Abstract
Background Context:
Recent work indicates no increased mortality risk with concurrent gabapentin and opioid use when using an active comparator control design. However, concurrent gabapentin and opioid prescriptions have been associated with greater risk of respiratory depression in some studies.
Purpose:
To compare the risk of respiratory events among Medicare enrollees with histories of spine-related diagnoses treated with gabapentin + opioids vs those treated with tricyclic antidepressants (TCA) or duloxetine + opioids. We hypothesized that enrollees treated with gabapentin + opioids would have increased risk of adverse respiratory events compared to those treated with an active control + opioids.
Study Design/Setting:
Propensity score-matched cohort study with an incident user, active comparator (TCA/duloxetine) control design. The primary analysis included those who concurrently (within 30 days) filled ≥1 incident gabapentin + ≥1 opioid or ≥1 incident TCA/duloxetine + ≥1 opioid prescription.
Patient Sample:
U.S. Medicare beneficiaries with histories of spine-related diagnoses 2017-2019. People treated with gabapentin + opioids (n=66,860) were matched on demographic and clinical factors to people treated with TCAs/duloxetine + opioids (n=66,860).
Outcome Measures:
Time to a composite respiratory outcome consisting of mechanical ventilation, intubation, respiratory failure, pneumonia, or acute respiratory distress syndrome.
Methods:
Cox proportional hazard regression was used to estimate adjusted hazard ratios (aHRs) and 95% confidence intervals (95% CIs).
Results:
Among 133,720 Medicare enrollees (median age 73.3 years; 66.9% female), 6277 (4.7%) experienced respiratory events before the end of follow-up. A total of 3469 (5.2%) of people who were treated with gabapentin + opioids (median initial dose/day of gabapentin was 300 mg) had respiratory events compared to 2808 (4.2%) of those treated with an active control + opioids. The increased risk in those treated with gabapentin + opioids was statistically significant after adjustment (HR 1.19; 95% CI 1.13, 1.25; p<0.0001). The most common respiratory events were pneumonia (3.7% of people in the gabapentin + opioids group versus 3.0% of people in the TCA/duloxetine + opioids group) and respiratory failure (2.3% in the gabapentin + opioids group versus 1.8% in the TCA/duloxetine + opioids group). Results were similar in analyses (a) restricted to ≤30-day follow-up and (b) that required ≥2 fills of each prescription.
Conclusions:
While recent work indicates no increased mortality risk with concurrent gabapentin and opioid use in this population, the current findings suggest clinicians should exercise caution in prescribing gabapentin to older adults with spine conditions who are using opioids, due to possible impacts on respiratory events. However, we cannot be certain that unmeasured confounding may explain these results and replication is needed.
Keywords: Anti-depressants, duloxetine, gabapentin, opioids, propensity matching, spine pain
Introduction
Given recommendations from the Centers for Disease Control and Prevention [1] and the American Geriatrics Society [2] for caution in prescribing opioids for pain, prescriptions of alternative medications to treat pain, such as gabapentin, have increased [3, 4]. In a large randomized controlled trial of patients with back pain in primary care [4], an epidemiologic intervention that decreased opioid use also resulted in a higher number of prescriptions of tricyclic antidepressants (TCAs) and non-steroidal anti-inflammatory drugs (NSAIDs). Combinations of medications are also commonly used to treat chronic pain in the United States [5, 6].
Although use of gabapentin alone is generally considered safe, some studies show associations with adverse respiratory events when it is used in combination with opioids [7]. One study found an association between gabapentin prescriptions and chronic obstructive pulmonary disease (COPD) exacerbations [8]. Other studies have found that perioperative use of gabapentinoids for pain is associated with postoperative respiratory depression [9], mechanical ventilation [10], and pneumonia compared to patients not exposed to gabapentin perioperatively [11]. On the other hand, some studies have found advantages associated with gabapentin use compared to a placebo, including reduced total morphine consumption when gabapentin was used preoperatively [12], and reduced fentanyl consumption and shorter durations of mechanical ventilation when gabapentin was used in an intensive care population [13].
While prior studies have found increased risks of respiratory events among people treated with gabapentin + opioids compared to opioids alone [14], these results may be explained by confounding due to the need for a second pain medication. That is, people taking both gabapentin and opioids may on average have more severe pain and comorbidities than those taking only opioids. These factors alone might account for greater risk of adverse events. Such confounding may be mitigated in research by using an active control medication as a comparator. TCAs and duloxetine, like gabapentin, are commonly prescribed to treat chronic pain. Therefore, patients receiving TCAs or duloxetine (TCA/duloxetine) are a reasonable active control group for patients receiving gabapentin.
In a recent study of the entire Medicare population with histories of spine-related conditions which used TCAs/duloxetine as an active control group [15], we found no increased risk of mortality among Medicare enrollees treated with gabapentin + opioids compared to those treated with TCAs/duloxetine + opioids. However, a secondary analysis found that people treated with gabapentin + opioids were at a slight but statistically significant increased risk of experiencing major medical complications. This raised the question of whether adverse respiratory events might explain the increased risk of major medical complications that were observed in people treated with gabapentin + opioids.
The purpose of this study was to compare time to adverse respiratory events among Medicare enrollees with histories of spine-related conditions who were treated with gabapentin + opioids compared to those treated with TCA/duloxetine + opioids. We hypothesized that enrollees treated with gabapentin + opioids would have increased risk of adverse respiratory events compared to those treated with an active control + opioids.
Methods
This manuscript adheres to RECORD-PE[16] and START-RWE[17] guidelines for reporting. This study was determined to be exempt from review by the University of Washington Institutional Review Board.
Study Population
We identified enrollees from the 100% sample of Centers for Medicare and Medicaid Services (CMS) beneficiaries from 2017-2019 who had International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) inpatient or outpatient diagnosis codes [18] for spine-related conditions (e.g., neck pain, low back pain, spinal stenosis, radiculopathy, others; eTable 1 [19]) supplemented by previously developed algorithms used by our research group [20-24] and crosswalks from ICD-9-CM to ICD-10-CM [25, 26]. We excluded individuals with the following on or before their “index” fill date of gabapentin, TCA/duloxetine, or opioids: age <65 years; not enrolled in CMS Part A, Part B and Part D for ≥1 year before index; enrolled in a health maintenance organization (HMO) health plan (because we did not have access to data from these plans). We also excluded people with diagnoses of “red flag” spine conditions (e.g., spine fracture) or seizure disorders (eTable 2) because we wanted to create a sample of people likely to have been prescribed the drugs of interest for pain not associated with these conditions.
