Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Jul 23.
Published in final edited form as: Ann Intern Med. 2018 Feb 13;168(6):396–404. doi: 10.7326/M17-1907

Opioid analgesic use and risk of invasive pneumococcal diseases: a nested case-control study

Andrew D Wiese *, Marie R Griffin *,†,, William Schaffner *,, C Michael Stein , Robert A Greevy ‡,§, Edward F Mitchel Jr *, Carlos G Grijalva *,
PMCID: PMC6647022  NIHMSID: NIHMS1030457  PMID: 29435555

Abstract

Background:

Although certain opioid analgesics have immunosuppressive properties and increase the risk of infections in animals, the clinical implications of prescription opioid use on infection risk among humans are unknown.

Objective:

Test the hypothesis that prescription opioid use is an independent risk factor for invasive pneumococcal disease (IPD).

Design:

Nested-case control study

Setting:

Tennessee Medicaid database linked to Medicare and Active Bacterial Core surveillance system databases (1995–2014)

Patients:

1,233 IPD cases ≥5 years matched to 24,399 controls on the diagnosis date, age, and county of residence.

Measurements:

Opioid use was measured using pharmacy prescription fills. IPD was defined by the isolation of S. pneumoniae from a normally sterile site. We compared the odds of current opioid use between cases and controls accounting for known IPD risk factors. Secondary analyses categorized opioid use by opioid characteristics, applied an IPD risk score to assure comparability between exposure groups and analyzed pneumonia and non-pneumonia IPD cases, separately.

Results:

IPD cases had a higher odds of being current opioid users compared with controls [adjusted odds ratio (aOR),1.62 (95% CI, 1.36 to 1.92)]. Associations were strongest for opioids that were long-acting [aOR,1.87 (CI, 1.24 to 2.82)], high-potency [aOR,1.72 (CI, 1.32 to 2.25)], and used in high doses [50–90mg/day aOR,1.71 (CI, 1.22 to 2.39) and ≥90mg/day aOR,1.75 (CI, 1.33 to 2.29)]. Results were consistent when accounting for the IPD risk score and when analyzing pneumonia and non-pneumonia IPD, separately.

Limitations:

Unmeasured confounding and measurement error, although sensitivity analyses suggested that neither was likely to affect results. Actual opioid use and other non-prescribing use (e.g., illicit opioid use) was not measured.

Conclusions:

Opioid use was associated with an increased risk of IPD and represents a novel risk factor for these diseases.

Primary Funding Source:

National Institutes of Health

INTRODUCTION

As opioid analgesic use has increased in the United States, the safety of prescription opioid use has come under further scrutiny.(14) Common safety concerns include the potential for opioid use disorders, overdose, and the development of serious adverse respiratory and cardiovascular events.(58) However, these known adverse effects only partially account for the excess morbidity and mortality observed among prescription opioid users.(911) There are also concerns about a potential excess of infections observed among prescription opioid users, but few studies have attempted to quantify the risk of infection among subjects using opioid analgesics.(1214)

Certain opioids have known immunosuppressive properties, and their use may increase the risk of infections.(15, 16) Animal and in–vitro experimental studies have demonstrated that some opioids disrupt lymphocyte and phagocyte proliferation, reduce innate immune cell activity, and inhibit cytokine expression and antibody production.(1618) In animal models, opioid-induced immune disruption also led to an increased susceptibility to bacterial infection, including infections caused by common human pathogens such as Streptococcus pneumoniae.(1921) However, the clinical implications of these observations for humans, including whether the risk differs by specific opioid properties or dose, remains unclear.

Invasive pneumococcal disease (IPD), caused by S. pneumoniae, includes serious illnesses such as bacteremia, meningitis and invasive pneumonia.(22) The case-fatality for IPD is high among adults (IPD pneumonia: 5–7%, IPD bacteremia: 20%, IPD meningitis: 22%), and thought to be even higher among older adults.(23) Diagnosis of IPD requires the isolation of S. pneumoniae from a normally sterile site.(23) Known risk factors for IPD include age (young children and older adults), decreased immune function, chronic high–risk medical conditions (e.g., lung, liver and kidney disease) and cigarette smoking.(22, 2426) Since IPD monitoring and prevention remains a public health priority, and opioid analgesic use represents a potentially novel and modifiable risk factor for serious infections including IPD, we sought to test the hypothesis that opioid analgesic use is an independent risk factor for laboratory-confirmed IPD.

METHODS

Data Sources

We conducted a nested case-control study among a retrospective cohort of persons enrolled in the Tennessee Medicaid (TennCare) program. TennCare, the managed Medicaid program in Tennessee, provides healthcare insurance to Tennessee residents who are Medicaid eligible. TennCare data provided information about enrollment, demographics, pharmacy use, healthcare encounters and comorbidities for each subject. These data were supplemented with State Vital Records information and hospital-based data from the Tennessee Hospital Discharge Data System. Pharmacy data were supplemented with Medicare Part D information for dual–eligible subjects. Laboratory-confirmed IPD cases were identified from the Tennessee Active Bacterial Core surveillance (ABCs) system. The ABCs system is funded by the Centers for Disease Control and Prevention to conduct active population and laboratory-based surveillance in 10 states for disease caused by selected pathogens of public health relevance, including IPD.(27) In Tennessee, Vanderbilt University Medical Center collaborates with the Tennessee Department of Health and the Tennessee Emerging Infections program to operate the ABCs system in 20 counties. The ABCs collects demographic, healthcare encounter, risk factor, and specimen/pathogen information for every detected case.

Institutional Review Board Approval

This study was approved by the Institutional Review Boards of Vanderbilt University and the Tennessee Department of Health, and the Bureau of TennCare.

Study Cohort

All TennCare enrollees with at least one filled study opioid prescription from 1995–2014 were identified (see Exposure) to exclude subjects with contraindications to opioids and those who may not be eligible to receive opioids. These subjects entered the study cohort on the earliest date (t0) when a study opioid prescription was filled, and the following criteria were met: >365 baseline days of continuous prior TennCare enrollment, age 5 years or older, documented access to pharmacy benefits, >1 healthcare encounter and no IPD identified during baseline, and free of non–study opioid (see Exposure) prescriptions during baseline or on t0. Subjects were also required to have ≥1 day of residence in a Tennessee county that reported to the ABCs system during the study period (see Case-Control Selection). Follow-up continued from t0 through the earliest of the following: end of the study (December 31, 2014), death, loss of enrollment, IPD, or first non-study opioid use. Subjects who ended follow-up due to loss of enrollment, IPD, or non-study opioid use were allowed to re-enter the cohort if they subsequently fulfilled all eligibility criteria.

