Jonathan Denne

Identifying a Subpopulation for a Tailored Therapy: Bridging Clinical Efficacy From a Laboratory-Developed Assay to a Validated In Vitro Diagnostic Test Kit

In the United States, regulatory approval of a therapy that is tailored to a subpopulation may require the coapproval of a companion in vitro diagnostic (IVD) tool for identifying that subpopulation. Unfortunately, for many reasons, development of the companion IVD may lag such that it is unavailable during a pivotal clinical trial of the therapy. Instead, a laboratory-developed test (LDT) may be used on clinical trial specimens to identify the subpopulation on whom to evaluate the therapy. However, remaining specimen material is saved so that when the companion IVD is ready for market, the specimens can be retested, in an effort to “bridge” from the LDT to the IVD. Unfortunately, retest results can be missing or invalid because some subjects lack remaining specimen material or because what remains is unevaluable (e.g., due to insufficient specimen material, inadequate specimen quality). We frame the bridging analysis problem as one of estimating drug efficacy in the IVD-defined subpopulation. We develop a closed-form approach, as well as approaches based on multiple imputation and bootstrapping to address the missing data problem. We discuss this in the context of a case study involving a recent submission and approval in the United States of a drug and IVD in oncology.

Attenuation of treatment effect due to measurement variability in assessment of progression-free survival.

For normally distributed data analyzed with linear models, it is well known that measurement error on an independent variable leads to attenuation of the effect of the independent variable on the dependent variable. However, for time-to-event variables such as progression-free survival (PFS), the effect of the measurement variability in the underlying measurements defining the event is less well understood. We conducted a simulation study to evaluate the impact of measurement variability in tumor assessment on the treatment effect hazard ratio for PFS and on the median PFS time, for different tumor assessment frequencies. Our results show that scan measurement variability can cause attenuation of the treatment effect (i.e. the hazard ratio is closer to one) and that the extent of attenuation may be increased with more frequent scan assessments. This attenuation leads to inflation of the type II error. Therefore, scan measurement variability should be minimized as far as possible in order to reveal a treatment effect that is closest to the truth. In disease settings where the measurement variability is shown to be large, consideration may be given to inflating the sample size of the study to maintain statistical power.

Comparison of KRAS genotype: therascreen assay vs. LNA-mediated qPCR clamping assay.

BACKGROUND Kirsten rat sarcoma virus (KRAS) wild-type status determined using a locked nucleic acid (LNA)-mediated quantitative polymerase chain reaction (qPCR) clamping assay (LNA assay) predicted response to therapy in the CRYSTAL (Cetuximab Combined With Irinotecan in First-Line Therapy for Metastatic Colorectal Cancer) study. A companion KRAS diagnostic tool has been developed for routine clinical use (QIAGEN therascreen kit) (QIAGEN Manchester Ltd, Manchester, UK). We wanted to assess the concordance between the validated US Food and Drug Administration (FDA)-approved therascreen assay and the LNA assay in determining the KRAS status of a subset of patients enrolled in the CRYSTAL study. PATIENTS AND METHODS DNA extracted from paraffin-embedded tumor sections was tested for KRAS status using the therascreen assay. Efficacy data from the CRYSTAL study were assessed to determine if the overall survival (OS) hazard ratio for cetuximab in patients identified as having KRAS wild-type status using the therascreen assay was equivalent to that in patients identified as KRAS wild-type using the LNA assay. This was determined by assessing if the concordance between the therascreen assay and the LNA assay met the minimum threshold (prespecified as 0.8) to achieve a significant difference in the OS hazard ratio in favor of the cetuximab + FOLFIRI (5-fluorouracil, leucovorin [folinic acid], irinotecan) arm in the KRAS wild-type population as identified using the therascreen assay. RESULTS Of the 148 samples determined to be KRAS wild-type (therascreen assay), 141 (95.3%) samples were also KRAS wild-type (LNA assay) and 7 samples (4.7%) were KRAS mutant (LNA assay). The prespecified primary concordance measure p was 141/148 = 0.953 (95% confidence interval [CI], 0.905-0.981). The concordance was statistically significantly higher than the prespecified threshold of 0.8 for concordance between the therascreen assay and the LNA assay. Consistent with the concordance exceeding the prespecified threshold, the OS hazard ratio (cetuximab + FOLFIRI arm vs. FOLFIRI arm) in the KRAS wild-type population, determined by the therascreen assay, supported a significant benefit for cetuximab (ie, the 95% CI excluded 1) and was comparable to the OS hazard ratio observed in the CRYSTAL study KRAS wild-type population (LNA assay) even after adjustment for potentially confounding baseline variables. CONCLUSION These results support the utility of the therascreen assay for identifying patients who may benefit from cetuximab therapy for metastatic colorectal cancer.