Medication Exposure Groups
Medication exposures (filled prescriptions only) were ascertained from Medicare Part D claims. People with incident gabapentin fills (eTable 3) and concurrent opioid fills within 30 days were compared to an active control group of people who had incident TCA or duloxetine fills and concurrent (within 30 days) opioid fills (eFigures 1a-1b). People in the gabapentin group were not allowed to have had gabapentin fills during the previous year; similarly, people in the TCA/duloxetine groups were not allowed to have had fills of TCA or duloxetine in the previous year. Selective serotonin reuptake inhibitors and serotonin-noradrenaline reuptake inhibitors (besides duloxetine) were not included in the active control group because they are less commonly prescribed for pain [27]. To ensure that opioids and the active comparator treatment were likely to have been taken concurrently, all beneficiaries included in our analysis had an “index fill” and a “qualifying fill.” For people in the gabapentin + opioids group, (1) the index fill could have been the calendar date of an opioid fill and the qualifying fill was the date on which they filled their first gabapentin prescription, or (2) the index fill could have been the calendar date on which they filled their first gabapentin prescription and the qualifying fill was the date of their opioid prescription fill. Similarly, for people in the TCA/duloxetine + opioids group, (1) the index fill could have been the calendar date on which a person filled an opioid prescription and the qualifying fill was the date on which they filled their first TCA/duloxetine prescription, or (2) the index fill could have been the calendar date on which they filled their first TCA/duloxetine prescription and the qualifying fill was the date of their opioid fill. If both relevant fills occurred on the same day, the index and qualifying fill dates were the same. We note that our analysis did not require opioid, gabapentin, TCA, or duloxetine fills to have been associated with any specific diagnosis and that these medications were not necessarily prescribed to address symptoms associated with the spine diagnoses that identified the patient sample. We believed this was appropriate because concerns regarding the effects of these medications on respiratory events would be relevant regardless of the indication for the medication.
The primary analysis examined people who filled ≥1 gabapentin + ≥1 opioid prescription within a 30-day period compared to those who filled ≥1 TCA/duloxetine + ≥1 opioid prescription within 30 days. This analysis was designed to capture early respiratory events based on the putative mechanism of risk (that the medications taken concurrently would increase the likelihood of respiratory depression) [7]. We also conducted a secondary analysis designed to increase the likelihood that people actually consumed the medications from their filled prescriptions. The typical period of time between refills for gabapentin is 1-3 months [28], so this secondary analysis examined people with ≥2 gabapentin fills + ≥2 opioid fills within a 120-day exposure period and compared them to those with ≥2 TCA/duloxetine fills + ≥2 opioid fills in 120 days (eFigures 1c-1d).
We also calculated the mean daily doses of opioids (in milligram morphine equivalents [MMEs][29]), gabapentin, duloxetine, and each TCA (in milligrams (mg)) for the index fills as well as the final dose closest to the end of follow-up. We focused on the last dose before the end of follow-up because we expected that this would be more relevant to respiratory outcomes than initial doses, which often begin low and are titrated upward if well tolerated. We compared the final dosages before the end of follow-up among those who did and did not have respiratory events.
Outcomes and Censoring Variables
Beginning from the day of the qualifying fill, we assessed occurrence of any of five respiratory events at any time and at 30 days, at 1 year, and at 2 years after the qualifying date: intubation, mechanical ventilation, respiratory failure, pneumonia, or acute respiratory distress syndrome (ARDS) (eTable 4). Because the posited mechanism of action of gabapentin + opioids is that the medications may increase risk of respiratory depression, [14] we chose respiratory events associated with respiratory depression and its sequelae as the main study outcomes. We did not include respiratory events that are not associated with respiratory depression (e.g., pulmonary embolism).
Censoring events were death; fills of benzodiazepine or pregabalin after the qualifying fill (because concurrent use of benzodiazepines independently increases respiratory depression among opioid users [30] and pregabalin is pharmacologically similar to gabapentin but often used differently, e.g., pregabalin is often prescribed when gabapentin is not tolerated or is ineffective, and pregabalin is classified as a Schedule V controlled substance by the US Drug Enforcement Agency, which may make providers reluctant to prescribe it to older adults [31]); no opioid refill for >45 days; TCA/duloxetine fill in comparison groups treated with gabapentin; no gabapentin refill for >180 days in comparison groups treated with gabapentin; gabapentin fill in comparison groups treated with TCA/duloxetine; no TCA/duloxetine refill for >180 days in comparison groups treated with TCA/duloxetine; or the end of the study (December 31, 2019), whichever occurred first.
Statistical Analysis
To increase internal validity and limit confounding, we used two approaches to ensure that the TCA/duloxetine group was as comparable as possible to the gabapentin group in terms of covariates that could have caused residual confounding. First, we included a wide range of variables in the logistic regression used for propensity score 1:1 matching based on clinical knowledge and prior literature (eTable 5). We chose to conduct 1:1 matching rather than 1:n matching because the power calculations that we performed a priori [32] indicated that 1:1 matching yielded sufficient power. In particular, given that duloxetine and TCAs are common treatments for mental health conditions, and the same is true for gabapentin (although to a lesser extent), we included several variables indicative of inpatient or outpatient psychiatric diagnoses in the propensity score, including whether people had diagnoses of alcohol abuse and opioid use disorder in the year prior to index, as well as the counts of the number of visit dates (as surrogates for severity) with diagnoses of mood disorders, generalized anxiety disorder (GAD), and post-traumatic stress disorder (PTSD) in the year prior.