Case-Control Selection

We used the ABCs system to identify laboratory-confirmed IPD among cohort members. IPD was defined by the isolation of S. pneumoniae from a normally sterile site (e.g., blood, cerebrospinal fluid).(24) The sample collection date was the index date for each case. We used incidence density sampling to randomly select up to 20 cohort members at risk but without laboratory-confirmed IPD (controls) per case. Controls were matched to cases on the index date, as well as age (individual years) and county of residence on that date. A subject could serve as a control for multiple cases and could later become an IPD case. Although nested case-control and cohort designs would provide the same conclusions, the nested case-control design was preferred for efficiency, especially with regards to exposure and covariates classification. Classifications were done relative to the index date, which simplifies the extensive computational challenge of tracking time-varying exposure and covariate information throughout a traditional long follow-up in a large cohort. Importantly, the nested case-control odds ratio provides an unbiased estimator of incidence rate ratios with negligible or no loss of precision.(28,29).

Exposure

The use of prescribed study opioids was the exposure of interest. Study opioid analgesics were prescribed oral and transdermal formulations. Non-study opioids included antitussive and antidiarrheal formulations (non-pain indications), injectable formulations for which timing of use and dose can be difficult to ascertain, and formulations used primarily for opioid use disorders (i.e. buprenorphine). Using pharmacy data, we defined four mutually exclusive exposure categories relative to the index date for cases and controls. Current users were subjects with a study opioid prescription overlapping the index date. To minimize exposure misclassification due to imperfect adherence or intermittent use, recent users were subjects whose most recent prescription ended 1–90 days before index date, and past users were subjects whose most recent prescription ended 91–182 days before the index date. Remote users included all other scenarios with no opioid prescription that ended within 182 days before the index date. New users were defined as a subset of current users whose prescription overlapping the index date was initiated after 182 days without an opioid prescription. Current opioid use, the main study exposure, was further classified according to the opioid duration of action (short/long-acting), potency (moderate/high), previously described immunosuppressive properties (immunosuppressive/non-immunosuppressive/unknown), and estimated daily dose in morphine milligram equivalents on the index date (<50mg, 50–90mg, >=90mg). Opioid characteristics were defined at the national drug classification code level of the opioid based on the previous literature and classifications used in earlier studies (Table 1).(7, 13, 14, 30) To avoid misclassification, current users of ≥1 opioid type were classified separately from those receiving only a single type.

Table 1.

Study Opioid Classifications*

Potency Duration of Action Previously described immunosuppressive properties MME Dose
Short–acting
Propoxyphene Medium Short–Acting Unknown 0.23
Codeine§ Medium Short–Acting Yes 0.15
Hydrocodone§ Medium Short–Acting No 1.0
Tramadol (immediate release) Medium Short–Acting No 0.10
Butalbital/codeine Medium Short–Acting Yes 0.15
Dihydrocodeine Medium Short–Acting Yes 0.25
Pentazocine Medium Short–Acting No 0.37
Tapentadol (immediate release) Medium Short–Acting Unknown 0.40
Short–Acting
Morphine sulfate High Short–Acting Yes 1.0
Codeine sulfate High Short–Acting Yes 0.15
Oxycodone High Short–Acting No 1.5
Hydromorphone (immediate release) High Short–Acting No 4.0
Meperidine hydrochloride High Short–Acting Unknown 0.1
Fentanyl (transmucosal)|| High Short–Acting Yes 125.0
Oxymorphone (immediate release) High Short–Acting No 3.0
Long–acting
Hydrocodone (extended release) High Long–Acting No 1.0
Levorphanol High Long–Acting Unknown 0.4
Tapentadol (extended release) High Long–Acting No 0.40
Tramadol (extended release) High Long–Acting No 0.1
Morphine sulfate (sustained release) High Long–Acting Yes 1.0
Oxycodone HCL controlled release High Long–Acting No 1.5
Methadone High Long–Acting Yes 3.0
Fentanyl (transdermal) High Long–Acting Yes 2.4
Oxymorphone (extended release) High Long–Acting No 3.0
Hydromorphone (extended release) High Long–Acting No 4.0
*

Opioid characteristics were defined based on previous literature and classifications used in earlier studies (7, 13, 14, 30)

Morphine milligram equivalent conversion per mg of opioid, with conversion factors based on classifications used in earlier studies (7, 13, 14, 30)

Alone or in combination

§

In combination

||

The conversion factor to milligram morphine equivalents for transmucosal fentanyl assumes that the measurement of opioid strength is measured as milligrams per oral dose, and it assumes 50% bioavailability of transmucosal fentanyl (e.g. 0.100 grams transmucosal fentanyl is equivalent to 12.5 of oral morphine) (7)

The conversion factor to milligram morphine equivalents for transdermal fentanyl assumes that the measurement of opioid strength is measured as micrograms per hours and assumes each patch remains in place for 3 days (e.g. 25 micrograms transdermal fentanyl/hour is equivalent to 60mg of oral daily morphine) (7)

Covariates

Relevant demographics, comorbidities (including IPD risk factors), conditions associated with pain, medication use and healthcare utilization were measured during the 365 days before the index date and considered as potential confounders. Demographics included sex and race. Other covariates, including healthcare resources use, were defined using diagnosis and procedure codes. Medications were identified using TennCare medication codes. Well-recognized risk factors for IPD, per the Advisory Committee on Immunization Practices, alcohol or substance use disorder, cardiovascular disease, serious hepatic and chronic lung disease, end-stage renal disease/hemodialysis, HIV, malignancy, immune disorders, diabetes, sickle-cell disease, and tobacco use.(25, 26) Other comorbidities included surrogate frailty markers (e.g., debility, pressure ulcers, impaired mobility).(31) Conditions associated with pain included abdominal, back, musculoskeletal, dental and neuropathic pain, as well as trauma/injury, headache, arthritis and pain not otherwise specified. Healthcare utilization included nursing home residence and the baseline number of hospitalizations, outpatient, and emergency department visits (Appendix Table 1). Per our selection criteria, only individuals with full benefits that demonstrated active use of those services were included. Thus, indicators for each study covariate were based on the presence of specific conditions and medication use. Lack of evidence meant the individual had no history of that condition or medication use, and so this information was not considered missing.

