Impact of the Proportion of Asian Patients on the Overall Outcome
Up to this point, this review has highlighted the effect of Asian ethnicity on clinical outcomes (survival, response rate and toxicity) in NSCLC. Here, we discuss the proportion of Asians in a randomized study having an impact on the overall HR for survival. The previously discussed FLEX and ISEL studies will be used as examples. In the FLEX study, the ethnic breakdown in the experimental (chemotherapy and cetuximab) and control (chemotherapy alone) treatment arms were similar: Asian 11 and 10%, caucasian 85 and 84% and others 5 and 5%, respectively. A subgroup analysis in the study showed differences in the HR for survival in the Asian (HR: 1.18; 95% CI: 0.73–1.90) and caucasian cohort (HR: 0.8; 95% CI: 0.69–0.93) and an overall HR of 0.87 (95% CI: 0.76–1.00) that was considered to be statistically significant. To simulate the effect of a difference in the proportion of Asian in the FLEX trial, we assumed that the proportion of other ethnic groups ('others') remain constant, whilst the percentage of Asian and caucasian patients vary. When the percentage of Asians in the study was doubled from 10 to 20%, based on a weighted average approach, the overall HR increased from 0.87 (95% CI: 0.76–1.00) to 0.89 (95% CI: 0.77–1.02). When the percentage of Asian patients in the study was reduced from 10 to 5%, the overall HR was reduced to 0.84 (95% CI: 0.73–0.97), and if no Asians were in the study, the HR was 0.83 (95% CI: 0.72–0.95) (Table 4). This simulation shows the effect on overall HR for survival based on a change in the percentage of Asian patients in the study. An increase in the percentage of Asian patients altered the overall result from a positive study, from a statistical viewpoint, to a negative study. Conversely, a survival benefit with the addition of cetuximab to chemotherapy was observed with a smaller number of Asian patients in the simulation, and was more evident if no Asians were in the study (Table 4).
In the ISEL study, the percentage of Asians and non-Asians, which we defined as a sum of caucasian, black and others, in the experimental (gefitinib) and control (placebo) treatment was 21 and 19%, respectively and 79 and 81%, respectively. When the percentage of Asians in the ISEL study population was increased from 20 to 25%, the overall HR was reduced from 0.89 (95% CI: 0.77–1.02) to 0.86 (95% CI: 0.75–0.98). When the percentage of Asians in the study was reduced to 0 or 10%, the HR was 0.92 (95% CI: 0.81–1.05) and 0.89 (95% CI: 0.78–1.02), respectively (Table 5). In this second simulation, with more Asians, the overall result was changed from a negative study (from a statistical viewpoint) to a positive study.
The two simulations emphasized the underlying importance of not just Asian ethnicity, but the percentage of Asians in the study population affecting the overall study result. We have highlighted first, the importance of subanalyzing different ethnic groups, and second, the awareness of the proportion of different ethnic groups when planning the study. Thus, even when Asian ethnicity is balanced in both arms, the proportion of Asians can be just as important.
In global trials, sample sizes have generally been calculated based on the assumption that consistent results are obtained from the subpopulations. This approach, however, may have limitations based on the example discussed above. As such, the Japanese Pharmaceuticals and Medical Devices Agency have suggested two alternative methods. The first method attempts to maintain a balance on the Japanese and total sample size required by calculating the subjects required in order to observe a prespecified relative benefit of the drug in Japanese as compared with the general study population, with a certain probability (usually 80%). As for the second method, recognizing that each subpopulation has a different expected effect from the treatment, the overall sample size calculated is such that the different benefits from treatment expected in respective subpopulations are satisfied with a minimum probability (as above, usually 80%). These two methods result in different sample size estimations, and the advantages and disadvantages of each method should be considered.
