SNPs in PI3K-PTEN-mTOR and Brain Metastases in NSCLC—Response

We thank Giovannetti and colleagues for their thoughtful comments about our recent article (1) entitled “Associations between Single-Nucleotide Polymorphisms in the PI3K-PTEN-AKT-mTOR Pathway and Increased Risk of Brain Metastasis in Patients with Non–Small Cell Lung Cancer.” In their letter, the authors express that the sample size should be determined according to the “rule of tens.” This is a simple and effective method used to evaluate prior events. In fact, for prospective studies, the method used commonly is based on the sample size estimation formula of survival analysis [|$(z_\beta + z_{1 - \alpha } )^2 /(P_A P_B \log \,_e^2 \Delta )$|⁠; refs. 2, 3]. Here, P is the proportion of each genotype in our study. According to this formula, the sample size is associated with the gene mutation frequency of the SNP. On the basis of our study and public data, there should be approximately 80 events for analysis of rs2498804 and rs2494732, and we have 99 patients with brain metastases. The rule of tens is more about the ability to control for other variables in a multivariate analysis. We agree that the data may have been too sparse for the control variables we included. However, the unadjusted hazard ratio (HR) is similar (Table 1).

Another issue is whether the Bonferroni correction should be used. It is important to correct the P value of each gene when performing a statistical test on a group of genes. GeneSpring offers four types of multiple testing corrections, and the Bonferroni correction is the most stringent. The more stringent a multiple testing correction, the less false-positive genes are allowed. The trade-off of a stringent multiple testing correction is that the rate of false negatives is very high. To reduce false negatives, P values were not corrected. This may be debatable but is done in practice (3, 4). Furthermore, the association remained significant after the Bonferroni correction for AKT1:rs2494732 (i.e., <0.05/16 = 0.003) in our study. Otherwise, AKT1:rs2498804 and rs2494732 were in strong linkage disequilibrium (D = 0.95), and we believe that AKT1: rs2498804 is not a false-positive gene (1). In addition, we are pleased to find that the association of AKT1:rs2498804 was confirmed in your independent cohorts. We think that this is a very good method to avoid false positives.

Giovannetti and colleagues also raised concerns about the treatment details in our study. We agree with them on this issue. In fact, we have described that the treatment details included surgery, chemotherapy, and radiotherapy. Unfortunately, we have no information on EGFR mutation status or ALK translocations, so we are unable to address this important comment. However, we can report early findings from an ongoing study of 70 patients receiving oral erlotinib in a program supported by the China Charity Federation. The results were the same as those in our study (Table 2). The data have been provided in our response to the reviewers.

Finally, we admit that there are limitations with our study and agree with Giovannetti and colleagues that further research is needed to unravel the functional significance of these polymorphisms. We are currently using immunohistochemistry to detect AKT protein expression in NSCLC primary tissues, and additional functional validation is ongoing.

See the original Letter to the Editor, p. 3623