Statistical Analysis

We compared the odds of being a current opioid user versus a remote user between IPD cases and controls. Multivariable conditional logistic regression was used to calculate adjusted odds ratios (aOR) and 95% confidence intervals (CI) accounting for the matching design and adjusting for all well-recognized risk factors for IPD. To assess model fit and fulfillment of assumptions, we conducted standard regression diagnostics for conditional logistic regression (Section 3-Appendix).(32) Planned secondary analyses stratified current opioid use in separate models by duration of action, potency, previously described immunosuppressive properties, and estimated daily dose.(7, 13, 14, 30) Since opioid-related respiratory depression may facilitate aspiration and the development of pneumonia, we also assessed IPD associated with pneumonia separately from non-pneumonia IPD outcomes.

As a complementary method of assuring that current and remote opioid users were similar concerning IPD risk, we conducted a separate planned analysis by calculating an IPD risk score that included all study covariates, excluding the well-recognized risk factors for IPD (Appendix Table 2). Analogous to propensity scores for cohort studies, disease risk scores provide an efficient strategy to account for potential differences in the risk of IPD between exposure groups in case-control designs, especially when the number of covariates is large, the exposure consists of multiple categories and the number of cases is limited.(3335) The IPD risk score was calculated among all non-current opioid users using a logistic regression model with IPD as the outcome and included 103 covariates assessed in the 365 days preceding the index date. The coefficients from this logistic regression model were used to calculate the predicted probability of IPD for each subject in the entire study population independent of opioid exposure and the presence of IPD risk factors. We incorporated the IPD risk score, categorized as deciles of predictive probabilities, together with the IPD risk factors into the conditional logistic regression model for opioid use and IPD.(33)

Since some opioid use may be prescribed for the initial clinical manifestations of IPD (e.g., chest pain associated with pneumonia), a planned sensitivity analysis excluded new users that initiated current opioid use within 4 days (inclusive) of the index date to address possible protopathic bias. Our primary analysis accounted for the use of the pneumococcal polysaccharide vaccine in the 365 days before the index date. Since polysaccharide vaccine protects against IPD for at least 5 years,(25) we examined pneumococcal vaccination history among cases and controls with >5 years of continuous enrollment preceding their index date. Since pneumococcal conjugate vaccines (PCVs) also provide long-term protection against IPD, we repeated our main analysis excluding data from 2012–2014, when PCV was recommended for use among adults.(25, 26) Finally, we assessed the sensitivity of our estimates to the impact of a potential unmeasured confounder.(36) All analyses were performed in Stata-IC, version 15.1 (College Station, TX).

RESULTS

Study population

Among the retrospective cohort of TennCare enrollees who fulfilled all selection criteria (n=221,096) [Appendix Figure 1], we identified 1,233 laboratory-confirmed IPD cases [73.9% (n=911) were invasive pneumonia] and 24,399 matched controls. Cases had a lower percentage of females and a higher prevalence of risk factors for IPD compared to controls, including cardiovascular and chronic lung disease, HIV, malignancy, and smoking. In addition, 25.2% of IPD cases were current users of opioids on the index date compared with 14.4% of controls (Table 2). Among current opioid users, a higher percentage of cases used long-acting and high potency opioids, and higher daily doses compared to controls (Table 3). Comparing characteristics between exposure groups, current opioid users had a higher prevalence of risk factors for IPD than remote users, including age ≥40 years, cardiovascular and chronic lung disease, malignancy, diabetes, smoking and higher prior levels of healthcare utilization. Vaccination with pneumococcal polysaccharide vaccine in the 365 days before the index date was more common among current users than remote users (Table 4).

Table 2.

Characteristics of IPD cases and matched controls, Tennessee Medicaid enrollees (1995–2014)

Characteristic IPD Cases (n=1,233) Controls (n=24,399)
Female sex – no. (%) 732 (59.4%) 16,731 (68.6%)
Race – no. (%)
 White 571 (46.3%) 11,378 (46.6%)
 Black 534 (43.3%) 10,378 (42.5%)
 Other 128 (10.4%) 2,643 (10.8%)
Age Category*, years – no. (%)
 <18 44 (3.6%) 880 (3.6%)
 18–39 262 (21.2%) 5,240 (21.5%)
 40–64 636 (51.6%) 12,688 (52.0%)
 65–74 159 (12.9%) 3,158 (12.9%)
 ≥75 132 (10.7%) 2,433 (10.0%)
Residence* – Type of County – no. (%)
 Non–Metropolitan 177 (14.4%) 3,336 (13.7%)
 Metropolitan 1,056 (85.6%) 21,063 (86.3%)
Comorbidities – no. (%)
 Alcohol and substance use disorder 165 (13.4%) 966 (4.0%)
 Cardiovascular disease 266 (21.6%) 2,761 (11.3%)
 Serious hepatic disease 68 (5.5%) 213 (0.9%)
 Chronic lung disease 354 (28.7%) 2,979 (12.2%)
 End–stage renal disease 69 (5.6%) 388 (1.6%)
 HIV 161 (13.1%) 296 (1.2%)
 Malignancy 144 (11.7%) 948 (3.9%)
 Immune disorder/transplant 22 (1.8%) 89 (0.4%)
 Diabetes 253 (20.5%) 3,928 (16.1%)
 Sickle cell disease 13 (1.1%) 37 (0.2%)
 Smoking and smoking-related diagnoses 272 (22.1%) 2,063 (8.5%)
Healthcare Utilization – no. (%)
 Pneumococcal polysaccharide vaccination 43 (3.5%) 561 (2.3%)
 Outpatient clinic visits
   0–4 565 (45.8%) 13,333 (54.6%)
   5–9 284 (23.0%) 5,958 (24.4%)
   10–19 276 (22.4%) 4,095 (16.8%)
   ≥ 20 108 (8.8%) 1,013 (4.2%)
 ED visits
   0 361 (29.3%) 11,973 (49.1%)
   1–2 489 (39.7%) 8,728 (35.8%)
   3–4 205 (16.6%) 2,270 (9.3%)
   ≥ 5 178 (14.4%) 1,428 (5.9%)
 Hospitalizations
   0 578 (46.9%) 18,128 (74.3%)
   1 297 (24.1%) 3,880 (15.9%)
   2 157 (12.7%) 1,302 (5.3%)
   ≥ 3 201 (16.3%) 1,089 (4.5%)
 Recent nursing home stay - past 30 days 69 (5.6%) 1,132 (4.6%)
Opioid Use§ – no. (%)
 Remote users 492 (39.9%) 12,690 (52.0%)
 Past users 118 (9.6%) 2,705 (11.1%)
 Recent users 312 (25.3%) 5,483 (22.5%)
 Current users 311 (25.2%) 3,521 (14.4%)
IPD Syndrome – no. (%)
 Invasive pneumonia 911 (73.9%) n/a
 Other IPD syndromes|| 322 (26.1%) n/a
*