Second, because we expected certain variables that occurred between the index and qualifying fills to be strongly related to whether people received the prescriptions of interest and to respiratory events, we matched on them “exactly” in addition to matching on the propensity score. These included (1) average daily MME [29] of opioid prescriptions filled on/between the index and qualifying fill dates, matched within 9 categories (>0 to 2.1; >2.1 to 5.3; >5.3 to 10.0; >10.0 to 15; >15 to 20; >20 to 30; >30 to 40; >40 to 64, and >64) based on MME distribution; (2) the length of time between the index and qualifying fills to make the timing of concurrent opioid treatment comparable between groups; (3) the presence of a respiratory event between the index and qualifying fills (eTable 4); and (4) the presence of non-respiratory complications between index and qualifying fills (eTable 6).
We calculated descriptive statistics (numbers, proportions, medians, interquartile ranges (IQRs), and standardized mean differences) to ensure the matched groups were similar. Kaplan-Meier curves were plotted to graphically assess the proportional hazards assumption, which was met for all analyses. Cox proportional hazard regression was used to estimate adjusted hazard ratios (aHRs) and 95% confidence intervals (95% CIs). Because death could have been a competing risk if it occurred prior to diagnosis of respiratory events, we conducted a sensitivity analysis using Fine and Gray’s sub-distribution hazard model [33] to estimate the risk of a respiratory event.
We conducted two exploratory post hoc analyses to examine possible effect modification of the gabapentin-respiratory event association by 1) whether the participant had neuropathic pain diagnoses and 2) initial dosage of gabapentin. For the latter, in three separate analyses, we restricted the gabapentin + opioids group to those with decreasing initial dosages of gabapentin (≤300, ≤200, ≤100 mg/day) to determine if there was an inflection point at which the risk of respiratory events became statistically non-significant. These analyses repeated the analytic approach, including the propensity and exact matching procedures, in strata of the possible effect modifier. We also evaluated the number of people who received spine surgery during the follow-up period. We performed a post hoc analysis in which we censored people at the time of a spine surgery and a separate post hoc analysis in which we included receipt of spine surgery as a time-varying covariate. We also performed a post hoc analysis comparing a group that filled prescriptions for gabapentin + opioids to a group that filled TCA (and not duloxetine) + opioids, and similarly, a comparison of those who filled gabapentin + opioids to those who filled duloxetine (and not a TCA) + opioids. Finally, we conducted a post hoc analysis in which we included diagnoses of mental health conditions (mood disorders, generalized anxiety disorder, and post-traumatic stress disorder) and whether people had fills of medications often used to treat mental health disorders (selective serotonin reuptake inhibitors (SSRIs), serotonin and norepinephrine reuptake inhibitors (SNRIs) except for duloxetine, bupropion, and mirtazapine) between the index and qualifying dates as exact matching variables. All regression models were adjusted for the variables in the propensity score model (eTable 5). Two-sided p-values of <0.05 were considered statistically significant and analyses were performed using SAS Enterprise Guide version 7.15 and SAS for Windows version 9.4 (using the OneToManyMTCH macro) [34], SAS Institute Inc., Cary, NC.
Results
The flow of Medicare beneficiaries in the primary analysis is shown in Figure 1. Of the initial 525,884 people who were eligible for matching, 66,860 people treated with gabapentin + opioids matched to 66, 860 people treated with TCA/duloxetine + opioids. An abbreviated list of cohort characteristics is shown in Table 1 and a complete list is shown in eTable 7. The median (range) follow-up time was 45 (1-1093) days. Although the unmatched treatment groups differed with respect to various characteristics (eTable 7), the matched groups were closely comparable (Table 1 and eTable 7). All subsequent results refer to the matched groups. The majority of people were female, white, and had a mean (standard deviation) Charlson comorbidity score of 2.7 (2.7). In both treatment groups, mental health diagnoses and fills of medications besides opioids, gabapentin, TCA, or duloxetine that are often prescribed for pain were common. In the TCA/duloxetine group, 63% of Medicare beneficiaries filled duloxetine for their index fill and 37% filled a TCA. The median (IQR) doses of the index fills were: 300 mg (300, 900) for gabapentin, 30 mg (30, 60) for duloxetine and 25 mg (10, 30) for amitriptyline, the most common TCA that was filled.
Figure 1.
Flow of patients for the primary analysis.
Table 1.
Comparison of patients who received gabapentin to those who received TCAs/duloxetine using 1:1 matching in the primary analysis (selected variables).
| N (%) unless otherwise indicated | Gabapentin n=66,860 (50%) n (%) |
TCA/Duloxetine n=66,860 (50%) n (%) |
Standardized Mean Difference |
|---|---|---|---|
| DEMOGRAPHIC VARIABLES | |||
