Updated Results from the Phase III ALEX Trial: Embracing Hazard Ratios

Dr. D. Ross Camidge

By D. Ross Camidge, MD, and Mary W. Redman, PhD

Posted: August 2018

Updates of the primary analysis of the phase III ALEX trial were presented at the American Society of Clinical Oncology Annual Meeting this year.1,2 ALEX was a first-line trial of crizotinib versus alectinib in advanced ALK-positive NSCLC, and the primary endpoint was investigator-assessed progression-free survival (PFS). After longer follow up, the PFS hazard ratio (HR) dropped from 0.47 to 0.43, with a median PFS of 10.9 months for crizotinib (previously 11.1 months) and 34.8 months (95% CI: 17.7-not evaluable; not calculable previously) for alectinib. Independent radiology review committee (IRC) PFS analyses were restricted to the primary analysis time-point, and in 2017 also demonstrated a statistically significant benefit (HR 0.5; medians 10.4 and 25.7 months, respectively). To some extent, this update is a victory lap for alectinib—its role as the first drug to be preferred over crizotinib in this population has been cemented. However, it also raises several interesting issues.

Investigator-Assessed Endpoints vs. IRC, Medians vs. Hazard Ratios

When a lesion is on the cusp of progression, an investigator’s measurements could reflect some clinical bias, and such datasets are usually associated with longer PFS and higher response rates than an IRC’s as in ALEX. However, investigator-assessed endpoints may also be viewed as data more reflective of treatment decisions in the real world. From a practical standpoint, the major reason whether quoting an investigator-assessed versus IRC-assessed datapoint matters is that when the inevitable cross-trial comparisons for other drugs—such as brigatinib, ensartinib, and lorlatinib, which are all completing ”ALEX-like” studies—occurs, we must compare like with like.

Beyond the effects of investigator versus IRC adjudication, the major jump in the median PFS from the IRC-predicted value in 2017 to the investigator-assessed median in 2018 also reflects how medians are most informative when the chances of an event occurs uniformly over time. When a flattening of the Kaplan-Meier curves is observed because a proportion of patients become “long-term survivors,” then the chance of an event is not uniform. If the flattening of the curves occurs around the estimate of the media (as in the alectinib arm of ALEX), more than a single time point can be associated with time points where 50% of patients have/have not experienced the event, and the confidence intervals around the median will be larger. As data mature resulting in more observed event times (less censoring), the point estimate of the median can shift tremendously in the horizontal direction even though the overall magnitude of benefit has changed relatively little. Consistent with this, although the change in the estimated alectinib arm median PFS from the IRC in 2017 to the investigator-assessed update in 2018 seems large, the median for the crizotinib arm, which has a more uniformly distributed PFS curve over time, did not differ that much either between investigator and IRC assessments or between the 2017 and 2018 analyses. Similarly, when next-generation ALK inhibitor drugs are examined in the pure second-line post-crizotinib setting, the median PFS estimates are incredibly robust for the same drug across trials because the durations of control are shorter and the PFS curve is steeper and more uniform over time. For example, the median PFS was 7.1, 8.1, and 8.9 months in three separate trials of alectinib in the post-crizotinib setting.3-5

Although the median is a convenient number to remember—as we start to see PFS curves that are more reminiscent of advanced breast cancer trials than the advanced NSCLC trials of only a few years ago—in reality, the hazard ratio is a better true estimate of benefit in first-line randomized trials and will likely become the number we embrace.

A Glimpse of Overall Survival Benefit?

The ALEX update consolidates the idea that the benefit seen in the first-line alectinib arm remains far superior to what might be expected from the thought experiment of adding the median PFS of first-line crizotinib to the median PFS of alectinib administered post-crizotinib. Why this is so remains hypothetical. Perhaps control of more subclones of disease at baseline, in the body and/or the brain, may lessen the development of biologic diversity over time and directly improve the natural history of the disease compared to chasing after resistance once it has manifested (when, with cells turning over, additional diversity will have been generated). If the administration order of drugs really does matter, such that one can’t play catch up later, this suggests that the nonsignificant trend in ALEX toward improved overall survival with alectinib (HR 0.76; 95% CI [0.50, 1.15]) could turn out to be real. However, now that patients with advanced ALK-positive NSCLC are surviving for years, it will take a long time for these data to mature and for us to know for certain. ✦


1. Peters S, Camidge RD, Shaw AT, et al. Alectinib versus Crizotinib in Untreated ALK-Positive Non-Small- Cell Lung Cancer. N Engl J Med. 2017;;377(9):829-838.
2. Camidge DR, Peters S, Mok T, et al. Updated efficacy and safety data from the global phase III ALEX study of alectinib versus crizotinib in untreated advanced ALK+ NSCLC. J Clin Oncol 36, 2018 (suppl; abstr 9043).
3. Novello S, Mazieres J, Oh IJ, et al. Primary Results from the Phase III ALUR Study of Alectinib versus Chemotherapy in Previously Treated ALK+ Non-Small-Cell Lung Cancer (NSCLC). Ann Oncol. 2017;28(Issue suppl_5): 12990_PR.
4. Shaw AT, et al. Alectinib in ALK-positive, crizotinib-resistant, non-small-cell lung cancer: a singlegroup, multicentre, phase 2 trial. Lancet Oncol. 2016;17(2):234–242.
5. Ou SH, Ahn JS, De Petris L, et al. Alectinib in Crizotinib-Refractory ALK-Rearranged Non-Small-Cell Lung Cancer: A Phase II Global Study. J Clin Oncol. 2016;34(7):661–668.