Controls were matched to cases on individual year of age, county of residence and eligibility on the index date (i.e. controls had to be eligible retrospective cohort members on the index date for the case)

Metropolitan counties were defined as those under Active Bacterial Core surveillance with at least one city with a population >100,000 according to 2015 Census estimates (Davidson, Hamilton, Knox, Rutherford and Shelby)

Comorbidities and healthcare utilization patterns were assessed in the 365–day period preceding the index date for cases and controls (with the exception of “Recent Nursing Home Stay”)

§

Opioid use (current, recent, past and remote) was assessed relative to the index date for cases and matched controls

||

Other invasive pneumococcal disease syndromes included meningitis, primary bacteremia and bacteremia secondary to other conditions (e.g. cellulitis)

Table 3.

Distribution of Opioid Characteristics in Current Opioid Users among IPD cases and matched controls, Tennessee Medicaid enrollees (1995–2014)

Characteristic* IPD Cases (n=311) Controls (n=3,521)
New Users – no (%) 21 (6.8%) 186 (5.3%)
Duration of Opioid Action – no. (%)
 Short–Acting (SA) 231 (74.3%) 2,869 (81.5%)
 Long–Acting (LA) 37 (11.9%) 256 (7.3%)
 Combination SA/LA 43 (13.8%) 396 (11.2%)
Previously described Immunosuppressive Properties– no. (%)
 Unknown 35 (11.3%) 408 (11.6%)
 Non–Immunosuppressive (NIS) 200 (64.3%) 2,446 (69.5%)
 Immunosuppressive (IS) 44 (14.1%) 368 (10.5%)
 Combination Unknown/NIS/IS 32 (10.3%) 299 (8.5%)
Potency of Opioid – no. (%)
 Medium 182 (58.5%) 2,479 (70.4%)
 High 100 (32.2%) 813 (23.1%)
 Combination Medium/High 29 (9.3%) 229 (6.5%)
Opioid Dose§ – no. (%)
 <50mg 170 (54.7%) 2,220 (63.1%)
 50–90mg 51 (16.4%) 509 (14.5%)
 ≥90mg 90 (28.9%) 792 (22.5%)
*

All opioid characteristics were defined a priori (Table 1)

Controls were matched to cases on individual year of age, county of residence and eligibility on the index date (i.e. controls had to be eligible retrospective cohort members on the index date for the case)

Each opioid was categorized a priori as potentially immunosuppressive, non–immunosuppressive and unknown based on existing literature (Table 1)

§

Categories of opioid dose were defined a priori according to morphine milligram equivalents per day based on categories outlined in the U.S. Centers for Disease Control and Prevention chronic pain opioid prescribing guidelines that recommend careful assessment of opioid prescriptions 50–90 morphine-milligram equivalents and ≥90 morphine milligram equivalents per day(7)

Table 4.

Characteristics of current opioid users compared to remote opioid users, Tennessee Medicaid enrollees (1995–2014)*

Characteristic Current Opioid Users (n=3,832) Remote Opioid Users (n=13,182)
Female sex – no. (%) 2,555 (66.7%) 8,757 (66.4%)
Race – no. (%)
 White 2,431 (63.4%) 5,385 (40.9%)
 Black 951 (24.8%) 6,380 (48.4%)
 Other 450 (11.7%) 1,417 (10.7%)
Age Category, years
 <18 17 (0.4%) 621 (4.7%)
 18–39 361 (9.4%) 3,049 (23.1%)
 40–64 2,334 (60.9%) 6,698 (50.8%)
 65–74 642 (16.8%) 1,552 (11.8%)
 ≥75 478 (12.5%) 1,262 (9.6%)
Residence – Type of County
 Non–Metropolitan 829 (21.6%) 1,542 (11.7%)
 Metropolitan 3,003 (78.4%) 11,640 (88.3%)
Comorbidities§ – no. (%)
 Alcohol/substance abuse 262 (6.8%) 418 (3.2%)
 Cardiovascular Disease 724 (18.9%) 1,110 (8.4%)
 Serious hepatic disease 69 (1.8%) 81 (0.6%)
 Chronic Lung Disease 908 (23.7%) 1,167 (8.9%)
 End–stage renal disease 91 (2.4%) 161 (1.2%)
 HIV 93 (2.4%) 198 (1.5%)
 Malignancy 293 (7.6%) 340 (2.6%)
 Immune Disorder/Transplant 27 (0.7%) 36 (0.3%)
 Diabetes 908 (23.7%) 1,696 (12.9%)
 Sickle Cell Disease 12 (0.3%) 14 (0.1%)
 Smoking and smoking-related diagnosis 665 (17.4%) 703 (5.3%)
Healthcare Utilization§ – no. (%)
 Pneumococcal polysaccharide vaccination 147 (3.8%) 234 (1.8%)
 Outpatient Visits
   0–4 1,111 (29.0%) 8,781 (66.6%)
   5–9 917 (23.9%) 2,895 (22.0%)
   10–19 1,332 (34.8%) 1,276 (9.7%)
   ≥ 20 472 (12.3%) 230 (1.7%)
 ED Visits
   0 1,603 (41.8%) 7,750 (58.8%)
   1–2 1,393 (36.4%) 4,217 (32.0%)
   3–4 466 (12.2%) 847 (6.4%)
   ≥ 5 370 (9.7%) 368 (2.8%)
 Hospitalizations
   0 2,429 (63.4%) 10,682 (81.0%)
   1 749 (19.5%) 1,631 (12.4%)
   2 308 (8.0%) 515 (3.9%)
   ≥ 3 346 (9.0%) 354 (2.7%)
 Recent Nursing Home Stay – Past 30 days 299 (7.8%) 503 (3.8%)
IPD Case Status – no. (%)
 Control 3,521 (91.9%) 12,690 (96.3%)
 Case 311 (8.1%) 492 (3.7%)
*

