Article Text
PDF
Abstract
Objective To evaluate the synergistic effects created by fine particulate matter (PM2.5) and corticosteroid use on hospitalisation and mortality in older adults at high risk for cardiovascular thromboembolic events (CTEs).
Design and setting A retrospective cohort study using a US nationwide administrative healthcare claims database.
Participants A 50% random sample of participants with high-risk conditions for CTE from the 2008–2016 Medicare Fee-for-Service population.
Exposures Corticosteroid therapy and seasonal-average PM2.5.
Main outcome measures Incidences of myocardial infarction or acute coronary syndrome (MI/ACS), ischaemic stroke or transient ischaemic attack, heart failure (HF), venous thromboembolism, atrial fibrillation and all-cause mortality. We assessed additive interactions between PM2.5 and corticosteroids using estimates of the relative excess risk due to interaction (RERI) obtained using marginal structural models for causal inference.
Results Among the 1 936 786 individuals in the high CTE risk cohort (mean age 76.8, 40.0% male, 87.4% white), the mean PM2.5 exposure level was 8.3±2.4 µg/m3 and 37.7% had at least one prescription for a systemic corticosteroid during follow-up. For all outcomes, we observed increases in risk associated with corticosteroid use and with increasing PM2.5 exposure. PM2.5 demonstrated a non-linear relationship with some outcomes. We also observed evidence of an interaction existing between corticosteroid use and PM2.5 for some CTEs. For an increase in PM2.5 from 8 μg/m3 to 12 μg/m3 (a policy-relevant change), the RERI of corticosteroid use and PM2.5 was significant for HF (15.6%, 95% CI 4.0%, 27.3%). Increasing PM2.5 from 5 μg/m3 to 10 μg/m3 yielded significant RERIs for incidences of HF (32.4; 95% CI 14.9%, 49.9%) and MI/ACSs (29.8%; 95% CI 5.5%, 54.0%).
Conclusion PM2.5 and systemic corticosteroid use were independently associated with increases in CTE hospitalisations. We also found evidence of significant additive interactions between the two exposures for HF and MI/ACSs suggesting synergy between these two exposures.
- heart failure
- cardiac epidemiology
- clinical pharmacology
- statistics & research methods
- myocardial infarction
- stroke
Data availability statement
Data may be obtained from a third party and are not publicly available. For data privacy reasons, the Medicare data used in this study cannot be made publicly available, but interested parties can request access by applying through the US Centers for Medicare and Medicaid Services. The PM2.5 exposure data are publicly available at the following link: https://doi.org/10.7927/0rvr-4538. Area-level covariates used herein are also publicly available from the US Census Bureau website. Code for fitting the models, plotting estimates and reproducing study findings reported within this manuscript is available at: https://github.com/kevjosey/pm-interaction.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
- heart failure
- cardiac epidemiology
- clinical pharmacology
- statistics & research methods
- myocardial infarction
- stroke
STRENGTHS AND LIMITATIONS OF THIS STUDY
We conduct analyses using robust causal learning techniques and incorporate a large set of individual-level factors that are typically absent in environmental health analyses with large claims datasets.
We present statistics that evaluate the synergy between fine particulate matter and corticosteroid therapy on the additive scale, which is more relevant for assessing excess risks and informing policy.
Patient medical history prior to receiving Medicare benefits is unknowable within a Fee-for-Service claims database, which may lead to exclusion of some high-risk individuals from the cohort.
We censor participants after the earlier of the end of first corticosteroid regimen or the first incidence of the outcome of interest, which makes the analyses tractable but sacrifices some information that may be gleaned from the data.