| Age (mean, standard deviation) | 75.0 ± 7.0 | 74.8 ± 7.2 | −0.02 |
| Female | 45,313 (67.8%) | 44,129 (66.0%) | 0.04 |
| Patient race | |||
| Unknown | 684 (1.0%) | 777 (1.2%) | 0.01 |
| White | 58,222 (87.1%) | 57,966 (86.7%) | −0.01 |
| Black | 5,226 (7.8%) | 5,324 (8.0%) | 0.01 |
| Other | 641 (1.0%) | 630 (0.9%) | 0.00 |
| Asian | 628 (0.9%) | 670 (1.0%) | 0.01 |
| Hispanic | 1,072 (1.6%) | 1,065 (1.6%) | 0.00 |
| North American Native | 387 (0.6%) | 428 (0.6%) | 0.01 |
| Buy-in from state | 14,903 (22.3%) | 15,132 (22.6%) | 0.01 |
| Indexa year | |||
| 2017 | 27,348 (40.9%) | 27,240 (40.7%) | 0.00 |
| 2018 | 21,655 (32.4%) | 21,692 (32.4%) | 0.00 |
| 2019 | 17,857 (26.7%) | 17,928 (26.8%) | 0.00 |
| CLINICAL VARIABLES | |||
| Number of days from first back pain diagnosis to index (mean, standard deviation) | 570 ± 360 | 560 ± 350 | −0.02 |
| Number of days from most recent spine-related diagnosis to index | 130 ± 220 | 120 ± 200 | −0.01 |
| Charlson co-morbidity score (mean, standard deviation) | 2.7 ± 2.7 | 2.7 ± 2.7 | 0.01 |
| Spine-related neuropathic pain in year prior to index | 33,955 (50.8%) | 35,117 (52.5%) | 0.03 |
| Cancer (excluding non-melanoma skin cancer) in year prior to index | 9,918 (14.8%) | 10,321 (15.4%) | 0.02 |
| Receipt of spine surgery in year prior to index | 1,339 (2.0%) | 1,658 (2.5%) | 0.03 |
| Respiratory events in year prior to index | |||
| # unique dates with diagnoses (mean ± standard deviation) | 0.7 ± 3.9 | 0.7 ± 4.2 | 0.01 |
| Intubation | 355 (0.5%) | 355 (0.5%) | 0.00 |
| Mechanical ventilation | 453 (0.7%) | 460 (0.7%) | 0.00 |
| Respiratory failure | 5,896 (8.8%) | 5,921 (8.9%) | 0.00 |
| Pneumonia | 3,066 (4.6%) | 3,160 (4.7%) | 0.00 |
| Acute respiratory distress syndrome | 238 (0.4%) | 257 (0.4%) | 0.00 |
| Mental health diagnoses/procedures in year prior to index | |||
| Alcohol abuse | 356 (0.5%) | 395 (0.6%) | 0.01 |
| Opioid use disorder | 12,448 (18.6%) | 13,056 (19.5%) | 0.02 |
| Insomnia | 8,971 (13.4%) | 8,566 (12.8%) | −0.02 |
| Electroconvulsive therapy | 15 (0.0%) | 11 (0.0%) | 0.00 |
| Therapeutic repetitive transcranial magnetic stimulation treatment | 16 (0.0%) | 18 (0.0%) | 0.00 |
| Number of unique dates with generalized anxiety disorder diagnoses (mean, standard deviation) | 0.3 ± 1.7 | 0.3 ± 1.7 | 0.01 |
| Number of unique dates with post-traumatic stress disorder diagnoses (mean, standard deviation) | 0.04 ± 0.8 | 0.04 ± 0.8 | 0.00 |
| Number of unique dates with depression diagnoses (mean, standard deviation) | 1.1 ± 4.0 | 1.1 ± 3.5 | 0.00 |
| Medication fills in year prior to index | |||
| NSAID | 25,223 (37.7%) | 25,190 (37.7%) | 0.00 |
| Acetaminophen | 720 (1.1%) | 710 (1.1%) | 0.00 |
| Muscle relaxer | 14,629 (21.9%) | 14,688 (22.0%) | 0.00 |
| Benzodiazepine | 19,123 (28.6%) | 18,627 (27.9%) | −0.02 |
| Pregabalin | 3,543 (5.3%) | 4,183 (6.3%) | 0.04 |
| Anti-psychotics | 3,141 (4.7%) | 3,043 (4.6%) | 0.00 |
| Selective serotonin/ norepinephrine reuptake inhibitors | 20,642 (30.9%) | 20,151 (30.1%) | −0.02 |
| Bupropion | 3,542 (5.3%) | 3,433 (5.1%) | 0.00 |
| Trazodone/nefazodone | 7,197 (10.8%) | 7,022 (10.5%) | 0.00 |
| Monoamine oxidase inhibitor | 2,704 (4.0%) | 2,738 (4.1%) | 0.00 |
| Duration of opioid use (mean, standard deviation days) | 160 ± 200 | 170 ± 190 | 0.03 |
| Average daily milligram morphine equivalents (MME) from index to qualifying dates (median, IQR) | 35.6 ± 62.9 | 36.3 ± 63.2 | 0.01 |
| Average daily milligram morphine equivalents (MME) from qualifying to end of follow-up (median, IQR) | 28.7 ± 55.8 | 36.3 ± 63.2 | 0.06 |
| Days from index to qualifying (median, IQR) | 7.6 ± 8.2 | 7.9 ± 8.4 | 0.04 |
| 0-7 | 39,903 (59.7%) | 39,903 (59.7%) | 0.00 |
| 8-30 | 26,957 (40.3%) | 26,957 (40.3%) | |
| Respiratory complications from index to qualifying dates | |||
| Intubation | <11 | <11 | 0.00 |
| Mechanical ventilation | <11 | <11 | 0.00 |
| Respiratory failure | 14 (0.0%) | 14 (0.0%) | 0.00 |
| Pneumonia | 49 (0.1%) | 49 (0.1%) | 0.00 |
| Acute respiratory distress syndrome (ARDS) | <11 | <11 | 0.00 |
| Non-respiratory complications from index to qualifying dates | |||
| High acute riskb | <11 | <11 | 0.00 |
| Moderate acute riskc | 2,050 (3.1%) | 2,050 (3.1%) | 0.00 |
| Low acute riskd | 1,862 (2.8%) | 1,862 (2.8%) | 0.00 |
Index=Date person had first fill of gabapentin, TCA, or duloxetine
High acute risk complications=cardiac arrest, cardiopulmonary resuscitation
Moderate acute risk complications=sepsis, fluid and electrolyte disorders, weight loss, cancer, acute myocardial infarction, paralysis
Low acute risk complications=coagulopathy, gastrointestinal bleed, cardiac arrhythmias, stroke
Abbreviations
MME: Milligram morphine equivalents; NSAID: Nonsteroidal anti-inflammatory drug; TCA: Tricyclic antidepressant
We observed a statistically significant increased risk of respiratory events in those treated with gabapentin + opioids over the entire follow-up time (eFigure 2; Figure 2) (aHR, 1.19; 95% CI, 1.13, 1.25). Comparing those treated with gabapentin + opioids versus those treated with TCA/duloxetine + opioids, the proportion at risk for respiratory events was very similar at 1-month, but the difference widened by 1-year and 2-years after index (Figure 2).
Figure 2.
Primary analysis Cox proportional hazard curves for any-time respiratory events.