Counts within rows of each variable type will not total to the full study population, as totals for past users (n=2,823) and recent users (n=5,795) are not included in this table

Opioid use (current, recent, past and remote) was assessed relative to the index date for cases and matched controls

Metropolitan counties were defined as those under Active Bacterial Core surveillance with at least one city with a population >100,000 according to 2015 Census estimates (Davidson, Hamilton, Knox, Rutherford and Shelby)

§

Comorbidities and healthcare utilization patterns were assessed in the 365–day period preceding the index date for cases and controls (with the exception of “Recent Nursing Home Stay”)

Opioid use and risk of IPD

Current use of opioids was significantly associated with IPD compared with remote opioid use in the multivariable conditional logistic regression model, which adjusted for well-known IPD risk factors [aOR,1.62 (95% CI,1.36 to 1.92)] (Table 5 and Appendix Table 3). When current use was classified based on opioid characteristics, current use of both long-acting [aOR,1.87 (CI,1.24 to 2.82)] and short-acting opioids [aOR,1.58 (CI,1.32 to 1.90)] was associated with IPD compared with remote use. The association was demonstrated across all daily opioid dose categories, with the highest aORs observed at morphine milligram equivalent doses ≥50mg [50–90mg aOR,1.71 (CI,1.22 to 2.39) and ≥90mg aOR,1.75 (CI,1.33 to 2.29)]. Additionally, the strongest associations were observed for the use of high-potency opioids and opioids with previously described immunosuppressive properties, although the confidence intervals were overlapping across each category (Table 5 and Appendix Table 4). Importantly, the association between current opioid use and IPD was demonstrated for both pneumonia IPD [aOR,1.54 (CI,1.26 to 1.88)] and non-pneumonia IPD [aOR,1.94 (CI,1.36 to 2.77)]. In the subset of current users identified as new users (n=21 cases), the aOR was higher compared with remote users [aOR,2.44 (CI,1.49 to 4.00)] (Table 5).

Table 5.

Crude and Adjusted Odds Ratios (aOR) for Laboratory–Confirmed Invasive Pneumococcal Disease by Opioid Use Type among Tennessee Medicaid Enrollees (1995–2014) [n=25,362]

Exposure* Cases Crude Odds Ratio (95% CI) Adjusted Odds Ratio (95% CI)
Recency of Opioid Use
 Remote Users 492 1.00 (reference) 1.00 (reference)
 Past Users 118 1.13 (0.92 to 1.39) 0.87 (0.70 to 1.08)
 Recent Users 312 1.50 (1.29 to 1.73) 1.03 (0.87 to 1.21)
 Current Users 311 2.47 (2.11 to 2.89) 1.62 (1.36 to 1.92)
   New Users 21 3.01 (1.90 to 4.78) 2.44 (1.49 to 4.00)
Duration of Opioid Action
 Remote Users 492 1.00 (reference) 1.00 (reference)
 Short–Acting (SA) Opioid Users 231 2.24 (1.89 to 2.66) 1.58 (1.32 to 1.90)
 Long–Acting (LA) Opioid Users 37 3.92 (2.73 to 5.61) 1.87 (1.24 to 2.82)
 Combination SA/LA Opioid Users 43 3.15 (2.25 to 4.42) 1.64 (1.12 to 2.38)
Previously described Immunosuppressive Properties
 Remote Users 492 1.00 (reference) 1.00 (reference)
 Unknown 35 2.26 (1.58 to 3.25) 1.79 (1.22 to 2.63)
 Non–Immunosuppressive (NIS) 200 2.31 (1.93 to 2.77) 1.55 (1.27 to 1.88)
 Immunosuppressive (IS) 44 3.23 (2.33 to 4.48) 1.74 (1.20 to 2.53)
 Combination Unknown/NIS/IS 32 3.07 (2.09 to 4.50) 1.72 (1.12 to 2.63)
Potency of Opioid
 Remote Users 492 1.00 (reference) 1.00 (reference)
 Medium 182 2.04 (1.70 to 2.45) 1.52 (1.25 to 1.85)
 High 100 3.50 (2.77 to 4.43) 1.72 (1.32 to 2.25)
 Combination Medium/High 29 3.62 (2.42 to 5.44) 2.20 (1.40 to 3.46)
Dose of Opioid§
 Remote Users 492 1.00 (reference) 1.00 (reference)
 <50mg 170 2.13 (1.77 to 2.58) 1.54 (1.26 to 1.88)
 50–90mg 51 2.82 (2.07 to 3.83) 1.71 (1.22 to 2.39)
 ≥90mg 90 3.19 (2.50 to 4.06) 1.75 (1.33 to 2.29)
*

Each set of opioid characteristics (recency, duration of action, immunosuppression, potency and dose) were examined using a separate conditional logistic regression model including the same covariate sets (Appendix Table 4)

Adjusted odds ratio are derived from the full model including sex, race, alcohol/substance use disorder, cardiovascular disease, serious hepatic disease, chronic lung disease, hemodialysis, human-immunodeficiency virus, cancer, immune disorders, diabetes, sickle cell disease, smoking, nursing home residency, pneumococcal polysaccharide vaccination and numbers of healthcare encounters (hospitalizations, emergency department and outpatient) taking into account the study design where controls were matched to cases on individual year of age, county of residence and eligibility on the index date (i.e. controls had to be eligible retrospective cohort members on the index date for the case)

Subset of current users

§

Morphine milligram equivalent per day

The IPD risk score was calculated among non-current users [n=21,800 and 922 IPD cases (Appendix Table 2)]. Including the IPD risk score in the model yielded results that were very similar [aOR,1.61 (CI,1.35 to 1.91)] (Appendix Table 5) to the main analysis findings.