Introduction
Climate change is ‘the single biggest health threat facing humanity’,1 and is expected to have a growing impact on human health through multiple pathways, including more frequent extreme weather events and worsening ambient air pollution.2 Air pollution is currently among the top five modifiable contributors to death and disease globally.3 The impacts of air pollution, specifically fine particulate matter (PM2.5), on the cardiovascular system are well-established. PM2.5 exposure has been linked to increased risk of stroke, myocardial infarction (MI), heart failure (HF), venous thromboembolism and other cardiovascular events.4–11 More than half of deaths attributable to air pollutants are due to cardiovascular thromboembolic events (CTEs).12 Epidemiological assessments in this area are also supported by cellular/toxicological experiments and by controlled animal/human studies, which both demonstrate the mechanisms by which PM2.5 may trigger acute events as well as prompt the chronic development of cardiovascular diseases.13
One of the most vulnerable populations, older adults, are at elevated risk for mortality and morbidity from PM2.5, particularly those with accessory comorbidities such as respiratory and cardiovascular diseases.14 Older adults are also at increased risk for CTE from certain medications taken to treat or prevent comorbidities.15 16 For example, systemic corticosteroids used for asthma/chronic obstructive pulmonary disease (COPD) exacerbations and to treat autoimmune diseases have direct vasoconstriction effects that inhibit fibrinolytic activity of the blood, leading to clinically recognised thrombogenicity.17 Further, systemic corticosteroids can cause sodium and fluid retention issues, leading to hypertension or HF exacerbations.18–20
Although the independent effects of air pollution and corticosteroids on CTE are well-known, no prior study has assessed the risk of both exposures simultaneously on CTE. Thus, it is unknown whether there is a synergy between these factors. Leveraging rich healthcare usage data on a large cohort of Medicare beneficiaries with comorbidities linked to residential PM2.5 concentrations, we examined whether simultaneously experiencing elevated PM2.5 concentrations and being exposed to corticosteroid therapies leads to an increased risk of CTE that is greater than the combination of these two effects independently. To our knowledge, this is the first study to examine interactions between a drug and air pollution exposure. We therefore provide a causal analytic framework that enables robust investigation of the contributing factors that explain individual-specific vulnerabilities to air pollution through the evaluation of additive interactions in survival models. Additionally, our analyses adjust for a large set of individual-level potential confounders that are typically unmeasured in environmental health analyses with large claims datasets, which lends added credibility to our findings.
Methods
Study population and cohort definition
The cohort used in this study has been previously described.14 Briefly, we used data from a 50% random sample of the 2008–2016 Medicare Part D-eligible Fee-for-Service beneficiary population and formed a cohort of individuals with conditions known to increase the risk of CTE. These high-risk conditions included pre-existing cardiovascular diseases, prior VTE, total joint arthroplasty and cancer. Any beneficiary who had an inpatient diagnosis/procedure at any position (primary or otherwise) for one or more of the above causes during a 1-year baseline period from their date of enrolment into the Medicare Fee-for-Service system was entered into the cohort at the end of the baseline period. This definition for high CTE risk has been shown to be highly predictive of future CTEs (see online supplemental table S1 for specific International Classification of Diseases (ICD-9/ICD-10) diagnosis codes used to define each high-risk condition).21
Supplemental material
Outcomes
We followed all participants until they developed one of the outcomes of interest, or until they experienced a censoring event—whichever occurred first. Outcomes of interest included hospitalisation for: (1) MI or acute coronary syndrome (MI/ACS); (2) ischaemic stroke or transient ischaemic attack (Stroke/TIA); (3) HF; (4) atrial fibrillation (Afib); (5) VTE or (6) death from any cause (see online supplemental table S2 for ICD-9/ICD-10 diagnosis and procedure codes identifying the outcomes). Non-administrative censoring events included death (when death was not the outcome under study), loss of eligibility for Medicare Part D and the participant’s moving to a ZIP-code without available PM2.5 exposure data. We also censored participants after the discontinuation of their first corticosteroid therapy, with a 30-day grace period.
