Quantifying bias from dependent left truncation in survival analyses of real world data

In real world data (RWD) studies, observed datasets are often subject to left truncation, which can bias estimates of survival parameters. Standard methods can only suitably account for left truncation when survival and entry time are independent. Therefore, in the dependent left truncation setting, it is important to quantify the magnitude and direction of estimator bias to determine whether an analysis provides valid results. We conduct simulation studies of common RWD analytic settings in order to determine when standard analysis provides reliable estimates, and to identify factors that contribute most to estimator bias. We also outline a procedure for conducting a simulation-based sensitivity analysis for an arbitrary dataset subject to dependent left truncation. Our simulation results show that when comparing a truncated real-world arm to a non-truncated arm, we observe the estimated hazard ratio biased upwards, providing conservative inference. The most important data-generating parameter contributing to bias is the proportion of left truncated patients, given any level of dependence between survival and entry time. For specific datasets and analyses that may differ from our example, we recommend applying our sensitivity analysis approach to determine how results would change given varying proportions of truncation.

Overall survival of patients with metastatic castrate-resistant prostate cancer (mCRPC) who have PTEN tumor suppressor gene loss of function.

58 Background: It is estimated that more than 40% of patients with mCRPC have functional loss of phosphatase and tensin homolog (PTEN) tumor suppressor gene, which is associated with unfavorable prognosis and reduced response to androgen receptor-targeting therapy. We describe patient characteristics and survival outcomes by PTEN LOF status among patients with mCRPC in real-world clinical practice. Methods: We conducted a retrospective cohort study using data from the nationwide Flatiron Health-Foundation Medicine mCRPC Clinico-Genomic database (FH-FMI CGDB), a de-identified database linking data derived from electronic health records with genomic data derived from FMI comprehensive genomic profiling (CGP) tests. The study included patients ≥18 years old, with a primary diagnosis of mCRPC between 1/1/2013 and 6/30/2019 who underwent FMI CGP testing and who had a valid PTEN LOF status. Patients were included if their PTEN report date and mCRPC diagnosis date occurred before death or censoring. PTEN LOF status was identified via FMI’s CGP testing. Kaplan-Meier (KM) methods assessed overall survival (OS) by PTEN LOF status from the date of mCRPC diagnosis (later of metastasis and castration resistance) until death or end of study follow-up. A stratified Cox regression model was used to estimate the hazard of death. The Cox model was adjusted for age, race and sequence of metastasis/CRPC diagnoses, and was stratified by the year of mCRPC diagnosis. Adjustments to account for left-truncation and survivorship bias were made in the KM analysis and the Cox regression model. Results: Among the 458 patients who met the eligibility criteria, 174 (38%) had PTEN LOF. The majority of the study sample (76%) was diagnosed with castration-resistance after metastasis. The PTEN LOF group had a higher percentage of white patients (80% vs. 68%; p= 0.01) compared to the PTEN non-LOF group. The mean age of the study sample was 68 years, and there was no difference in mean age at diagnosis by PTEN LOF status ( p= 0.17). Based on the KM estimates adjusted for left-truncation, the median OS was 14.3 months (95% confidence interval [CI]: 11.1-19.7) in the PTEN LOF group compared to 18.3 months (95%CI: 15.5-21.5) in the PTEN non-LOF group (log-rank p= 0.049). In the multivariable Cox model, the PTEN LOF group had numerically 30% higher risk of death compared to the PTEN non-LOF group (hazard ratio = 1.30; 95% CI: 0.99-1.71; p= 0.057). Conclusions: Among real-world patients with mCRPC in the CGDB, PTEN LOF could be associated with poorer survival outcomes, potentially highlighting the unmet need among these patients. Additional studies with larger cohorts are needed to better evaluate the survival outcomes of patients with PTEN LOF. Therapeutic agents acting on the PTEN/PI3K/AKT/mTOR pathway are being tested in clinical trials, and could potentially improve outcomes in this subgroup of patients with mCRPC.

Penalized regression for left-truncated and right-censored survival data

High-dimensional data are becoming increasingly common in the medical field as large volumes of patient information are collected and processed by high-throughput screening, electronic health records (EHRs), and comprehensive genomic testing. Statistical models that attempt to study the effects of many predictors on survival typically implement feature selection or penalized methods to mitigate the undesirable consequences of overfitting. In some cases survival data is also left-truncated which can give rise to an immortal time bias, but penalized survival methods that adjust for left truncation are not commonly implemented. To address these challenges, we apply a penalized Cox proportional hazards model for left-truncated and right-censored survival data and assess implications of left truncation adjustment on bias and interpretation. We use simulation studies and a high-dimensional, real-world clinico-genomic database (CGDB) to highlight the pitfalls of failing to account for left truncation in survival modeling.

Capture method affects survival estimates and subsequent interpretation of ecological covariates for a long-lived cervid

Understanding what variables affect ungulate neonate survival is imperative to successful conservation and management of the species. Predation is commonly cited as a cause-specific source of mortality and ecological covariates often influence neonate survival. However, variation in survival estimates related to capture methodology has been documented with opportunistically captured neonates generally displaying greater survival than those captured via aid of vaginal implant transmitters (VITs), likely because of increased left truncation observed in the opportunistically captured datasets. Our goal was to assess if 3- and 6-month survival estimates varied by capture method while simultaneously assessing if capture method affected model selection and interpretation of ecological covariates for white-tailed deer neonates captured from three study sites in North Dakota and South Dakota, USA. We found survival varied by capture method for 3-month neonate survival with opportunistically captured neonates displaying up to 26% greater survival than their counterparts captured via VITs; however, this relationship was not present for 6-month survival. We also found model selection and subsequent interpretation of ecological covariates varied when analyzing datasets comprised of neonates captured via VITs, neonates captured opportunistically, and all neonates combined regardless of capture method. When interpreting results from our VIT only analysis for 3-month survival, we found survival varied by three time intervals and was lowest in the first two weeks of life. Capture method did not affect 6-month survival which was most influenced by total precipitation occurring during 3 – 8 weeks of a neonate’s life and percent canopy cover found at a neonate’s capture site. Our results support previous research that capture method must be accounted for when deriving survival estimates for ungulate neonates as it can impact derived estimates and subsequent interpretation of results.