Archives

  • 2018-07
  • 2018-10
  • 2018-11
  • 2019-04
  • 2019-05
  • 2019-06
  • 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2019-12
  • 2020-01
  • 2020-02
  • 2020-03
  • 2020-04
  • 2020-05
  • 2020-06
  • 2020-07
  • 2020-08
  • 2020-09
  • 2020-10
  • 2020-11
  • 2020-12
  • 2021-01
  • 2021-02
  • 2021-03
  • 2021-04
  • 2021-05
  • 2021-06
  • 2021-07
  • 2021-08
  • 2021-09
  • 2021-10
  • 2021-11
  • 2021-12
  • 2022-01
  • 2022-02
  • 2022-03
  • 2022-04
  • 2022-05
  • 2022-06
  • 2022-07
  • 2022-08
  • 2022-09
  • 2022-10
  • 2022-11
  • 2022-12
  • 2023-01
  • 2023-02
  • 2023-03
  • 2023-04
  • 2023-05
  • 2023-06
  • 2023-07
  • 2023-08
  • 2023-09
  • 2023-10
  • 2023-11
  • 2023-12
  • 2024-01
  • 2024-02
  • 2024-03
  • 2024-04

    • Strengths of this analysis include that

    2018-11-13

    Strengths of this analysis include that the biomarker assay was analytically validated, the data were from a randomized clinical trial, and the hypothesis was prospectively stated prior to data unblinding (Khleif et al., 2010). The fact that a single biomarker was identified minimized the potential for Type I error, which can be encountered with multiple comparisons (Beckman et al., 2011). Internal validation was provided through exploratory analyses, which showed that decreasing ∆Ct values (i.e., increasing HRG mRNA orexin agonist manufacturer levels) showed a trend for improved clinical benefit (i.e., lower HR) for patients treated with patritumab plus erlotinib compared with placebo plus erlotinib. The maximum likelihood analysis showed that the median ΔCt value was within the range of optimal ΔCt values resulting in the lowest P-value for the HRs. An initial provisional choice of the median value appears to be a reasonable approach for continuous biomarkers with unimodal distributions, where the cutoff value is not already defined, since it provides the largest sample size for subgroup comparisons. However, subsequent validation and iterative refinement of any provisional cutoff against clinical data will always be required as a key component of companion diagnostic development for continuous biomarkers (Fridlyand et al., 2013). Importantly, the distribution of HRG mRNA observed in the HERALD study was unimodal and similar to the distribution observed in commercial samples; thus, it is likely to be representative of patients with NSCLC. The effort to translate this research assay method to a companion diagnostic assay is also an important consideration after the identification of a predictive biomarker. Optimizing an assay method for these purposes may result in changes to the defined cutoff value, the amount of tumor tissue required, and other logistical aspects of clinical trial conduct. Despite samples being obtained from most patients, only 103 samples (i.e., <50% of samples) could ultimately be analyzed for HRG mRNA, resulting in relatively small numbers of patients in the HRG-high and HRG-low subgroups (Khleif et al., 2010). The protocol allowed patients to enter the study if they had an available specimen, allowing them to begin therapy promptly. But it is clear from our results that a quality check of the specimens should have been required before enrollment, despite the possible resultant delays in beginning therapy. As the AACR-FDA-NCI Cancer Biomarkers Collaborative states, absence of high-quality biospecimens is one of the most significant roadblocks to developing and validating biomarkers (Mandrekar and Sargent, 2009). The tumor samples were collected from patients prior to treatment and prior to the selection of the primary biomarker hypothesis or the validation of the HRG assay method. Therefore, it was impossible to stratify the patient population by HRG status. In the absence of stratification and with fewer than half of the tumor samples ultimately being suitable for biomarker measurement, the sample size was not sufficiently large enough to assure that confounding factors did not bias the results (Patterson et al., 2011). In this case, simulations to assess the impact of potential interactions for various significant factors, such as undetected EGFR mutations, were an important mechanism to further qualify the robustness of clinical results when a predictive biomarker was not used in stratification. A large percentage of patients in the HERALD study had an unknown EGFR mutation status (70.3%), and sensitizing EGFR mutations were less prevalent (6.6%) than might be expected among patients with known EGFR mutation status and based on the published literature (~10% incidence rate (Paez et al., 2004; Yang et al., 2008)). This suggests the possibility that the group of patients with unknown EGFR mutation status may have had a high percentage of sensitizing EGFR mutations, potentially leading to an imbalance in EGFR sensitizing mutations in the HRG-high subgroup between the arms of the HERALD study. However, potential imbalances in EGFR mutation status were simulated based on tissue results and found to have been unlikely to have biased the results. Further, the patient demographics and disease characteristics of the HRG-high subgroup were similar to that of the overall ITT population. Finally, EGFR mutation status was also assessed by plasma measurements (data not shown), reducing the percentage of patients with unknown EGFR mutation status to approximately 30%, with no apparent imbalance in EGFR-sensitizing mutations. Nonetheless, the presence of confounding interactions cannot be fully dismissed.