PM2.5 exposures
Seasonal-average PM2.5 concentrations were derived from spatially and temporally aggregated predictions from a well-validated, high-resolution PM2.5 model.22 This model predicts PM2.5 concentrations at 1 km2 grids across the USA and consists of an ensemble of neural net and machine learning submodels trained on integrated high-resolution satellite, land use, emissions, ground monitoring and weather data. Daily gridded estimates were aggregated and linked to participants by residential ZIP-code, and then averaged within seasons. Figure 1 demonstrates the significant between-season variation in PM2.5 patterns in the USA, which motivated our choice to examine seasonal-average PM2.5 exposures rather than more traditional yearly average exposures as shorter-term increases in exposure may have harmful effects in this vulnerable population. We are particularly focused on contrasting outcomes under PM2.5 exposures of 12 μg/m3 versus 8 μg/m3, which are policy-relevant thresholds currently in review by the US Environmental Protection Agency (EPA; 12 μg/m3 being the current US limit for annual average PM2.5).23 We also compare outcomes for the contrast between PM2.5 levels of 10 μg/m3 versus 5 μg/m3 in a secondary analysis, informed by the WHO’s updated guidelines recommending an annual average limit of 5 μg/m3 (recently reduced from 10 μg/m3).24
Season-specific average PM2.5 measurements for every ZIP-code across the USA over the period 2008–2016. Note that the PM2.5 measurements are most severe in the southern states during the summer and in the upper-midwest states during winter.
Corticosteroid exposures
We used Medicare Part D drug dispensing data to identify systemic corticosteroid exposure. Systemic corticosteroids of interest included cortisone, hydrocortisone, prednisone, prednisolone, methylprednisolone, triamcinolone, dexamethasone and betamethasone. Initiation and duration of each corticosteroid were estimated based on the dispensing date, dispensing dose and days’ supply of the participants’ prescriptions. Because allowing for continuous follow-up was computationally infeasible, corticosteroid therapy status was updated quarterly until one of the study endpoints was achieved for each participant.
To ensure that individuals’ quarterly follow-up times aligned with key dates of corticosteroid usage, we constructed unique drug exposure panels for each cohort member. The anchor point of the drug exposure panels is the date of initiation of corticosteroid therapy during follow-up for users. In other words, for a participant who uses corticosteroids at some point during follow-up, the first day of their corticosteroid therapy always coincides with the first day of a quarter. We then constructed individual-specific panels spanning quarterly intervals extending backward in time to the participant’s index date and forward in time to the participant’s end date (see online supplemental figure S1, for example). For individuals who never use corticosteroids during follow-up, quarter start times coincide with changes in season.
Covariates
We identified individual-level sociodemographic characteristics, comorbidities and health services usage information derived from Medicare enrolment files and inpatient, outpatient, and drug dispensing data from files pertaining to Medicare Parts A, B and D, respectively. Using Medicare enrolment files, we extracted the following individual-level baseline variables: age, sex, race/ethnicity and Medicaid eligibility (a proxy for low-income status). Various pre-enrolment measurements were assessed based on diagnosis codes for inpatient and outpatient visits during each participant’s baseline period (see online supplemental table S3 for complete list of comorbidities). We also derived metrics of health services usage during the baseline period, including the number of hospitalisations, number of emergency department visits, number of outpatient visits and number of generic medications dispensed. We consider this collection of variables as time-invariant and treat them as potential confounders between corticosteroid use and CTE.
Additional temporal and neighbourhood-level features were also identified to enable further confounding adjustment, for both PM2.5-CTE and corticosteroid-CTE associations. These variables included season, year, region and PM2.5 from the prior four seasons as well as area-based measures of population density, proportion of residents living below the federal poverty line, proportion of housing units that are owner-occupied, median home value, median household income, proportion of residents identifying as Hispanic, proportion of residents identifying as Black and proportion of residents 25+ with at least a high school diploma that were linked to Medicare by ZIP-code of beneficiaries’ residence. These demographic and socioeconomic features were considered as time-varying covariates updated yearly. We also accounted for ZIP-code changes that occurred during follow-up and updated participants’ PM2.5 exposures and neighbourhood features accordingly.
