In this real-world, retrospective, observational study, we assessed factors associated with TTNT in patients with OC who received 1LM niraparib monotherapy after 1L platinum-based chemotherapy. In univariable analyses, age ≥ 75 years; BRCAwt/HRd, BRCAwt/HRp, or BRCAwt/HRunk status; receipt of ICS; no surgery; and visible residual disease were statistically significantly associated with shorter observed TTNT, whereas stage I/II or stage III disease at diagnosis were associated with longer observed TTNT. Conversely, BMI, race and ethnicity, platelet count, niraparib starting dose status, time to 1LM therapy from end of 1L therapy, duration of 1L therapy, and use of a bevacizumab-containing 1L regimen were not statistically significantly associated with observed TTNT in this population. After adjustment, BRCAwt/HRd, other BRCA/HRD status, or unknown BRCA/HRD status, receipt of ICS, no surgery, and having visible residual disease were associated with shorter observed TTNT in the multivariable model.

The purpose of TTNT is to function as a proxy for PFS in the real-world setting. In the phase 3 PRIMA trial, median PFS was 13.8 months (95% CI 11.3–14.2 months) in the overall cohort, 19.6 months (95% CI 13.6 to not estimable) in the HRd/BRCAwt subgroup, and 8.1 months (95% CI 5.7–9.4) in the HRp subgroup [6]. These values are consistent with the observed TTNT in the current study of a primarily BRCAwt and/or HRp population and supports the continued use of TTNT as a proxy for real-world PFS.

The observed median TTNT in this study was also consistent with that of another real-world analysis. In that study, patients diagnosed with stage III/IV OC who received any type of 1LM, including bevacizumab, a PARPi, paclitaxel, or gemcitabine in the maintenance setting after 1L treatment, had a median TTNT of 13.3 months (95% CI 11.7–15.8 months) [10]. Although these values are numerically larger than the observed 11.2-month median TTNT in this analysis, in that study, TTNT was estimated from the end of 1L treatment, whereas in this analysis, TTNT was measured from the date of 1LM niraparib initiation. Notably, the time from end of 1L therapy to initiation of 1LM therapy was more than 30 days for most patients in this analysis. Therefore, numeric differences in observed TTNT may be due to how TTNT was measured rather than significant differences in patient outcomes.

Identification of factors that are prognostic of positive outcomes could inform patient selection and improve response to maintenance therapy. Indeed, previous reports and other available evidence have identified several factors that are associated with longer TTNT. One analysis of patient electronic health records data from 2011–2018 investigating risk factors for progression after 1L platinum-based chemotherapy found that patients with stage IV disease or those who underwent ICS had shorter observed TTNT than those with stage III disease or those who underwent PCS [11]. A more recent report found that high-risk factors, including stage IV disease, not undergoing surgery or receiving neoadjuvant therapy plus ICS, visible residual disease after surgery, and BRCAwt disease or unknown BRCA status, were all associated with shorter TTNT than that for patients without any high-risk factors. These findings were regardless of whether the patient underwent maintenance therapy or active surveillance [10]. Consistent with these previous studies, this study provides supportive data that BRCAwt/HRd, other, or unknown BRCA/HRD status; receiving interval cytoreductive or no surgery; and having visible residual disease were associated with shorter observed TTNT for patients receiving niraparib 1LM monotherapy in this real-world population.

Interestingly, some factors shown to be statistically significant in previous analyses, such as race and SES [12,13,14,15], were not identified as prognostic in this study. Data on race and ethnicity are not required in electronic medical records and may not be accurately reflected in this study population. Furthermore, SES was measured at an area level for this analysis, complicating the assessment of its association with individual-level outcomes. Notably, neither the use of bevacizumab in 1L induction treatment nor the dose of niraparib were associated with TTNT after adjusting for other variables, although the latter typically affects other outcomes such as tolerability [16].

As with all studies, the findings presented here should be considered within the context of potential limitations. As a retrospective, observational study, this was not a predictive analysis and could not assess causality; future investigations should evaluate potential mechanisms by which these factors contribute to 1LM effectiveness, including niraparib effectiveness. TTNT was selected as an outcome given the available data, but future studies may consider assessing factors prognostic of real-world PFS. For ease of interpretability, the prognostic variables were modeled as categorical variables, but this reduced statistical power. In addition, the database that was used contains information primarily from USA patients cared for in community-based practices, so these findings may not be generalizable to the overall OC population. Lastly, as with any retrospective database study, the database used in this investigation is subject to missing data and possible errors, and data regarding any care or treatment received outside of this network were not available.

Opportunities for future research include evaluating which factors predict niraparib treatment benefit and if results are similar for other PARPis. Additionally, the association of patient characteristics with other outcome measures such as OS and real-world PFS should be assessed. However, such analyses would be contingent on the availability of these variables in real-world datasets with sufficient follow-up time (for OS).