Morbidity and Mortality Reduction Associated With Polysomnography Testing in Idiopathic Pulmonary Fibrosis

A Population-Based Cohort Study

Nicholas T. Vozoris; Andrew S. Wilton; Peter C. Austin; Tetyana Kendzerska; Clodagh M. Ryan; Andrea S. Gershon


BMC Pulm Med. 2021;21(185) 

In This Article


Our population-level study demonstrates the novel and important finding that undergoing PSG is associated with significantly lower rates of respiratory-related hospitalization and all-cause mortality among individuals with IPF. The credibility of our findings is further corroborated by significantly reduced respiratory-related hospitalization in association with PSG receipt amongst sicker subgroup of individuals, like those with prior history of respiratory-related hospitalization and prior systemic corticosteroid receipt. The finding that only a very small proportion of individuals with IPF undergo PSG highlights that this potentially helpful testing is being infrequently utilized.

Obviously, PSG in of itself is not responsible for the observed positive health outcomes. Rather, subsequent downstream events following PSG (i.e., diagnosis and treatment of sleep breathing disorders) are presumed to explain the observed benefits, with PSG simply reflecting a surrogate marker for these events. Although our study is the first population-based study demonstrating positive outcomes in IPF following PSG, our findings are consistent with the results of two earlier, smaller observational studies that showed improved survival among individuals with combined IPF and OSA in association with PAP adherence.[15,17] Our results are also consistent with three, small observational studies reporting improvements in scores on multiple quality-of-life instruments in association with treatment of OSA in IPF.[15,17,39] Our findings of improved health outcomes differ from another observational study, where neither having OSA, nor being adherent with PAP therapy, generally influenced all-cause mortality or progression-free survival, although in this study only a minority of individuals with interstitial lung disease had IPF (32.5%).[26] Although significant reductions in morbidity and mortality were observed in association with PSG receipt among individuals with IPF, only a very small proportion of our cohort (4.0%) underwent this beneficial testing, underscoring its wide under-utilization in the IPF population. Early identification and treatment of sleep breathing disorder may be important, as there is some evidence to suggest that more advanced degrees of sleep breathing disorder in IPF are associated with worse health outcomes.[16]

To account for measured differences between our exposed and control groups, we performed rigorous propensity score matching, adequately balancing on 41 covariates, including on multiple markers on IPF severity, general health status, comorbidities and health care system utilization. Furthermore, the fact that, as with the exposed group, control group entry was based on receipt of a test (i.e., spirometry) serves to makes it less likely that changes in overall health status or health-seeking behaviour explain our results. Reduced respiratory-related hospitalization in association with PSG in the subgroups of individuals with prior history of respiratory-related hospitalization and previous systemic corticosteroid receipt (which are sicker subgroups of individuals) decreases the likelihood that our findings are as result of 'healthy user' bias. Small sample size may account for the fact that rate of all-cause was not significantly lower in association with PSG in the aforementioned two subgroups, and that significantly improved outcomes were not observed in the sensitivity analyses by CHF comorbidity, as point estimates were below 1.00 for all outcomes across all subgroups.