Sensitivity analyses

In the planned sensitivity analysis that excluded cases and controls with an index date in the first 4 days of new use, the aOR for current users was relatively unchanged from the main analysis [aOR,1.56 (CI,1.31 to 1.86)]. The aOR comparing the subset of new users to remote users was reduced but had limited precision [aOR,1.51 (CI,0.66 to 3.45)].

In a quantitative analysis to determine the sensitivity of our findings to a potential unmeasured confounder, we estimated that an unknown confounder would need to be an independent, strong risk factor for IPD with an odds ratio of 2 or higher, and need to have an absolute difference in prevalence of >35% between current opioid users and remote users to explain the observed lower confidence interval bound of the aOR (1.36) from our primary analysis (Appendix Figure 2). At lower absolute differences in prevalence (25%, 15%, and 10%), the unknown confounder would need to be a stronger, independent risk factor for IPD (odds ratios of 2.5, 3.5 and 5.5, respectively). Of note, none of the covariates in the study (including all IPD risk factors and pneumococcal vaccination history) met such requirements for the absolute difference in prevalence and the strength of the independent association (Table 4 and Appendix Table 3). Therefore, weaker confounders and those with lower exposure prevalence differences could partially attenuate the observed association, but not fully account for it.

In the examination of pneumococcal polysaccharide vaccination history among cases and controls with >5 years of continuous enrollment before the index date, vaccination was higher among cases than controls (14.9% vs. 10.1%, n=18,354) and among current opioid users compared to remote opioid users (16.4% vs. 8.5%). Therefore, due to the protective effect of polysaccharide vaccination, differences in polysaccharide vaccination observed in the study population history could not explain our findings and our estimates may be conservative. Similarly, excluding individuals with an index date in years when PCVs were recommended for adults yielded results similar to the main findings [n=23,065; aOR,1.64 (CI,1.36 to 1.97)].

DISCUSSION

We report a strong association between use of prescribed opioids and the risk of laboratory-confirmed IPD. The association was strongest for current users of long-acting, high-potency, previously described as immunosuppressive and high-dose opioid formulations, and was consistent across clinical syndromes of IPD.

The immunosuppressive properties of certain opioid analgesics, including morphine and fentanyl, have been well established.(15, 16) In animal models, exposure to certain opioids increased the risk of infections due to common pathogens, including S. pneumoniae.(19, 20) Among humans, opioid use has been previously linked to an increased risk of infection among hospitalized surgical, burn and cancer patients.(3739) Two previous studies have also reported an association between outpatient prescription opioid use and the risk of serious infection in specific high-risk groups. One study, restricted to community-dwelling older adults enrolled in a private health insurance system, reported that patients with pneumonia had a 39% increased odds of opioid exposure compared to controls.(13) In another study of patients with rheumatoid arthritis, there was a 38% increased frequency of hospitalization for serious infection during periods of current opioid use compared to non-opioid use, and results were consistent across pneumonia and serious non-pneumonia infections.(14) In both previous studies, and consistent with our current findings, the occurrence of serious infections was highest during periods of exposure to long-acting opioids, high opioid doses, and opioids previously described as immunosuppressive.(13, 14)

A unique strength of this study was the use of ABCs data to identify laboratory-confirmed IPD cases. We minimized misclassification by using only laboratory-confirmed outcomes. The specificity of laboratory-confirmed IPD is very high, supporting its use for assessment of relative measures of association. Furthermore, IPD is a prototypical community-acquired infection, and thus less affected by other factors (e.g. recent hospitalization, IV drug use) that may impact assessments of serious infections as a whole.(26) In context of the previous literature, our assessment of IPD complements previous studies and suggests that opioid analgesic use increases the risk of serious infections among humans.(13, 14)

An important limitation is that opioid use was based on pharmacy prescription fills, but actual use was not observed. We attempted to minimize misclassification of the exposure by defining recent and past use categories to ensure that the current use category represented periods with the highest likelihood of opioid use. Although we accounted for evidence of alcohol/substance use disorders in the analysis, we could not assess illicit opioid use. Another limitation was the inability to make direct comparisons across opioid types while accounting for duration of action, potency, and dose of each opioid. Although we observed the strongest associations for long-acting, high-potency, immunosuppressive and high-dose opioids, laboratory-confirmed IPD was relatively rare, and we were underpowered to account for these factors simultaneously and to make direct comparisons among individual opioids. Because there are differences in the bioavailability, half-life, and amount of active metabolites among opioids, we would expect that the association between opioid use and serious infections might vary across opioids. Future studies will be important to characterize the role of individual opioids and inform prescribers and patients regarding appropriate opioid selection.

Our analyses accounted for a substantial number of relevant covariates. However, we cannot rule out the possibility of residual confounding. We estimated that a potential unmeasured confounder would need to fulfill two criteria: be a very strong risk factor for IPD and have a substantial distribution imbalance between exposure groups, to explain our findings. Nevertheless, the consistency of results in the more comprehensive IPD risk score analysis and the primary analysis should reduce concerns about residual confounding. Although our main analysis directly accounted for pneumococcal polysaccharide vaccination history during the year preceding the index date, an extended assessment also examined the history of vaccination during the 5 years preceding the index date and considered this as a potentially unmeasured confounder. Since pneumococcal vaccination was more common among current than remote opioid users, accounting for the protective effect of this factor would result in a stronger association between opioid use and IPD. Similarly, we demonstrated that the availability of PCVs for adults starting in 2012 had no impact on our findings. Although our study covered several years, each IPD case was matched on the index date to eligible at-risk controls, so by design, comparisons accounted for changes in prescribing practices and disease incidence throughout the study period. This addressed concerns about the indirect protection derived from vaccination of infants with PCVs. Finally, since the study population consisted only of TennCare enrollees, the results may not be generalizable to other populations.

