Assessing a biomarker's ability to reduce invasive procedures in patients with benign lung nodules: Results from the ORACLE study

Abstract

A blood-based integrated classifier (IC) has been clinically validated to improve accuracy in assessing probability of cancer risk (pCA) for pulmonary nodules (PN). This study evaluated the clinical utility of this biomarker for its ability to reduce invasive procedures in patients with pre-test pCA ≤ 50%. This was a propensity score matching (PSM) cohort study comparing patients in the ORACLE prospective, multicenter, observational registry to control patients treated with usual care. This study enrolled patients meeting the intended use criteria for IC testing: pCA ≤ 50%, age ≥40 years, nodule diameter 8-30 mm, and no history of lung cancer and/or active cancer (except for non-melanomatous skin cancer) within 5 years. The primary aim of this study was to evaluate invasive procedure use on benign PNs of registry patients as compared to control patients. A total of 280 IC tested, and 278 control patients met eligibility and analysis criteria and 197 were in each group after PSM (IC and control groups). Patients in the IC group were 74% less likely to undergo an invasive procedure as compared to the control group (absolute difference 14%, p <0.001) indicating that for every 7 patients tested, one unnecessary invasive procedure was avoided. Invasive procedure reduction corresponded to a reduction in risk classification, with 71 patients (36%) in the IC group classified as low risk (pCA < 5%). The proportion of IC group patients with malignant PNs sent to surveillance were not statistically different than the control group, 7.5% vs 3.5% for the IC vs. control groups, respectively (absolute difference 3.91%, p 0.075). The IC for patients with a newly discovered PN has demonstrated valuable clinical utility in a real-world setting. Use of this biomarker can change physicians' practice and reduce invasive procedures in patients with benign pulmonary nodules. Trial registration: Clinical trial registration: ClinicalTrials.gov NCT03766958.

Fig 1. Patient enrollment in ORACLE and eligibility for analysis.

All IC tested patients in the eligible population could be analyzed for reclassification by the integrated classifier. Registry patients were determined to be ineligible (n = 53) for the following reasons: pre-test pCA > 50% (n = 23), patients not meeting the study eligibility criteria (n = 13), history of non-lung cancer within 5 years of study enrollment (n = 5), previous history of lung cancer (n = 4), baseline CT scan over 90 days prior to enrollment (n = 3), no baseline CT scan (n = 2), < 40 years of age (n = 2), no blood sample submitted (n = 1). Registry patients were excluded from the analysis population (n = 139) for the following reasons: procedure performed prior to receipt of test result (n = 51), lost to follow (n = 37), incomplete clinical data (n = 24), patient deceased prior to endpoint (n = 8), presumptive treatment (e.g. SBRT) without diagnosis (n = 6), withdrew consent (n = 5), tumor other than NSCLC or SCLC (n = 4), other (n = 3), and nodule diagnosis prior to study initiation (n = 1). Control patients were deemed ineligible for the following reasons (n = 47): pCA >50% (n = 15), did not meet eligibility criteria (n = 18), and no diagnosis (n = 14). Controls were excluded from the analysis (n = 17) for the following reasons: presumptive treatment (n = 10), tumor other than NSCLC or SCLC (n = 5) and incomplete clinical data (n = 2).

Fig 2. Propensity score matching metrics.

Panel A shows the distribution of propensity scores for the IC tested (mean score 0.27(SD = 0.71)) and the control group (mean score -0.28 (SD = 0.77)) prior to propensity score matching, indicating a significant difference in propensity scores (p <0.001). Panel B shows the distribution of propensity scores following nearest neighbor matching (1:1, without replacement), with the mean score for the IC group of 0.03 (SD = 0.66) and control group mean score 0.02 (SD = 0.65), comparison propensity scores following matching did not indicate a significant difference (p 0.89). Panels C and D are a comparison of propensity scores for matched IC group patients (y-axis) and control group (x-axis) utilizing a caliper of 1000 (matches every case to a control subject) and a caliper of 0.2 for PSM, respectively. With a caliper of 0.2, the paired differences between propensity scores for the IC and control groups were 0.01 (SD = 0.04), respectively.

Fig 3. Comparison of Pre- and Post-Integrated Classifier (IC) risk distribution for Likely Benign test results.

Distribution of the pre-test risk pCA and post-test risk classifications for IPNs classified as “Likely Benign” by the IC test in the PSM IC Tested group. Prior to IC Testing, 8% (6/72) of patients had a pre-test risk <5% pCA, following testing, 99% (71/72) of patients with a Likely Benign test result had a post-test risk <5% pCA. Indicating that 90% (65/72) of patients with Likely Benign test results were reclassified from the 5–65% risk category to the <5% risk category.

Fig 4. Invasive procedure utilization for patients with benign nodules managed with or without the Integrated Classifier (IC).

The utilization of invasive procedures for patients with benign nodules was compared between the PSM IC and control groups. In the IC group, the proportion of invasive procedures for benign nodules was 5% (95% CI 1.6% to 8.3%, n = 8/162), whereas in the control group it was 19% (95% CI 12.6% to 24.7%, n = 30/161), with an absolute difference, 14%, (95% CI -19.5% to -7.9%) p <0.001, and relative reduction 74%.