In our large population-based study comparing individuals who received IPC insertion to chemical pleurodesis for malignant pleural effusions, no significant difference in post-procedure all-cause mortality was found after balancing baseline characteristics. There were significant differences across several variables at baseline. IPCs were inserted significantly later following a cancer diagnosis compared to chemical pleurodesis procedures, and in more frail individuals. In general, IPCs were more commonly used, however there was significant practice variation regionally, at the hospital-level, and by physician specialty. Despite this, the majority of IPCs were still inserted in the outpatient setting.

Post-procedure time-to-death is similar to a previous study of health administrative data in the US which found the median survival from first thoracentesis for MPE was 88 days (IQR 26–320) [14]. The significantly shorter time from index procedure until death (in the unweighted comparison) in our IPC group compared to chemical pleurodesis group may be explained by the longer time from initial cancer diagnosis to index procedure, as there was no significant difference from time of initial cancer diagnosis until death between groups. Guidelines recommend definitive management for a MPE after an initial thoracentesis if it is symptomatic and recurrent [7, 8]. A previous study of administrative data in the US found that only 24% of individuals received guideline consistent care for MPE [14]. The longer time from cancer diagnosis to index procedure in the IPC group may be partially explained by a higher proportion of IPC patients receiving repeated thoracenteses prior to their index procedure, guideline inconsistent care, compared to chemical pleurodesis patients (27.5% vs. 18.1%). There was also a higher proportion of frail individuals in the IPC group. This may be due to more frail individuals being less likely to be offered chemical pleurodesis due to the potential side effects, or a delay in referral for IPC treatment, during which time they become frail, an independent risk factor for mortality [23].

In our study, 50.2% of IPC patients and 31.5% of chemical pleurodesis patients died before 90 days suggesting there is a significant discrepancy between individuals eligible for inclusion in most RCTs (expected survival more than 3 months) compared to those receiving these interventions in the real-world. In previous RCTs and a network meta-analysis comparing these procedures, no significant survival difference has been found, regardless of whether individuals with expected survival of less than 3 months were excluded [9,10,11,12]. A retrospective cohort study previously found improved survival from time of index procedure and from time of first effusion in the IPC group, however is limited by the non-contemporaneous time periods during which the procedures were used [24]. Our sensitivity analyses all showed consistent findings with our main analysis, except the doubly robust method which showed IPC individuals had higher mortality, which should be further investigated in future studies. We did not analyze cause-specific mortality (such as death resulting from post-procedure complications) due to the potential risk of misclassification bias and the delay in the availability of cause-of-death information within health administrative data [25].

Our study also revealed the drastic practice change which has occurred over the last two decades with increasing IPC use. Between 2015 and 2019 we found IPCs were used to manage MPEs nearly four times more frequently than chemical pleurodesis. This is the opposite found in an earlier observational study from Australia and Spain, where chemical pleurodesis was used twice as often as IPC insertion for MPEs between 2007 and 2013 [26]. Our results likely represent an ongoing trend noted from earlier US data, where the proportion of IPCs for definitive treatment of MPEs increased from 15% in 2007 to 28% in 2011, whereas chemical pleurodesis rates declined between 2009 and 2013 [14, 27]. The steady level of IPC insertion seen in our study is similar to a multicentre study showing a plateau in IPC use after 2012 [28].

We found significant practice variation by specialty, with pulmonologists more likely to place IPCs and surgeons more likely to treat with chemical pleurodesis. This practice variation is consistent with surveys revealing that the majority of interventional pulmonologists favoured IPCs as the primary intervention while thoracic surgeons preferred chemical pleurodesis [29, 30]. One-quarter of pulmonologists previously reported referring to thoracic surgeons for chemical pleurodesis, potentially contributing to the procedure imbalances between specialities [31]. Consistent with our data showing an increased proportion of females in the IPC group, it has previously been found that women are less likely to be referred to thoracic surgery for definitive management of their MPEs, even when gynecologic and breast cancer diagnoses were excluded [32]. Future research is required to better understand what is driving these disparities in care, such as local resource availability and the beliefs of referring physicians.

Individuals tended to receive the treatment that was performed more frequently at the hospital where their index procedure occurred. We also found practice variation at the level of regional health authorities, with some regions performing similar numbers of IPC and chemical pleurodesis procedures, while others performed many more IPC insertions, and only one performing more chemical pleurodesis procedures. In some regions IPC insertion may be limited due to lack of physician or home nursing expertise to facilitate insertions and home drainage. Expanding training may allow all patients to have more equitable access to these treatments. Although guidelines recommend patient characteristics and preferences be taken into account for deciding between interventions, our results suggest that the choice of procedure may be influenced more by the specialty of the treating physician, and referral and practice patterns within the region [7, 8].