Statistical analysis
We first described summary measures of the individual-level and neighbourhood-level characteristics and calculated the number of person-years at risk, number of events, and event rates per 1000 person-years for each of the six outcomes examined, both overall and stratified by corticosteroid status. We then fitted history-adjusted marginal structural Cox proportional hazard models, facilitated by inverse probability weight (IPW) estimates, to investigate both the independent and synergistic effects of PM2.5 and corticosteroid use on the CTE outcomes.25 26 Separate models were used to estimate the IPWs for each of the outcomes considered, and separate weighted Cox models were fit over the age-time scale.27 We included penalised spline components in the weighted Cox models to account for potential nonlinear effects of PM2.5, in addition to a main effect for corticosteroid use and an interaction between corticosteroid use and PM2.5 (the penalised spline representation). Given that our cohort comprises individuals with diverse diseases each potentially affecting thrombosis risk, we employed stratification in the Cox models based on disease indication. Particularly, we allowed for disease-specific baseline hazards for autoimmune diseases and COPD/asthma (see online supplemental table S4 for ICD-9/ICD-10 codes), as these two categories of diseases exhibit unique pathways for cardiovascular and thromboembolic events independent of corticosteroid therapy.28–33
The final IPW for a given participant and follow-up period was constructed as the product of three distinct IPWs accounting for different potential sources of bias: an inverse probability of treatment weight for each of the PM2.5 and corticosteroid exposures, to adjust for confounding, and an inverse probability of censoring weight to account for informative censoring. The ZIP-code and season-specific IPWs for PM2.5 were constructed by taking the inverse of estimated generalised propensity scores modelled using gradient boosting regression. The individual-level IPWs for the quarterly corticosteroid use indicators were obtained by inverting propensity scores estimated using gradient boosting classification. Additionally, over the same corticosteroid use quarters, we modelled the probability of censoring with gradient boosting classification to produce inverse probability of censoring weights. The three IPWs were stabilised by the marginal probabilities of treatment/censoring. Extreme weights were truncated at the 1st and 99th percentiles of the final IPW distribution (see the online supplemental file for additional details on the construction of the IPWs).
We report the HR estimates and 95% CIs associating PM2.5 with the five CTE outcomes and all-cause mortality (comparing average hazards evaluated at PM2.5 levels of 12 vs 8 µg/m3 and 10 vs 5 µg/m3) with corticosteroid status held fixed (both on and off treatment). We also provide HR estimates associating corticosteroid use with each outcome, with PM2.5 held fixed at 8 μg/m3. We assessed synergy between PM2.5 and corticosteroids by calculating the relative excess risk due to interaction (RERI)—a measure of interaction on the additive scale that can be interpreted as the relative increase in risk due to the combined effect of the two exposures versus the individual effects of the two exposures summed together (presented as a percentage).34–38 Cluster m-out-of-n bootstrap samples of the Cox model parameters were used to compute standard errors for the RERI which account for the correlation between units in the same ZIP-code while maintaining adequate computational efficiency.39 Additional details for estimating the RERI are provided in the online supplemental file.
All analyses were conducted using R V.4.2.0. Data cleaning was performed using SAS V.9.4.
Patient and public involvement
None.
Results
The cohort included 1 936 786 beneficiaries with a total of 4 629 432 person-years of follow-up. Average age at index date was 76.8 years, with 60.0% of cohort members female, 15.9% Medicaid eligible, 87.4% white and 8.2% Black (table 1). The average participant follow-up time was 2.4±2.3 years, although this figure along with the total person-years of follow-up fluctuates depending on the outcome being evaluated (online supplemental table S5). Among the Medicare beneficiaries evaluated, 37.7% had at least one prescription for corticosteroid therapy during follow-up. Participants who received corticosteroid therapy were slightly younger than those who never received corticosteroid therapy (75.7 vs 77.4 years old), were more likely to be white (89.8% vs 85.9%) and were less likely to be Medicaid eligible (14.0% vs 17.0%). Table 1 and online supplemental table S3 show summary statistics of several comorbidities included into the IPW models, stratified by disease indications listed in online supplemental table S4. Table 2 contains data on demographics and season-specific PM2.5 measurements over 329 544 ZIP-code years from 35 695 unique ZIP-codes. The average PM2.5 level was 8.3±2.4 µg/m3, average population density was 1425 people per square mile, and the overall poverty rate was 10.3%.
