To the Editor:

In their recent article, Li et al¹ reported that survival and neurocognitive function in patients with brain metastases were related to the imaging response assessed at 2 months after whole-brain radiotherapy with 30 Gy.

In patients with brain metastases, the prospective evaluation of neurocognitive function or quality of life after treatment is always problematic because of high dropout rates and the potential introduction of bias. The recursive-partitioning analysis criteria—integrating age, Karnofsky performance status, and uncontrolled primary tumor or extracranial metastases²—have been validated to predict overall survival in patients with brain metastases. Therefore, patients with poorer initial performance status and, thus, shorter life expectancy must always be expected to be over-represented in the first months after treatment, whereas longer follow-up will select for patients with initially favorable status and with good response to treatment.

In some previous reports on experimental treatment approaches for brain metastases, a continuously improving quality of life has been reported where, in fact, the number of assessable patients was drastically reduced with time, apparently introducing a selection bias toward patients with favorable prognostic criteria.³ In their recent report, Li et al avoid this type of bias by presenting a time course of neurocognitive function only for the patients alive at 4 months and alive at 15 months, respectively, thereby ascertaining that identical patients were evaluated at each time point and that results represent the true effect of treatment. Interestingly, they showed that certain domains of neurocognitive function deteriorated in the group alive at 4 months (probably including many with progressive disease), but stabilized or improved in the 15-month survivors.

While this type of analysis is recommendable for future prospective studies of neurocognitive function and quality of life in patients with brain metastases, the response rate at each time point also should be indicated in such studies. Li et al¹ present neurocognitive function data for n = 9 patients alive at 15 months. From the overall survival graph, it appears that approximately 18% of the overall cohort of n = 135 (about 24 patients) were alive and potentially assessable at this time point. The authors do not state how many patients the 4-month data are based on and how many patients were assessable at the 4-month time point. Although response rates in the range of 80% may be unrealistic in patients with brain metastases, information on the compliance to questionnaires or neurocognitive testing is essential to interpret the representativity of the results.

In summary, selection effects on survivors and responders should always be considered in the reporting and interpretation of prospective neurocognitive function or quality-of-life studies in patients with brain metastases.

AUTHOR'S DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.