Several approaches can be used to investigate the presence of ethnic-specific determinants that confer interethnic differences in sensitivity and toxicity to systemic therapy. One method is the adequate representation of Asian and caucasian patients in a study to enable a comparison of interethnic effects. This usually requires a multicenter, international collaboration. An example using this approach was a Phase II study of patients with NSCLC treated with carboplatin and docetaxel in sites from Australia and Singapore. In this study, neutropenic fever was seen in 26% of patients overall; however, neutropenic fever among Asian patients in the initial cohort was 50%. Following a reduction in the dose of carboplatin for subsequent Asian patients, the neutropenic fever rate was 40%. A second method is the use of a common arm-study protocol, described previously, where the study design, eligibility criteria, staging and treatment was identical, permitting comparison between ethnic groups. Another approach is to study phenotypically enriched populations such as east Asians. This method enables the detection of ethnic-specific gene variants in the population studied, which can be further confirmed in other ethnic populations. For example, the c-met inhibitor, ARQ 197 (tivantinib), is mainly metabolized by CYP2C19. The variants CYP2C19*2 and CYP2C19*3 result in enzyme deficiency leading to the reduced metabolism of ARQ 197 and are more common among Asians compared with caucasians, with an allele frequency of 35 and 15%, respectively. In a Phase I study of Japanese patients treated with ARQ 197, two different recommended Phase II doses were determined: 360 mg twice a day (for extensive metabolisers) and 240 mg twice a day (for poor metabolisers).
Differences in outcome may be due to underlying characteristics related to ethnicity. Ethnicity is assumed to act as a surrogate for genetic, disease, social, behavioral or clinical characteristics and can be classified as intrinsic or extrinsic. These factors tend to be different between Asians and caucasians, being mindful that the North American population is more heterogenous than the Asian population. Intrinsic factors may affect the ability to extrapolate clinical data between geographical regions and include genetic polymorphisms, gene expression, gene mutations, age, gender, end-organ function, lean body mass and body composition. For example, activating EGFR mutations is a prognostic and predictive factor,[51–54] and given the higher frequency of activating EGFR mutations in Asians,[13,25] this factor is likely to contribute to improved survival in Asian patients. In a study reported in Abstract form, putative biomarkers of chemosensitivity such as excision-repair cross-complementing, ribonucleotide reductase M1 and thymidylate synthetase were more likely in adenocarcinoma.
Subtle differences in the genetic backgrounds of Asians and caucasians exist. Three polymorphisms associated with increased EGFR protein production (shorter CA-SSR1 length and variant forms of single nucleotide polymorphisms -216 and -191) were found to be rare in east Asians as compared with other ethnicities, suggesting that the cells of east Asians may make relatively less intrinsic EGFR protein. Colombo et al. recently identified single nucleotide polymorphism rs12072037 as the major polymorphism modulating MFSD2A promoter activity, significant because the MFSD2A gene maps within a linkage disequilibrium block containing the MYCL1-EcoRI polymorphism associated with prognosis and survival in lung cancer patients. Interestingly, the single nucleotide polymorphism rs107232037 is not polymorphic in caucasians, whereas in Asian populations, the frequency of the minor allele of this polymorphism was quite high.
Extrinsic ethnic factors are factors associated with the environment and culture in which a person resides. Extrinsic factors tend to be less genetically and more culturally and behaviorally determined and include the social and cultural aspects of a region such as dietary habits, tobacco and alcohol consumption, environmental pollution, socioeconomic status, attitude to the management of malignancies and treatment compliance. Although differences in medical care such as level of complexity and accessibility to healthcare facilities exist and may influence outcomes,[59–61] recent data suggest lung cancer outcomes are similar if there is unlimited access to care.[62–64]
It should be emphasized that treatment selection in patients with NSCLC is not based on ethnicity. Advances in the understanding of the molecular biology of NSCLC have led to the identification of various molecular subsets of NSCLC. Whilst treatment selection is traditionally based on histology, there has been a shift to personalized treatment in patients with NSCLC with the identification of specific driver oncogenes. EGFR mutations and ALK rearrangement best illustrate the therapeutic relevance of such molecular subsets of driver oncogenes.[37,52] Other subsets of driver mutations in genes including HER2, BRAF, PIK3CA, AKT1, MAP2K1 and MET have been identified. In a recent study in Chinese patients with adenocarcinoma, Li et al. reported that 89.1% of samples harbor known oncogenic driver mutations. In a study of US patients, the Lung Cancer Mutation Consortium reported that 54% of NSCLC samples had single-driver mutations. This information is now being used in 'real-time' to select patients for first-line therapy with targeted therapies specifically directed at the tumor mutation.
Future Oncol. 2012;8(4):451-462. © 2012 Future Medicine Ltd.