The numbers and percentages of people who experienced each respiratory event are shown in Table 2. Although respiratory events were uncommon in both drug treatment groups, a greater proportion of people in the gabapentin group experienced a respiratory event (5.2% versus 4.2%). The most common types of respiratory events were pneumonia (3.7% of people in the gabapentin + opioids group versus 3.0% in the TCA/duloxetine + opioids group) and respiratory failure (2.3% in the gabapentin + opioids group versus 1.8% in the TCA/duloxetine + opioids group). People in the gabapentin group were somewhat more likely to have been censored because they did not refill an opioid prescription for >45 days (51.8% versus 46.4%) and less likely than people in the TCA/duloxetine + opioids group to have been censored for filling pregabalin prescriptions (2.7% in the gabapentin + opioids group versus 4.6% in the TCA/duloxetine + opioids group) (Table 2). Results were similar when we examined 30-day respiratory events (eTable 8; eFigures 3-4).
Table 2.
Respiratory Outcomes at any time.
| Gabapentin n=66,860 (50%) n (%) |
TCA/Duloxetine n=66,860 (50%) n (%) |
|
|---|---|---|
| Any respiratory complication | 3,469 (5.2%) | 2,808 (4.2%) |
| Days to first respiratory complication (median, IQR) | 36 (12, 109) | 38 (12, 115) |
| Intubation | 197 (0.3%) | 178 (0.3%) |
| Mechanical ventilation | 245 (0.4%) | 216 (0.3%) |
| Respiratory failure | 1,538 (2.3%) | 1,196 (1.8%) |
| Pneumonia | 2,464 (3.7%) | 2,011 (3.0%) |
| Acute respiratory distress syndrome (ARDS) | 58 (0.1%) | 38 (0.1%) |
| Reasons for follow-up end | ||
| Disenrolled | 8,826 (13.2%) | 8,922 (13.3%) |
| No opioid refill >45 days | 34,616 (51.8%) | 31,034 (46.4%) |
| Pregabalin fill | 1,796 (2.7%) | 3,070 (4.6%) |
| Benzodiazepine fill | 11,370 (17.0%) | 11,667 (17.5%) |
| TCA/duloxetine fill in gabapentin group | 1,815 (2.7%) | -- |
| Gabapentin fill in TCA/duloxetine group | -- | 3,790 (5.7%) |
| No gabapentin refill >180 days | 4,362 (6.5%) | - |
| No TCA/duloxetine refill >180 days | -- | 4,873 (7.3%) |
| Respiratory complication | 3,469 (5.2%) | 2,808 (4.2%) |
| Died | 606 (0.9%) | 696 (1.0%) |
People with respiratory events had similar median opioid doses per day in both drug treatment groups (Table 3). The median dose of gabapentin in the last fill prior to the end of follow-up was higher among people with respiratory events compared to those without (median (IQR) 600 (300, 900) mg in those with respiratory events versus 400 (300, 900) mg in those without). Those who did not have respiratory events tended to have similar dosages of duloxetine and but somewhat higher doses of each of the TCAs (Table 3).
Table 3.
Median daily doses of opioids, gabapentin, TCA, and duloxetine in their last doses before the end of follow-up, stratified by whether person had respiratory complication.
| Gabapentin | TCA/Duloxetine | |||||||
|---|---|---|---|---|---|---|---|---|
| Medication | No respiratory complication |
Respiratory complication |
No respiratory complication |
Respiratory complication |
||||
| n | Median (IQR) |
n | Median (IQR) |
n | Median (IQR) |
n | Median (IQR) |
|
| Opioid (MME/day) | 63,391 | 30.0 (15.4, 45.0) | 3469 | 30.0 (20.0, 60.0) | 64,052 | 30.0 (15.3, 45.0) | 2808 | 32.0 (20.0, 60.0) |
| Gabapentin (mg/day) | 63,391 | 400 (300, 900) | 3469 | 600 (300, 900) | -- | -- | -- | -- |
| Duloxetine (mg/day) | -- | -- | -- | -- | 40,013 | 30.0 (30.0, 60.0) | 1884 | 30.0 (30.0, 60.0) |
| Amitriptyline (mg/day) | -- | -- | -- | -- | 11,775 | 25.0 (10.0, 50.0) | 450 | 25.0 (10.0, 50.0) |
| Imipramine (mg/day) | -- | -- | -- | -- | 662 | 25.0 (25.0, 50.0) | 26 | 47.5 (25.0, 75.0) |
| Nortriptyline (mg/day) | -- | -- | -- | -- | 5308 | 25.0 (1.0, 30.0) | 187 | 25.0 (10.0, 30.0) |
| Doxepin (mg/day) | -- | -- | -- | -- | 6136 | 20.0 (10.0, 37.5) | 256 | 22.5 (10.0, 40.0) |
| Other TCAa (mg/day) | -- | -- | -- | -- | 297 | 25.0 (20.0, 75.0) | 12 | 37.5 (17.5, 75.0) |
Amoxapine, clomipramine, desipramine, maprotiline, protriptyline, trimipramine maleate
Abbreviations
EOF: End of follow-up; IQR: Interquartile range; Mg: Milligram; MME: Milligram morphine equivalents; TCA: Tricyclic anti-depressant
Results of the secondary analysis requiring ≥2 fills of each medication of interest were similar to those of the primary analysis. A total of 165,585 people who filled ≥2 gabapentin and opioid prescriptions were eligible to be matched with 42,344 people who filled ≥2 TCA/duloxetine and opioid prescriptions; among these, 15,635 were in each matched group (eFigure 5). The groups were well-matched (eTable 9). We observed a greater proportion of respiratory events at any time among people treated with gabapentin + opioids (eTable 10) and an increased adjusted risk of a respiratory event among people treated with gabapentin + opioids (eFigures 6-7; aHR, 1.20; 95% CI, 1.08, 1.32). Results for 30-day respiratory events in the secondary analysis were similar to the primary analysis though the adjusted HR did not achieve statistical significance (aHR, 1.14; 95% CI 0.94, 1.37) (eTable 11; eFigures 8-9).