In conclusion, we found that current opioid use was strongly and consistently associated with the risk of IPD and that the association was strongest for the use of long-acting and high potency formulations, opioids previously described as immunosuppressive, and high dose opioids. Our study findings complement the experimental evidence from animal models and initial studies among humans and indicate that prescription opioid use is an independent, novel risk factor for IPD. These findings should be considered when formulating IPD prevention recommendations, including vaccination. Furthermore, this previously unrecognized association between opioid use and IPD highlights the need for judicious use of opioid analgesics considering both the benefits and risks of these medications. As the strongest associations were observed for opioids with certain characteristics, these findings should be considered during opioid analgesics selection for pain management.

Supplementary Material

Appendix Figure 1
Appendix Figure 2
Appendix Figure 3
Appendix Tables
Appendix Text

Acknowledgement

We are indebted to the Tennessee Bureau of TennCare of the Department of Finance and Administration, which provided data for the study. We are also indebted to the Tennessee Department of Health for providing data for the study. The corresponding author affirms that he has listed everyone who contributed significantly to the work.

Role of the funding source

The study was supported by the National Institutes of Health (NIH) - National Institute on Aging, through grants R03-AG042981, R01-AG043471 and TL1TR000447. The NIH had no role in the design, data collection, analysis, or interpretation of the study or in the decision to approve publication of the finished manuscript.

Footnotes

Reproducible Research Statement: Study protocol and statistical code is available from Carlos Grijalva (carlos.grijalva@vumc.org). The data is not available.

This is the prepublication, author-produced version of a manuscript accepted for publication in Annals of Internal Medicine. This version does not include post-acceptance editing and formatting. The American College of Physicians, the publisher of Annals of Internal Medicine, is not responsible for the content or presentation of the author-produced accepted version of the manuscript or any version that a third party derives from it. Readers who wish to access the definitive published version of this manuscript and any ancillary material related to this manuscript (e.g., correspondence, corrections, editorials, linked articles) should go to Annals.org or to the print issue in which the article appears. Those who cite this manuscript should cite the published version, as it is the official version of record.