Characteristics of high-risk Medicare beneficiaries (total N=1 936 786)
Neighbourhood-level characteristics averaged over 329 727 ZIP-code years across 35 695 unique ZIP-codes
During an average follow-up of 2.4 years per person, we observed a total of 244 451 hospitalisation for HF, 118 754 hospitalisations for Afib, 101 611 hospitalisations for Stroke/TIA, 93 191 hospitalisations for MI/ACS, 41 635 hospitalisations for VTE and 491 445 deaths. The incidence rates per 1000 person-years were 57.1 for HF, 27.0 for Afib, 22.8 Stroke/TIA, 20.8 for MI/ACS, 9.1 for VTE and 106.2 for death (online supplemental table S5).
Corticosteroid use was associated with higher risks of CTE and death, with significant associations for all six outcomes examined. Holding PM2.5 fixed at 8 μg/m3, the HRs (95% CI) for corticosteroid use were 2.04 (1.94, 2.14) for MI/ACS, 1.51 (1.42, 1.61) for Stroke/TIA, 2.18 (2.11, 2.25) for HF, 3.39 (3.19, 3.61) for VTE, 2.25 (2.11, 2.40) for Afib and 2.64 (2.57, 2.72) for death.
Seasonal-average PM2.5 exposure was also significantly associated with an increased risk of each of the six CTE and mortality outcomes. Increasing the PM2.5 concentration from 8 μg/m3 to 12 μg/m3, in the absence of corticosteroid therapy, resulted in HRs (95% CI) of 1.244 (1.226, 1.263) for MI/ACS, 1.252 (1.234, 1.271) for Stroke/TIA, 1.336 (1.323, 1.349) for HF, 1.307 (1.278, 1.337) for VTE, 1.181 (1.163, 1.200) for Afib and 1.227 (1.218, 1.236) for death. Figure 2 contains the estimated HRs associated with increasing PM2.5 from 5 μg/m3 to 10 μg/m3 and the estimated PM2.5 HRs while receiving corticosteroid therapy.
HRs (and 95% CIs) for corticosteroid use and increasing PM2.5 while both on and off corticosteroids. We contrasted the effects of setting PM2.5 concentrations to 10 versus 5 µg/m3 and 12 versus 8 µg/m3 which were chosen based on WHO and Environmental Protection Agency guidelines.23,24
Evaluating the interactions between PM2.5 and corticosteroid use on the additive scale, we observed significant interactions (RERI (95% CI)) associated with increasing PM2.5 from 8 μg/m3 to 12 μg/m3 for HF (15.6% (4.0%, 27.3%)), with a borderline significance interaction detected for death (12.5 (−0.6%, 25.5%)). Increasing PM2.5 from 5 μg/m3 to 10 μg/m3 resulted in a significantly increased excess risk due to interaction (RERI (95% CI)) for HF (32.4% (14.9%, 49.9%)) and MI/ACS (29.8% (5.5%, 54.0%)). Figure 3 plots the RERI curves corresponding to various PM2.5 contrasts across the range of observed exposure levels for each outcome. For most outcomes, the increase in RERI is steepest when PM2.5 is less than 10 μg/m3, indicating more intense synergy between PM2.5 and corticosteroids even at PM2.5 concentrations below current US annual average PM2.5 standards.
Relative excess risk due to interaction (RERI) between PM2.5 and corticosteroid usage for each of the six outcomes (and their 95% CI bands), comparing a range of PM2.5 concentrations to reference values of 5 μg/m3 (salmon) and 8 μg/m3 (blue). Curves represent the change in RERI due to simultaneously initiating corticosteroid treatment and increasing PM2.5 exposure to any given level above the corresponding PM2.5 reference level. Note that the curves intersect zero at their respective reference levels, as there can be no excess risk increase due to the interaction without changing both exposures concomitantly. ACS, acute coronary syndrome; TIA, transient ischaemic attack.