The number of individuals with missing data excluded from our survival analysis was predominantly driven by missing institutional level data (hospital type and annual procedure volume) in the IPC group receiving outpatient procedures. The majority of baseline characteristics were not significantly different between those with missing data and the analyzed groups. Although more individuals who were excluded from the IPC group had undergone previous thoracenteses, their time from cancer diagnosis to index procedure was not significantly different, suggesting that the procedures were not necessarily performed later in the disease course, and thus not likely to affect the mortality outcomes. Excluded individuals, had less mesothelioma and other cancers in the IPC group, and less mesothelioma but more other cancers in the chemical pleurodesis group. Although there were statistically significant differences in cancer types between excluded and included individuals, these differences were still small and not felt to be clinically significant.

We did not include cancer therapies received by patients at the time of their index procedure due to the difficulty in categorizing these for such a diverse group of cancers. Prognostic scores (e.g. LENT, PROMISE) have been developed for patients with MPE, in addition to having limitations, the components of these scores were not available for the majority of patients and therefore we were unable to include them [13, 15, 33].

Strengths of our study include the use of large databases to identify individuals treated in a variety of real-world clinical settings and practices, a perspective not previously captured in studies utilizing administrative data and not always available in clinical trials [14, 27, 34]. Previous population-based studies have assessed individuals with MPE, but lacked codes to identify IPC insertions and only evaluated inpatients [27, 34]. Individuals receiving IPCs represented 77.3% of our cohort with 61% of those procedures performed in the outpatient setting. Another study evaluating characteristics associated with ‘guideline consistent care’ in MPE patients between 2007 and 2011 was limited to Medicare patients older than 65, with IPCs only accounting for up to 28% of definitive procedures [14]. Individuals under 65 represented 36% of our overall cohort and 36% of IPC insertions. Given the variation in healthcare systems and practice patterns between countries, the real-world data from Ontario can provide valuable insights for other countries with similar health structures.

Limitations of our study include potential misclassification bias and unmeasured confounding (e.g., systemic anti-cancer treatments received in post-procedure period, re-expanding lung, sclerosing agent used, concurrent diagnostic biopsies) inherently related to using administrative data. Due to the absence of imaging data, we were not able to evaluate the presence of non-expanding lung, which would typically exclude individuals from receiving chemical pleurodesis, or determine when the MPE first developed. We therefore used patients’ initial cancer diagnosis date as a time reference. This may explain the prolonged time from cancer diagnosis to index procedure in some individuals who initially had early-stage disease, before developing recurrence or progression to advanced-stage disease, which ultimately lead to their MPE (Supplementary Fig. 7, Additional File 1). Further, the specific type of sclerosing agent used for chemical pleurodesis was not documented in the health administrative data. A recent network meta-analysis found no significant differences in pleurodesis failure rates between talc slurry and talc poudrage, but did indicate that talc slurry may have fewer failures compared to bleomycin and doxycycline [12]. Furthermore, it remains unclear whether the surgeries performed for chemical pleurodesis were also used for diagnostic purposes. However, the time from cancer diagnosis to the index procedure was not significantly different between patients who underwent surgical chemical pleurodesis and those who received the procedure percutaneously. The absence of a shorter interval for the surgical group may suggest that fewer procedures were performed for diagnostic purposes. We are unable to determine if repeat pleural procedures were ipsilateral or contralateral to the index procedure. However, for the individuals who had repeat procedures, there were higher proportions of individuals with malignancies that are more likely to cause bilateral effusions (e.g. breast and gynecologic cancers), which may require contralateral pleural drainage.

Johns Hopkins ADG and frailty flag were used as surrogates for performance status, which was not available. These have previously been found to accurately predict one-year mortality [35]. Due to the lack of a validated definition of MPE, previous studies of administrative data have defined MPE using inpatient diagnostic codes, however in order to capture the over 60% of individuals who received their IPC as an outpatient, we used a conservative definition to identify procedures performed for MPEs [27, 34, 36]. In general, patient characteristics in our cohort are similar by baseline characteristics to those from other population-based studies of patients with MPEs [14, 27, 34, 36].