I presented at a conference in Hawaii where I met my first cardiooncologist, Monika Leja, MD. As scientists, writers and researchers we often laud a longer life expectancy for many cancer survivors but often neglect the related comorbidities associated with toxic drug regimens contributing to increased mortality. Because the risk of cardiotoxic outcomes depends on type of drug, dose, cumulative dose, dose and schedule of administration, route, other drugs in regimen, inclusion of radiotherapy as well as patient demographicspreexisting CV risk factorsit is often hard to tease out universal guidance. This context informs the debate around a recent article published in the New England Journal of Medicine. I think it is particularly relevant as the objective of the study is to potentially spare patients chemotherapy regimens with the use of a prediction tool, MammaPrint, a 70gene signature test. A comparison of two FDA available gene expression profiling tests examined incremental cost effectiveness (ICE). The scatter plot diagram has 4 quadrants running counter clockwise from the upper right corner (I). Quadrant II data points indicate that  Oncotype DX is more costly and less effective than MammaPrint. Comparatively data points in Quadrant IV suggests that Oncotype DX is less costly and more effective than MammaPrint. The graphic demonstrates the majority of data points are in Quadrant II, indicating that MammaPrint is the dominant strategy. The graphic isn't clearly depicted in the reference article for some reason but the trend is somewhat visible. When looking at efficacy and cost MammaPrint seems like a worthwhile clinical tool. 70Gene Signature as an Aid to Treatment Decisions in EarlyStage Breast CancerAs described by the study authors, the primary goal was to "provide evidence of the clinical utility of the addition of the 70gene signature to standard clinicalpathological criteria in selecting patients for adjuvant chemotherapy." Confidence intervals gauge the utility of the data and accuracy of the estimates in the data. You may not be familiar with the onesided bound described in the paperlower onesided bound but it means that 95% of the population is greater than 92% or the noninferiority boundary. The Adjuvant!Online tool was not available but as an FYI predict looks like an interesting model when looking at baseline characteristics. In a welldesigned clinical trial we assume that authors expect no difference between the two treatment arms (chemotherapy and no chemotherapy for example). We fail to reject the null hypothesis unless we observe strong enough evidence to reject. The null hypothesis actually describes no difference between the two groups. The authors here did no such thing. Instead "The primary goal was to assess whether, among patients with highrisk clinical features and a lowrisk geneexpression profile who did not receive chemotherapy, the lower boundary of the 95% confidence interval for the rate of 5year survival without distant metastasis would be 92% (i.e., the noninferiority boundary) or higher." The maximum level of where we reject the null hypothesis is alpha (significance level) and it is typically 0.05. The relevant data is somewhat hard to decipher. If you review the appendix you are reminded that red indicates chemotherapy and blue indicates no chemotherapy. You can also see pvalues that should fail to reject the null hypothesis. I can only think that the expanded y axis leads to a bit of the distortion (inset). The diseasefree survival (panel C and D) seem to show a benefit for patients on chemotherapy and in Table 2 you observe a P value of 0.03, HR 0.64 and 95% CI (0.430.95) FIgure (from article) Survival without Distant Metastasis, Diseasefree Survival, and Overall Survival in the Two DiscordantRisk Groups, According to Randomized Treatment. Shown are the three major survival outcomes — survival without distant metastasis (Panels A and B), diseasefree survival (Panels C and D), and overall survival (Panels E and F) — among patients with discordant risk (i.e., high clinical risk and low genomic risk or low clinical risk and high genomic risk), according to the randomized treatment. This analysis was performed in the intentiontotreat population, which included patients who had discordant risk at the time of enrollment and who were analyzed according to treatment assignment. Timetoevent curves were estimated by means of the Kaplan–Meier method. The insets show the same data on an expanded y axis. From the article: RESULTS A total of 1550 patients (23.2%) were deemed to be at high clinical risk and low genomic risk. At 5 years, the rate of survival without distant metastasis in this group was 94.7% (95% confidence interval, 92.5 to 96.2) among those not receiving chemotherapy. The absolute difference in this survival rate between these patients and those who received chemotherapy was 1.5 percentage points, with the rate being lower without chemotherapy. Similar rates of survival without distant metastasis were reported in the subgroup of patients who had estrogenreceptor–positive, human epidermal growth factor receptor 2–negative, and either nodenegative or nodepositive disease.(also not significant) It is nontrivial to stress that statistical significance is not the same as clinical significance but when the data is so underwhelming how should we decide? The data suggest that it would be a roll of the dicestrictly due to chance. Are you feeling lucky? Thoughtful discussions about content development and outcomes analytics that apply the principles and frameworks of health policy and economics to persistent and perplexing health and health care problems...
Comments are closed.

Bonny is a data enthusiast applying curated analysis and visualization to persistent tensions between health policy, economics, and clinical research in oncology.
Archives
November 2020
Categories 