Our study has several limitations. Our study is hypothesis-generating and causation cannot be concluded as the explanation for our findings. Unmeasured differences between our exposed and control groups could explain our findings. Information on symptoms, lung function and exercise capacity measures, oxygenation status, and extent of fibrosis on imaging, were not available in our health administrative databases. While we balanced the exposed and control groups on a number of important indicators of IPF severity (including history of respiratory-related hospitalization, history of ICU admission, CHF, and previous systemic corticosteroid and respiratory antibiotic receipt), we acknowledge that these are not all validated severity markers. Any unmeasured IPF severity markers would most likely track disease severity markers that we did have information on, and these were in fact consistently more prevalent among exposed versus control individuals, before propensity score matching. Specifically, before propensity score matching, compared to controls, individuals undergoing PSG were more frequently admitted to hospital or ICU for respiratory-related reasons in the preceding year, had CHF and other forms of cardiovascular disease, had other pulmonary comorbidities, and received a systemic corticosteroid and a respiratory antibiotic. Any unmeasured IPF severity markers would logically be anticipated to track these, and if persistently unbalanced after propensity score matching, then bias the analysis against the exposed group, and yet, better health outcomes were observed in association with PSG receipt among these individuals. Our IPF identification algorithm, while developed by internationally-recognized IPF experts[1] and previously applied,[1,27,28,29] has not been validated, and we also lacked a validated measure of disease duration. The IPF algorithm largely relies on a J84.1 coding occurring in the context of a hospitalization or emergency room (ER) visit, which has the potential to under-capture individuals with milder degrees of disease, who would less likely present to hospital. Therefore, our results may not be applicable to all individuals with IPF. However, if individuals with milder degrees of IPF were less likely included in our study, this would only serve to decrease possible 'healthy user' bias influencing our results. Although we used PSG receipt as a marker for diagnosis and treatment of sleep breathing disorder, we acknowledge the limitations of this approach, that undergoing a PSG does not mean that sleep breathing disorder was indeed established in an individual or that any diagnosed sleep breathing disorder was being appropriately treated. However, given the high frequency with which sleep breathing disorders is known to occur among individuals with IPF,[10–17] it logically follows that the vast majority of individuals in our control group will have undiagnosed/untreated sleep breathing disorder, and therefore, PSG receipt becomes not an unreasonable surrogate marker. A sensitivity analysis where receipt specifically of therapeutic PSG was used to identify exposed individuals (which may be a superior marker for diagnosis and treatment of sleep breathing disorder) could not be undertaken because of small sample size (only 12/201 [6%] of exposed individuals received a therapeutic PSG). The very low number of individuals undergoing therapeutic PSG is likely largely explained by the advent of auto-titrating PAP units, with data recording and download capabilities, which has substantially shifted airway pressure determination from lab to home. A sensitivity analysis by PAP or supplemental oxygen receipt was also not feasible because of small sample size (only 37/201 [18%] of our exposed individuals were recorded as having subsequently received PAP or supplemental oxygen) and this is as a result of known incomplete recording of these therapies in our health administrative databases. While we propose that our findings of better health outcomes in association with PSG testing are likely as a result of diagnosis and treatment of sleep breathing disorder, the institution of other cardio-pulmonary interventions/treatments as a consequence of PSG results may have also possibly contributed to observed improvements. It is possible that undergoing PSG testing is reflective of having a more thorough health care provider, and that the observed improved health outcomes in association with PSG are then as a consequence of receipt of more thorough overall medical care, rather than direct downstream consequences following PSG. However, our propensity score model included multiple markers of health care utilization, including number of outpatient physician visits in the preceding year, receipt of other types of investigations (including chest computed tomography, echocardiography and exercise oximetry testing) and receipt of multiple types of pharmacotherapies (including anti-fibrotic therapy, systemic corticosteroids, respiratory antibiotics, inhalers, smoking cessation drugs and cardiac medications), and exposed and controls were well-balanced on all these variables, thereby making it less likely that differences in overall medical care received explain our findings. Our health administrative databases also do not contain objective information relating to sleep breathing disorder diagnosis (e.g., apnea–hypopnea index, oxygen desaturation measures). Our findings also potentially do not apply to individuals with IPF under the age of 66 years old, who were excluded from our study.

Receipt of PSG was found to be associated with significantly lower rates of respiratory-related hospitalization and all-cause mortality among individuals with IPF. Furthermore, we observed that PSG testing was being infrequently performed in the IPF population. Our findings have potentially important implications for the management of IPF, a disease that is progressive and for which there are limited treatment options. While our study was limited by the use of PSG as a surrogate marker, our results raise the possibility that evaluation for and treatment of sleep breathing disorders is beneficial in IPF and that is management strategy is being suboptimally applied. Further research, particularly clinical trials, would be needed to exclude possible unresolved confounding and establish causation.