Sensitivity analyses using a Fine-Gray model for competing risks of death showed no material differences from the primary analyses in which individuals who died were censored (data not shown). Exploratory post hoc analyses examining effect modification found similar-magnitude gabapentin-respiratory event association irrespective of baseline neuropathic pain diagnosis (data not shown). In our post hoc analysis restricting people in the gabapentin + opioids group to decreasing initial dosages of gabapentin, we observed statistically significantly increased adjusted HRs in the gabapentin + opioids group compared to the TCA/duloxetine + opioids group, even when we restricted the gabapentin + opioids group to those who filled prescriptions for ≤100 mg of gabapentin per day (data not shown). A total of n=1984 (1.5%) people received spine surgery during the follow-up period, of whom n=1406 were treated with gabapentin + opioids and n=578 were treated with TCA/duloxetine + opioids. The results of post hoc analyses in which we censored people upon receipt of spine surgery and in which we considered receipt of spine surgery as a time-dependent covariate were not meaningfully different from those of our main analyses (data not shown). In the post hoc analyses in which we compared people who filled gabapentin + opioids to those who filled a TCA (and not duloxetine) + opioids, results were almost identical to the original analysis (aHR, 1.19; 95% CI 1.10-1.28, eFigure 10). The result was slightly attenuated but still statistically significant when we compared people who filled gabapentin + opioids to those who filled duloxetine (and not TCA) + opioids (aHR, 1.12; 95% CI 1.06-1.19, eFigure 11). In the post hoc analysis in which we included whether people had encounters with mood disorder, generalized anxiety disorder, post-traumatic stress disorder, and certain medications used to treat mental health between the index and qualifying dates as exact match variables, results were again almost identical to the original analysis (aHR, 1.19; 95% CI 1.13-1.25, eFigure 12).
Discussion
In this analysis of Medicare beneficiaries with histories of spine-related diagnoses, we found that people treated with gabapentin + opioids were at a small but statistically significant increased risk of respiratory events compared to people treated with TCA/duloxetine + opioids, with a 19% greater relative risk of a respiratory event within 1 year and a 2.0% greater absolute risk within 1 year. This relationship was consistent when we examined respiratory events within 30 days, as well as when we required people in the matched cohorts to have filled ≥2 prescriptions for gabapentin/active control + opioids and therefore were more likely to have consumed their medications.
The combination of gabapentin + opioids may lead to increased risk of respiratory events by causing respiratory depression [35-38]. This hypothesis is supported by studies comparing people exposed to gabapentin versus non-users with respect to outcomes including COPD exacerbations [9], post-operative mechanical ventilation [10], pneumonia [11], and respiratory depression [8, 37]. However, by comparing gabapentin users to non-users without an active control group [9-11], these analyses were subject to substantial residual confounding and it is possible that some unmeasured confounder rather than gabapentin was causing the respiratory events. One of these studies [8] did perform a secondary analysis with an active control group. These authors examined the risk of COPD exacerbations in people in Quebec, Canada with neuropathic or chronic pain who were treated with gabapentinoids versus NSAIDs. They found that those treated with gabapentinoids were at increased risk of COPD exacerbations. Therefore, our study is consistent with and adds to the limited previous work demonstrating increased risk of adverse respiratory events with combined use of gabapentin and opioid medication.
The current study also builds on past work by showing that the increased risk of respiratory events with combined opioid and gabapentin applies not only to select groups of people with certain medical conditions (e.g., breast cancer survivors [38], surgical patients [9], those with COPD [8]) but more broadly to people with histories of spinal pain, the most common pain condition associated with office visits that result in opioid prescriptions [39]. This is important given that spine-related diagnoses are common diagnoses and are a substantial driver of increasing health care costs in the United States [40].
It is important to note that the significantly increased risk of respiratory events with concurrent gabapentin and opioid use in the current analysis was present even though the doses of gabapentin used were quite low, with a median daily gabapentin dose of about 300 mg at the time of gabapentin initiation, increasing to a 400 mg daily gabapentin dose throughout follow-up among people who did not have subsequent respiratory events and to a 600 mg daily gabapentin dose among people who did have subsequent respiratory events. This highlights that, among older adults who are using opioids, even low doses of gabapentin may increase the likelihood of respiratory events and supports the current recommendation from the Beers criteria [2] that the combination of gabapentin and opioids should be used cautiously in older adults, or avoided.
The current findings are important to consider in the context of our recent work using nearly identical methods that found no significant difference in the risk of mortality (OR= 0.98 [0.90, 1.06]; p=0.63) among those treated with gabapentin + opioids versus those treated with TCA/duloxetine + opioids [15], with a point estimate that did not suggest even a slight tendency towards a detrimental effect on mortality. Together, these results suggest the possibility that adverse respiratory events after gabapentin initiation among older adults using opioids may lead to clinical changes, such as not refilling gabapentin and opioid prescriptions in people who have had respiratory events, that prevent subsequent mortality. This underscores the importance of educating patients and caregivers about the increased risk of respiratory events with opioid use, and potential augmentation of this increased risk when initiating gabapentin, so that patients and their caregivers can be watchful for changes in respiratory status. This also underscores the importance of healthcare professionals asking about shifts in respiratory status when seeing older patients who use opioids who are returning to clinic after recently initiating gabapentin.
We emphasize strongly that the current study’s findings apply only to gabapentin treatment in the context of concurrent opioid treatment among older adults and do not apply to gabapentin when opioids are not being used concurrently. Future studies are needed in samples of people not taking opioids to compare rates of respiratory events and other important outcomes such as falls in those taking gabapentin versus those taking an active control medication. Also, studies conducted in broad populations who do not have pre-existing respiratory diagnoses or other major comorbidities would be invaluable.
By using an active comparator group and extensive matching in this large, representative group of older US adults, we evaluated respiratory events in cohorts that were as alike as possible (in terms of characteristics that can be assessed from claims data) except with respect to their drug treatment regimens. The active comparator group also illustrated the high frequency of respiratory events among older adults using opioids in general, even those who did not use gabapentin. Despite use of an active control group, however, it remains possible that residual confounding by factors not available in claims data such as pain severity and functional limitations remained. Furthermore, although we carefully matched on a variety of mental health diagnoses and procedures and conducted additional post hoc analyses to better control for residual confounding by these diagnoses, it is possible that some people may have had undiagnosed mood disorders which could have affected our attempts to create comparison groups as alike as possible. Because our study population included only older adults with histories of spine-related conditions, we cannot be certain that our results would apply to younger populations or people with other pain conditions.