An interesting result worth noting concerns the HRs associated with PM2.5 while receiving corticosteroid therapy. Observe that nearly every estimate of the HR in the third subplot of figure 2 is attenuated toward the null value of one relative to the second subplot in figure 2 examining the effects of PM2.5 while not receiving corticosteroid therapy. This implies that the multiplicative interaction between PM2.5 and corticosteroid use is negative. Thus, our results provide an example of the discordance that can occur between additive and multiplicative measures of synergy, which is well-established in the literature, and further demonstrates the caution one needs to consider when evaluating potential causal interactions.40
Discussion
In this study, we examined the interaction between seasonal-average PM2.5 exposure and corticosteroid use on the risk of CTE in a cohort of Medicare beneficiaries with high-risk conditions for CTE from the broadly generalisable set of Fee-for-Service enrollees. Using marginal structural models from the causal inference literature, which adjust for time-varying confounding attributable to several observed neighbourhood-level and individual-level covariates, we found that the escalation in risk for certain CTE outcomes during periods of simultaneous high PM2.5 exposure and corticosteroid use was larger than what would be expected from the independent effects of the two factors added together. In particular, we detected synergism between these two exposures for HF and MI/ACS.
Numerous studies have reported that older adults and those with comorbidities, particularly respiratory and cardiovascular disease,41 are at elevated risk for mortality and morbidity from air pollution. Older adults may be more vulnerable not only because of age and pre-existing diseases, but also because of the multiple medications they receive.42 Despite making up only 13% of the US population, older adults account for more than one-third of all prescriptions dispensed.43 44 Yet, current evidence on the health impacts of air pollution lacks consideration of additional factors to characterise individuals at risk.40 45 In particular, studies lack considerations for medication use, a prevalent risk factor that may further increase vulnerability in older adults. To our knowledge, our study provides the first epidemiological evidence of synergistic effects of air pollution and medication on CTE outcomes in older adults.
In addition, examining the independent effects of PM2.5 and corticosteroid use on CTE and mortality, we observed results that corroborate those already found in the current literature. PM2.5 has been significantly associated with increased risk of MI/ACS,8 Stroke/TIA,7 HF,46 VTE,14 Afib5 and all-cause mortality.47 Our results sometimes yielded associations larger in magnitude than those found in previous studies. This is unsurprising given that our cohort consists only of participants already at high risk for CTE.14 Likewise, corticosteroid use was strongly and significantly associated with increased risk of the five CTE outcomes and all-cause mortality in our study. The deleterious effects that corticosteroids can have on CTE outcomes have already been described in several other reports.19 20
There are several potential biological mechanisms explaining the synergistic interactions between prescription systemic corticosteroids and PM2.5 on CTE. First, both PM2.5 and glucocorticoids have been shown to induce hypercoagulable states in humans. As PM2.5 is small enough to translocate into the bloodstream, chronic PM2.5 exposure may increase coagulability indirectly through production of pro-oxidative and pro-inflammatory factors that can then induce production of coagulation factors and fibrinogen. Steroids may complement this thrombogenicity by stimulating Plasminogen Activator Inhibitor-1, which decreases dissolution of fibrinogen.48 Second, PM2.5 may also lead to atherosclerotic changes and autonomic cardiac dysfunction (ie, reduced heart rate variability), which in conjunction with adverse metabolic changes seen with systemic glucocorticoid use, can increase risk of cardiovascular disease-related outcomes. Third, both PM2.5 inhalation49 and glucocorticoids50 have been shown to have vasoconstrictive effects, which can increase blood pressure, risk of hypoxia in cardiac/brain tissue, and ultimately lead to MI or stroke. Some steroids also exhibit mineralocorticoid activity at higher doses which can lead to fluid retention and potassium efflux.51 In combination with PM2.5’s effects on autonomic dysfunction and modulation of vascular tone,52 this could potentially exacerbate HF or induce arrhythmias.