REFERENCES

  • 1.Okie S A flood of opioids, a rising tide of deaths. N Engl J Med 2010;363(21):1981–5. [DOI] [PubMed] [Google Scholar]
  • 2.Dart RC, Surratt HL, Cicero TJ, Parrino MW, Severtson SG, Bucher-Bartelson B, et al. Trends in opioid analgesic abuse and mortality in the United States. N Engl J Med 2015;372(3):241–8. [DOI] [PubMed] [Google Scholar]
  • 3.Berterame S, Erthal J, Thomas J, Fellner S, Vosse B, Clare P, et al. Use of and barriers to access to opioid analgesics: a worldwide, regional, and national study. Lancet 2016;387(10028):1644–56. [DOI] [PubMed] [Google Scholar]
  • 4.Chou R, Turner JA, Devine EB, Hansen RN, Sullivan SD, Blazina I, et al. The effectiveness and risks of long-term opioid therapy for chronic pain: a systematic review for a National Institutes of Health Pathways to Prevention Workshop. Ann Intern Med 2015;162(4):276–86. [DOI] [PubMed] [Google Scholar]
  • 5.Solomon DH, Rassen JA, Glynn RJ, Garneau K, Levin R, Lee J, et al. The comparative safety of opioids for nonmalignant pain in older adults. Arch Intern Med 2010;170(22):1979–86. [DOI] [PubMed] [Google Scholar]
  • 6.McCann DJ, Skolnick P. Trends in opioid analgesic abuse and mortality in the United States. N Engl J Med 2015;372(16):1572–3. [DOI] [PubMed] [Google Scholar]
  • 7.Dowell D, Haegerich TM, Chou R. CDC Guideline for Prescribing Opioids for Chronic Pain - United States, 2016. MMWR Recomm Rep 2016;65(1):1–49. [DOI] [PubMed] [Google Scholar]
  • 8.Larochelle MR, Liebschutz JM, Zhang F, Ross-Degnan D, Wharam JF. Opioid Prescribing After Nonfatal Overdose and Association With Repeated Overdose: A Cohort Study. Ann Intern Med 2016;164(1):1–9. [DOI] [PubMed] [Google Scholar]
  • 9.Ray WA, Chung CP, Murray KT, Hall K, Stein CM. Prescription of Long-Acting Opioids and Mortality in Patients With Chronic Noncancer Pain. Jama 2016;315(22):2415–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Ekstrom MP, Bornefalk-Hermansson A, Abernethy AP, Currow DC. Safety of benzodiazepines and opioids in very severe respiratory disease: national prospective study. Bmj 2014;348:g445. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Vozoris NT, Wang X, Fischer HD, Gershon AS, Bell CM, Gill SS, et al. Incident opioid drug use among older adults with chronic obstructive pulmonary disease: a population-based cohort study. Br J Clin Pharmacol 2016;81(1):161–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Veldhuizen S, Callaghan RC. Cause-specific mortality among people previously hospitalized with opioid-related conditions: a retrospective cohort study. Ann Epidemiol 2014;24(8):620–4. [DOI] [PubMed] [Google Scholar]
  • 13.Dublin S, Walker RL, Jackson ML, Nelson JC, Weiss NS, Von Korff M, et al. Use of opioids or benzodiazepines and risk of pneumonia in older adults: a population-based case-control study. J Am Geriatr Soc 2011;59(10):1899–907. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Wiese AD, Griffin MR, Stein CM, Mitchel EF Jr., Grijalva CG. Opioid Analgesics and the Risk of Serious Infections Among Patients With Rheumatoid Arthritis: A Self-Controlled Case Series Study. Arthritis Rheumatol 2016;68(2):323–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Pergolizzi J, Boger RH, Budd K, Dahan A, Erdine S, Hans G, et al. Opioids and the management of chronic severe pain in the elderly: consensus statement of an International Expert Panel with focus on the six clinically most often used World Health Organization Step III opioids (buprenorphine, fentanyl, hydromorphone, methadone, morphine, oxycodone). Pain Pract 2008;8(4):287–313. [DOI] [PubMed] [Google Scholar]
  • 16.Plein LM, Rittner HL. Opioids and the immune system - friend or foe. Br J Pharmacol 2017. [DOI] [PMC free article] [PubMed]
  • 17.Roy S, Ninkovic J, Banerjee S, Charboneau RG, Das S, Dutta R, et al. Opioid drug abuse and modulation of immune function: consequences in the susceptibility to opportunistic infections. J Neuroimmune Pharmacol 2011;6(4):442–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Vallejo R, de Leon-Casasola O, Benyamin R. Opioid therapy and immunosuppression: a review. Am J Ther 2004;11(5):354–65. [DOI] [PubMed] [Google Scholar]
  • 19.Wang J, Barke RA, Charboneau R, Schwendener R, Roy S. Morphine induces defects in early response of alveolar macrophages to Streptococcus pneumoniae by modulating TLR9-NF-kappa B signaling. J Immunol 2008;180(5):3594–600. [DOI] [PubMed] [Google Scholar]
  • 20.Wang J, Barke RA, Charboneau R, Roy S. Morphine impairs host innate immune response and increases susceptibility to Streptococcus pneumoniae lung infection. J Immunol 2005;174(1):426–34. [DOI] [PubMed] [Google Scholar]
  • 21.Breslow JM, Monroy MA, Daly JM, Meissler JJ, Gaughan J, Adler MW, et al. Morphine, but not trauma, sensitizes to systemic Acinetobacter baumannii infection. J Neuroimmune Pharmacol 2011;6(4):551–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Huang SS, Johnson KM, Ray GT, Wroe P, Lieu TA, Moore MR, et al. Healthcare utilization and cost of pneumococcal disease in the United States. Vaccine 2011;29(18):3398–412. [DOI] [PubMed] [Google Scholar]
  • 23.Epidemiology and Prevention of Vaccine-Preventable Diseases In: Hamborsky JKA, Wolfe C, ed. 13 ed: Communication and Education Branch, National Center for Immunization and Respiratory Diseases, Centers for Disease Control and Prevention; 2015. [Google Scholar]
  • 24.Moore MR, Link-Gelles R, Schaffner W, Lynfield R, Lexau C, Bennett NM, et al. Effect of use of 13-valent pneumococcal conjugate vaccine in children on invasive pneumococcal disease in children and adults in the USA: analysis of multisite, population-based surveillance. Lancet Infect Dis 2015;15(3):301–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Use of 13-valent pneumococcal conjugate vaccine and 23-valent pneumococcal polysaccharide vaccine for adults with immunocompromising conditions: recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Morb Mortal Wkly Rep 2012;61(40):816–9. [PubMed] [Google Scholar]
  • 26.Tomczyk S, Bennett NM, Stoecker C, Gierke R, Moore MR, Whitney CG, et al. Use of 13-valent pneumococcal conjugate vaccine and 23-valent pneumococcal polysaccharide vaccine among adults aged >/=65 years: recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Morb Mortal Wkly Rep 2014;63(37):822–5. [PMC free article] [PubMed] [Google Scholar]
  • 27.Schuchat A, Hilger T, Zell E, Farley MM, Reingold A, Harrison L, et al. Active bacterial core surveillance of the emerging infections program network. Emerg Infect Dis 2001;7(1):92–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Essebag V, Platt RW, Abrahamowicz M, Pilote L. Comparison of nested case-control and survival analysis methodologies for analysis of time-dependent exposure. BMC Med Res Methodol 2005;5(1):5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Suissa S Novel Approaches to Pharmacoepidemiology Study Design and Statistical Analysis Pharmacoepidemiology: John Wiley & Sons, Ltd; 2002:785–805. [Google Scholar]
  • 30.Von Korff M, Saunders K, Thomas Ray G, Boudreau D, Campbell C, Merrill J, et al. De facto long-term opioid therapy for noncancer pain. Clin J Pain 2008;24(6):521–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Kim DH, Schneeweiss S. Measuring frailty using claims data for pharmacoepidemiologic studies of mortality in older adults: evidence and recommendations. Pharmacoepidemiol Drug Saf 2014;23(9):891–901. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Hosmer DWLS, Sturdivant RX. Applied Logistic Regression, Third Edition. John Wiley & Sons, Inc.; 2013. [Google Scholar]
  • 33.Arbogast PG, Seeger JD, Group DMCSVW. Summary Variables in Observational Research: Propensity cores and Disease Risk Scores. In: Quality AfHRa, ed; 2012.
  • 34.Arbogast PG, Ray WA. Performance of disease risk scores, propensity scores, and traditional multivariable outcome regression in the presence of multiple confounders. Am J Epidemiol 2011;174(5):613–20. [DOI] [PubMed] [Google Scholar]
  • 35.Arbogast PG, Ray WA. Use of disease risk scores in pharmacoepidemiologic studies. Stat Methods Med Res 2009;18(1):67–80. [DOI] [PubMed] [Google Scholar]
  • 36.Schneeweiss S Sensitivity analysis and external adjustment for unmeasured confounders in epidemiologic database studies of therapeutics. Pharmacoepidemiol Drug Saf 2006;15(5):291–303. [DOI] [PubMed] [Google Scholar]
  • 37.Inagi T, Suzuki M, Osumi M, Bito H. Remifentanil-based anaesthesia increases the incidence of postoperative surgical site infection. J Hosp Infect 2015;89(1):61–8. [DOI] [PubMed] [Google Scholar]
  • 38.Schwacha MG, McGwin G Jr., Hutchinson CB, Cross JM, Maclennan PA, Rue LW 3rd. The contribution of opiate analgesics to the development of infectious complications in burn patients. Am J Surg 2006;192(1):82–6. [DOI] [PubMed] [Google Scholar]
  • 39.Shao YJ, Liu WS, Guan BQ, Hao JL, Ji K, Cheng XJ, et al. Contribution of Opiate Analgesics to the Development of Infections in Advanced Cancer Patients. Clin J Pain 2017;33(4):295–9. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Appendix Figure 1
Appendix Figure 2
Appendix Figure 3
Appendix Tables
Appendix Text

RESOURCES