In summary, among older adults with histories of spine conditions who use opioids, gabapentin initiation, even at low doses, conferred a small but statistically significant risk of adverse respiratory events compared to an active control group. Thus, healthcare professionals should carefully consider whether the benefits of initiating gabapentin in older adults who have already been prescribed opioids outweigh the risks. If gabapentin is prescribed in older adults taking opioids, our findings suggest the importance of informing patients and caregivers of the risk of respiratory events and carefully monitoring them for respiratory depression and other respiratory problems.
Supplementary Material
Funding:
This work was supported by the University of Washington Clinical Learning, Evidence, And Research (CLEAR) Center for Musculoskeletal Disorders, Administrative, Methodologic and Resource Cores and NIAMS/NIH grant P30AR072572. The funding source had no role in the study design, collection, analysis and interpretation of the data, writing of the report, or the decision to submit this article for publication.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Competing Interests: The authors report no conflicts of interest.
References
- 1.Dowell D, Haegerich TM, Chou R. CDC Guideline for Prescribing Opioids for Chronic Pain--United States, 2016. JAMA. 2016;315(15):1624–45. 10.1001/jama.2016.1464. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Croke L. Beers Criteria for Inappropriate Medication Use in Older Patients: An Update from the AGS. Am Fam Physician. 2020;101(1):56–7. [PubMed] [Google Scholar]
- 3.Peet ED, Dana B, Sheng FY, Powell D, Shetty K, Stein BD. Trends in the Concurrent Prescription of Opioids and Gabapentin in the US, 2006 to 2018. JAMA Intern Med. 2022. 10.1001/jamainternmed.2022.5268. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Marcum ZA, Gold LS, James KT, et al. Effects of Including Epidemiologic Data in Lumbar Spine Imaging Reports on Prescribing Non-Opioid Medications for Pain. J Gen Intern Med. 2021;36(8):2237–43. 10.1007/s11606-021-06627-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hackbarth G, Christianson J, Miller M, editors. Polypharmacy and opioid use among Medicare Part D enrollees. Report to the Congress: Medicare and the Health Care Delivery System; 2015: MedPAC.gov Washington, DC. [Google Scholar]
- 6.Shmagel A, Ngo L, Ensrud K, Foley R. Prescription Medication Use Among Community-Based U.S. Adults With Chronic Low Back Pain: A Cross-Sectional Population Based Study. J Pain. 2018;19(10):1104–12. 10.1016/j.jpain.2018.04.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Food and Drug Administration. FDA warns about serious breathing problems with seizure and nerve pain medicines gabapentin (Neurontin, Gralise, Horizant) and pregabalin (Lyrica, Lyrica CR). Food and Drug Administration; 2019. [cited 2023 Dec 1]; Available from: https://www.fda.gov/media/133681/download. [Google Scholar]
- 8.Rahman AA, Dell'Aniello S, Moodie EEM, et al. Gabapentinoids and Risk for Severe Exacerbation in Chronic Obstructive Pulmonary Disease : A Population-Based Cohort Study. Ann Intern Med. 2024. 10.7326/M23-0849. [DOI] [PubMed] [Google Scholar]
- 9.Cavalcante AN, Sprung J, Schroeder DR, Weingarten TN. Multimodal analgesic therapy with gabapentin and its association with postoperative respiratory depression. Anesthesia & Analgesia. 2017;125(1):141–6. [DOI] [PubMed] [Google Scholar]
- 10.McGauvran MM, Ohnuma T, Raghunathan K, et al. Association Between Gabapentinoids and Postoperative Pulmonary Complications in Patients Undergoing Thoracic Surgery. J Cardiothorac Vasc Anesth. 2022;36(8 Pt A):2295–302. 10.1053/j.jvca.2021.10.003. [DOI] [PubMed] [Google Scholar]
- 11.Park CM, Inouye SK, Marcantonio ER, et al. Perioperative Gabapentin Use and In-Hospital Adverse Clinical Events Among Older Adults After Major Surgery. JAMA Intern Med. 2022;182(11):1117–27. 10.1001/jamainternmed.2022.3680. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Menda F, Koner O, Sayin M, Ergenoglu M, Kucukaksu S, Aykac B. Effects of single-dose gabapentin on postoperative pain and morphine consumption after cardiac surgery. J Cardiothorac Vasc Anesth. 2010;24(5):808–13. 10.1053/j.jvca.2009.10.023. [DOI] [PubMed] [Google Scholar]
- 13.Salarian S, Memary E, Taheri F, Bagheri B. Gabapentin as Add-On to Fentanyl and Midazolam in Patients Receiving Mechanical Ventilation: A Randomized, Blinded Study. Turk J Anaesthesiol Reanim. 2022;50(2):101–6. 10.5152/TJAR.2022.21366. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Tambon M, Montarnal B, Lepetit M, Lapeyre-Mestre M. Non-opioid antinociceptive drugs : risk of respiratory depression and death related to concomitant use of gabapentinoids in addition to opioids. Expert Opin Drug Saf. 2023;22(3):183–94. 10.1080/14740338.2023.2193397. [DOI] [PubMed] [Google Scholar]
- 15.Gold LS, Heagerty PJ, Hansen RN, et al. Mortality after concurrent treatment with gabapentin and opioids in older adults with spine diagnoses. Pain. 2024. 10.1097/j.pain.0000000000003448. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Langan SM, Schmidt SA, Wing K, et al. The reporting of studies conducted using observational routinely collected health data statement for pharmacoepidemiology (RECORD-PE). bmj. 2018;363. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Wang SV, Pinheiro S, Hua W, et al. STaRT-RWE: structured template for planning and reporting on the implementation of real world evidence studies. Bmj. 2021;372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.World Health Organization. ICD-10-CM. International Classification of Diseases, 10th revision, Clinical Modification. . 2015. [updated Oct 1 2015; cited 2022 December 1]; Available from: https://www.cdc.gov/nchs/icd/icd-10-cm.htm. [Google Scholar]
