90% of patients would accept PCI for a survival benefit of more than 6 months; 52% of patients pre-treatment (78% post-treatment) would have accepted PCI with no survival benefit but for a reduction in BM. These results verify that although survival improvement is most important to patients, a reduction in BM incidence (especially after treatment) is important for a significant proportion of patients. Therefore, if patient preference is taken into account, they may choose PCI despite the lack of evidence for significant survival advantage.

Drawing firm conclusions from the data in this review was challenging due to the small number of trials, which often included a heterogeneous patient population, treatment regimes and PCI dose. In addition, some of the trials had low patient numbers and two RCTs closing prematurely due to poor accrual.