An enhanced prognostic score for overall survival of patients with cancer derived from a large real-world cohort.

BACKGROUND: By understanding prognostic biomarkers, we gain insights into disease biology and may improve design, conduct, and data analysis of clinical trials and real-world data. In this context, we used the Flatiron Health Electronic Health Record-derived deidentified database that provides treatment outcome and biomarker data from >280 oncology centers in the USA, organized into 17 cohorts defined by cancer type. PATIENTS AND METHODS: In 122 694 patients, we analyzed demographic, clinical, routine hematology, and blood chemistry parameters within a Cox proportional hazard framework to derive a multivariable prognostic risk model for overall survival (OS), the 'Real wOrld PROgnostic score (ROPRO)'. We validated ROPRO in two independent phase I and III clinical studies. RESULTS: A total of 27 variables contributed independently and homogeneously across cancer indications to OS. In the largest cohort (advanced non-small-cell lung cancer), for example, patients with elevated ROPRO scores (upper 10%) had a 7.91-fold (95% confidence interval 7.45-8.39) increased death hazard compared with patients with low scores (lower 10%). Median survival was 23.9 months (23.3-24.5) in the lowest ROPRO quartile Q1, 14.8 months (14.4-15.2) in Q2, 9.4 months (9.1-9.7) in Q3, and 4.7 months (4.6-4.8) in Q4. The ROPRO model performance indicators [C-index = 0.747 (standard error 0.001), 3-month area under the curve (AUC) = 0.822 (0.819-0.825)] strongly outperformed those of the Royal Marsden Hospital Score [C-index = 0.54 (standard error 0.0005), 3-month AUC = 0.579 (0.577-0.581)]. We confirmed the high prognostic relevance of ROPRO in clinical Phase 1 and III trials. CONCLUSIONS: The ROPRO provides improved prognostic power for OS. In oncology clinical development, it has great potential for applications in patient stratification, patient enrichment strategies, data interpretation, and early decision-making in clinical studies.

Performance of pharmacovigilance signal-detection algorithms for the FDA adverse event reporting system.

Signal-detection algorithms (SDAs) are recognized as vital tools in pharmacovigilance. However, their performance characteristics are generally unknown. By leveraging a unique gold standard recently made public by the Observational Medical Outcomes Partnership (OMOP) and by conducting a unique systematic evaluation, we provide new insights into the diagnostic potential and characteristics of SDAs that are routinely applied to the US Food and Drug Administration (FDA) Adverse Event Reporting System (AERS). We find that SDAs can attain reasonable predictive accuracy in signaling adverse events. Two performance classes emerge, indicating that the class of approaches that address confounding and masking effects benefits safety surveillance. Our study shows that not all events are equally detectable, suggesting that specific events might be monitored more effectively using other data sources. We provide performance guidelines for several operating scenarios to inform the trade-off between sensitivity and specificity for specific use cases. We also propose an approach and demonstrate its application in identifying optimal signaling thresholds, given specific misclassification tolerances.

Pharmacovigilance using clinical notes.

With increasing adoption of electronic health records (EHRs), there is an opportunity to use the free-text portion of EHRs for pharmacovigilance. We present novel methods that annotate the unstructured clinical notes and transform them into a deidentified patient-feature matrix encoded using medical terminologies. We demonstrate the use of the resulting high-throughput data for detecting drug-adverse event associations and adverse events associated with drug-drug interactions. We show that these methods flag adverse events early (in most cases before an official alert), allow filtering of spurious signals by adjusting for potential confounding, and compile prevalence information. We argue that analyzing large volumes of free-text clinical notes enables drug safety surveillance using a yet untapped data source. Such data mining can be used for hypothesis generation and for rapid analysis of suspected adverse event risk.

Prediction of 60 day case-fatality after aneurysmal subarachnoid haemorrhage: results from the International Subarachnoid Aneurysm Trial (ISAT).

Aneurysmal subarachnoid haemorrhage (aSAH) is a devastating event with substantial case-fatality. Our purpose was to examine which clinical and neuro-imaging characteristics, available on admission, predict 60 day case-fatality in aSAH and to evaluate performance of our prediction model. We performed a secondary analysis of patients enrolled in the International Subarachnoid Aneurysm Trial (ISAT), a randomised multicentre trial to compare coiling with clipping in aSAH patients. Multivariable logistic regression analysis was used to develop a prognostic model to estimate the risk of dying within 60 days from aSAH based on clinical and neuro-imaging characteristics. The model was internally validated with bootstrapping techniques. The study population comprised of 2,128 patients who had been randomised to either endovascular coiling or neurosurgical clipping. In this population 153 patients (7.2%) died within 60 days. World Federation of Neurosurgical Societies (WFNS) grade was the most important predictor of case-fatality, followed by age, lumen size of the aneurysm and Fisher grade. The model discriminated reasonably between those who died within 60 days and those who survived (c statistic = 0.73), with minor optimism according to bootstrap re-sampling (optimism corrected c statistic = 0.70). Several strong predictors are available to predict 60 day case-fatality in aSAH patients who survived the early stage up till a treatment decision; after external validation these predictors could eventually be used in clinical decision making.