Meta-analysis examining impact of age on overall survival with pemetrexed for the treatment of advanced non-squamous non-small cell lung cancer

OBJECTIVE In clinical practice, elderly patients are often undertreated relative to younger patients. This meta-analysis was designed to determine whether older patients with non-squamous non-small cell lung cancer (NSCLC) could derive an overall survival (OS) benefit from pemetrexed treatment comparable to that experienced by younger patients in the first-line, second-line, or maintenance settings. METHODS Data from 2671 patients with non-squamous NSCLC participating in four pemetrexed phase III studies were included in a meta-analysis using a random-effects model. Studies included were: JMEI (second-line pemetrexed, N=399); JMDB (first-line pemetrexed/cisplatin, N=1252); JMEN (pemetrexed maintenance after non-pemetrexed/platinum doublet, N=481); and PARAMOUNT (pemetrexed maintenance after first-line pemetrexed/cisplatin, N=539). Patients were predominantly Eastern Cooperative Oncology Group performance status (PS) 0/1. The ratio of OS hazard ratio (HR) (pemetrexed versus control) for younger patients over that for older patients within each study was used as the measure of the differential effect of pemetrexed. Data were examined using age cutoffs of 65 and 70 years. RESULTS Among the four studies, 32% of patients were aged ≥65 years and 14% were aged ≥70 years. The test of heterogeneity among studies was non-significant for subgroups defined by age 65 (P=0.083) and age 70 (P=0.848). The pooled ratio of the OS HR (pemetrexed versus control) in patients <65years to that in patients ≥65 years was 0.92 (95% confidence intervals [CI] 0.67-1.25). Similar results were seen for the analysis using the age 70 years cut-off (0.80 [95% CI 0.62-1.04]). CONCLUSIONS In patients with non-squamous NSCLC with good PS, the effect of pemetrexed on OS was not found to be different in younger and older patients undergoing treatment in the first-line, second-line, or maintenance settings.

Substantial Evidence From a Replicated Secondary Analysis, Followed by a Single Prospective Confirmatory Study

In the context of a New Drug Application in the United States, the term “substantial evidence” of clinical efficacy is usually understood to mean two well-controlled confirmatory studies in which a statistically significant difference is shown on the primary endpoint(s) in both studies. We examine whether this usual understanding can be safely extended to include data packages based on other combinations of studies. For instance, it is not uncommon in drug development for two simultaneous phase 3 confirmatory studies to be designed on basis of limited phase 2 data and then for those studies to fail to show a statistically significant difference on the primary endpoint, but there is a difference seen in both studies on a secondary endpoint or in a particular subgroup. The standard paradigm would suggest that this “finding” would need to be replicated in two further confirmatory studies with the revised endpoint/study population as primary. However, given the evidence from the original two studies, replication of this “finding” in just one further study may provide “substantial evidence” for the purposes of registering an efficacy claim. We examine this and other similar scenarios in terms of whether accepting such registration packages by the Food and Drug Administration would inflate the statistical risk of accepting a false efficacy claim.