- 19.Martin BI, Lurie JD, Tosteson AN, et al. Indications for spine surgery: validation of an administrative coding algorithm to classify degenerative diagnoses. Spine (Phila Pa 1976). 2014;39(9):769–79. 10.1097/BRS.0000000000000275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Cherkin DC, Deyo RA, Volinn E, Loeser JD. Use of the International Classification of Diseases (ICD-9-CM) to identify hospitalizations for mechanical low back problems in administrative databases. Spine (Phila Pa 1976). 1992;17(7):817–25. 10.1097/00007632-199207000-00015. [DOI] [PubMed] [Google Scholar]
- 21.Deyo RA, Andersson G, Bombardier C, et al. Outcome measures for studying patients with low back pain. Spine (Phila Pa 1976). 1994;19(18 Suppl):2032S–6S. 10.1097/00007632-199409151-00003. [DOI] [PubMed] [Google Scholar]
- 22.Deyo RA, Bryan M, Comstock BA, et al. Trajectories of symptoms and function in older adults with low back disorders. Spine (Phila Pa 1976). 2015;40(17):1352–62. 10.1097/BRS.0000000000000975. [DOI] [PubMed] [Google Scholar]
- 23.Suri P, Stanaway IB, Zhang Y, et al. Genome-wide association studies of low back pain and lumbar spinal disorders using electronic health record data identify a locus associated with lumbar spinal stenosis. Pain. 2021;162(8):2263–72. 10.1097/j.pain.0000000000002221. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kazberouk A, Martin BI, Stevens JP, McGuire KJ. Validation of an administrative coding algorithm for classifying surgical indication and operative features of spine surgery. Spine (Phila Pa 1976). 2015;40(2):114–20. 10.1097/BRS.0000000000000682. [DOI] [PubMed] [Google Scholar]
- 25.Mayhew M, DeBar LL, Deyo RA, et al. Development and Assessment of a Crosswalk Between ICD-9-CM and ICD-10-CM to Identify Patients with Common Pain Conditions. J Pain. 2019;20(12):1429–45. 10.1016/j.jpain.2019.05.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.National Bureau of Economic Research. ICD-9-CM to and from ICD-10-CM and ICD-10-PCS Crosswalk or General Equivalence Mappings. 2016. [cited 2024 October 1]; Available from: https://www.nber.org/research/data/icd-9-cm-and-icd-10-cm-and-icd-10-pcs-crosswalk-or-general-equivalence-mappings.
- 27.Bonilla-Jaime H, Sanchez-Salcedo JA, Estevez-Cabrera MM, Molina-Jimenez T, Cortes-Altamirano JL, Alfaro-Rodriguez A. Depression and Pain: Use of Antidepressants. Curr Neuropharmacol. 2022;20(2):384–402. 10.2174/1570159X19666210609161447. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Grauer JS, Cramer JD. Association of State-Imposed Restrictions on Gabapentin with Changes in Prescribing in Medicare. J Gen Intern Med. 2022;37(14):3630–7. 10.1007/s11606-021-07314-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Centers for Disease Control. CALCULATING TOTAL DAILY DOSE OF OPIOIDS FOR SAFER DOSAGE. 2023. [cited 2023 October 1]; Available from: https://www.cdc.gov/opioids/providers/prescribing/pdf/calculating-total-daily-dose.pdf.
- 30.Xu KY, Hartz SM, Borodovsky JT, Bierut LJ, Grucza RA. Association Between Benzodiazepine Use With or Without Opioid Use and All-Cause Mortality in the United States, 1999-2015. JAMA Netw Open. 2020;3(12):e2028557. 10.1001/jamanetworkopen.2020.28557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Bockbrader HN, Wesche D, Miller R, Chapel S, Janiczek N, Burger P. A comparison of the pharmacokinetics and pharmacodynamics of pregabalin and gabapentin. Clin Pharmacokinet. 2010;49(10):661–9. 10.2165/11536200-000000000-00000. [DOI] [PubMed] [Google Scholar]
- 32.Gold LS, P. Gabapentin in CMS Patients. Open Science Foundation; 2023. [cited 2024 January 8]; Available from: https://osf.io/9u6re/. [Google Scholar]
- 33.Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. Journal of the American statistical association. 1999;94(446):496–509. [Google Scholar]
- 34.Parsons L. Performing a 1:N Case-Control Match on Propensity Score. [cited 2024 December 17]; Available from: https://support.sas.com/resources/papers/proceedings/proceedings/sugi29/165-29.pdf. [Google Scholar]
- 35.Hahn J, Jo Y, Yoo SH, Shin J, Yu YM, Ah YM. Risk of major adverse events associated with gabapentinoid and opioid combination therapy: A systematic review and meta-analysis. Front Pharmacol. 2022;13:1009950. 10.3389/fphar.2022.1009950. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Peckham AM, Fairman KA, Sclar DA. All-Cause and Drug-Related Medical Events Associated with Overuse of Gabapentin and/or Opioid Medications: A Retrospective Cohort Analysis of a Commercially Insured US Population. Drug Saf. 2018;41(2):213–28. 10.1007/s40264-017-0595-1. [DOI] [PubMed] [Google Scholar]
- 37.Bykov K, Bateman BT, Franklin JM, Vine SM, Patorno E. Association of Gabapentinoids With the Risk of Opioid-Related Adverse Events in Surgical Patients in the United States. JAMA Netw Open. 2020;3(12):e2031647. 10.1001/jamanetworkopen.2020.31647. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Lakkad M, Martin B, Li C, Harrington S, Dayer L, Painter JT. The use of gabapentinoids and opioids and risk of developing opioid-induced respiratory depression among older breast cancer survivors with neuropathic pain. J Cancer Surviv. 2024;18(3):917–27. 10.1007/s11764-023-01338-9. [DOI] [PubMed] [Google Scholar]
- 39.Sherry TB, Sabety A, Maestas N. Documented Pain Diagnoses in Adults Prescribed Opioids: Results From the National Ambulatory Medical Care Survey, 2006-2015. Ann Intern Med. 2018;169(12):892–4. 10.7326/M18-0644. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Dieleman JL, Baral R, Birger M, et al. US Spending on Personal Health Care and Public Health, 1996-2013. JAMA. 2016;316(24):2627–46. 10.1001/jama.2016.16885. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.