Our analysis is not without its limitations. First, using model-based PM2.5 aggregated to ZIP-codes carries the potential for attenuation created by exposure measurement error.53 However, even with such potential attenuation, we still obtained significant results. Second, comorbidities were captured and fixed at the index date and not allowed to vary over time. However, most comorbidities that we accounted for are chronic diseases that are rarely reversed. Third, we censored participants after their first corticosteroid therapy ended, and repeated corticosteroid exposure was not considered in the analyses. A recurrent events model might have been constructed to alleviate this issue; however, fitting marginal structural models in this design is both more time-intensive and new to the causal inference space. Moreover, our approach to consider the first course of exposure makes epidemiological sense given that repeated drug exposures are likely to be associated with worsening of the disease or comorbidities, which is difficult to correct for in a model of the exposure responses.54 Fourth, exposure to PM2.5 may contribute to the development of diseases that necessitate corticosteroid therapies, such that corticosteroids may mediate the overall impact of PM2.5. However, in this paper we did not specifically assess this role of corticosteroids as a potential mediator of PM2.5 effects. Instead, our focus was on examining the combined effects of PM2.5 and corticosteroid use. Finally, while we allowed for differing baseline risks of the outcomes for certain disease indications, we did not investigate the potential differential exposure effects experienced by distinct groups of patients using corticosteroids. Based on the findings of this initial analysis, further investigation of these heterogeneous effects is essential in future studies.
Conclusion
Using a cohort of nearly 2 million adults at high risk for CTE, we found evidence of a synergistic effect between seasonal PM2.5 exposure and corticosteroid use on several CTEs. We used advanced causal inference methods to control for potential confounding attributable to a large set of individual-level and neighbourhood-level covariates. We also observed strong independent impacts of PM2.5 and corticosteroids on each of the six outcomes examined. Our study demonstrates that certain combinations of medication and PM2.5 can work synergistically to impose increased health risks on older adults, even when PM2.5 concentrations fall below EPA standards. While our results should not discourage clinicians/older adults from prescribing/taking medications needed for treatment, they do shape our conceptual model of disease risk, which we believe should incorporate potential synergisms between individual-level and environmental-level risk factors. Our results also emphasise the need for stricter control of PM2.5 concentrations to help protect these vulnerable populations for whom corticosteroid medications are commonly indicated.
Data availability statement
Data may be obtained from a third party and are not publicly available. For data privacy reasons, the Medicare data used in this study cannot be made publicly available, but interested parties can request access by applying through the US Centers for Medicare and Medicaid Services. The PM2.5 exposure data are publicly available at the following link: https://doi.org/10.7927/0rvr-4538. Area-level covariates used herein are also publicly available from the US Census Bureau website. Code for fitting the models, plotting estimates and reproducing study findings reported within this manuscript is available at: https://github.com/kevjosey/pm-interaction.
Ethics statements
Patient consent for publication
Ethics approval
The study was approved by the Rutgers Institutional Review Board (ID Pro20170001685).
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors KJ, RN and SS conducted the analysis, interpreted the data and drafted the article. AV, BB, AP and DR were responsible for writing portions of the literature review and aided in drafting the introduction and discussion of the article, particularly the clinical consequences of the results. PG, AP and MR were responsible for preprocessing and collating the data, along with other data management and administrative duties. SS guided and supervised the research process. RN and SS provided funding support. All authors provided critical feedback on the study design and revised the manuscript. All authors approved the final draft of the manuscript and are accountable for all aspects of the work reported herein. SS is responsible for the overall content as the guarantor.
Funding This study was supported by National Institutes of Health grants R01AG060232, 1K01ES032458 and 5T32ES007142.
Map disclaimer The inclusion of any map (including the depiction of any boundaries therein), or of any geographic or locational reference, does not imply the expression of any opinion whatsoever on the part of BMJ concerning the legal status of any country, territory, jurisdiction or area or of its authorities. Any such expression remains solely that of the relevant source and is not endorsed by BMJ. Maps are provided without any warranty of any kind, either express or implied.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.