Concurrent Gabapentin and Opioid Use and Risk of Mortality in Medicare Recipients with Non-Cancer Pain

Abstract

Gabapentin is prescribed for pain and is perceived as safe generally. However, gabapentin can cause respiratory depression, exacerbated by concomitant central nervous system depressants (e.g., opioids), a concern for vulnerable populations. We compared mortality rates among new users of either gabapentin or duloxetine with or without concurrent opioids in the 20% Medicare sample. We conducted a new-user design retrospective cohort study, in Medicare enrollees ages 65–89 with non-cancer chronic pain and no severe illness who filled prescriptions between 2015–2018 for gabapentin (n=233,060) or duloxetine (n=34,009). Daily opioid doses, estimated in morphine milligram equivalents (MME), were classified into none, low (0< MME< 50), and high (≥50 MME), based on Centers for Disease Control and Prevention (CDC) recommendations. The outcomes were all-cause mortality (primary) and out-of-hospital mortality (secondary). We used inverse probability of treatment weighting to adjust for differences between gabapentin and duloxetine users. During 116,707 person-years of follow-up, 1,379 patients died. All-cause mortality rate in gabapentin users was 12.16/1,000 person-years versus 9.94/1,000 in duloxetine users. Risks were similar for users with no concurrent opioids (aHR=1.03, 95%CI: 0.80, 1.31) or low-dose daily opioids (aHR=1.06, 95%CI: 0.63, 1.76). However, gabapentin users receiving concurrent high-dose daily opioids had an increased rate of all-cause mortality compared to duloxetine users on high-dose opioids (aHR=2.03, 95%CI: 1.19, 3.46). Out-of-hospital mortality yielded similar results. In this retrospective cohort study of Medicare beneficiaries, concurrent use of high-dose opioids and gabapentin was associated with a higher all-cause mortality risk than that for concurrent use of high-dose opioids and duloxetine.

Introduction

Gabapentin was the 10^th most commonly prescribed drug in the U.S. in 2020, with over 49 million prescriptions filled by more than 10 million patients¹. An anticonvulsant, gabapentin is indicated for treating epilepsy and postherpetic neuralgia², though it is prescribed for many off-label uses in the U.S., including fibromyalgia and neuropathic pain³. Systematic reviews of randomized clinical trials have shown that gabapentin can improve chronic pain due to neuropathic conditions such as diabetic peripheral neuropathy and postherpetic neuralgia^{4, 5, 6}.

Gabapentin prescribing has steadily increased over the years, nearly doubling from 2009 to 2016⁷ associated with off label promotion. In addition, amidst the opioid epidemic in the U.S., the Centers for Disease Control and Prevention (CDC) released a set of guidelines⁸ that encouraged providers to prescribe alternative non-opioid analgesics to patients suffering from chronic pain. While gabapentin is commonly used for the management of pain, its safety profile has received increased scrutiny. Several studies have demonstrated potential neurologic adverse effects^{4, 9, 10, 11, 12} as well as increasing evidence of abuse and illicit use^{13 14 15 16 17 18}.

Respiratory depression, which can be life-threatening, is also a major safety concern with gabapentin use. In 2019, the FDA warned that the use of gabapentinoids, including gabapentin, may lead to serious breathing difficulties among individuals with respiratory risk factors¹⁹. These respiratory risk factors include the use of opioids or other medications that depress the central nervous system¹⁹. Between 2012 and 2018, 12 deaths due to respiratory depression were reported to the FDA Adverse Event Reporting System among individuals who took gabapentinoids and had at least one risk factor¹⁹. Respiratory depression among gabapentin users have been reported in case reports^{20 21 22}, observational studies^{23 24 25}, and a clinical trial²⁶, however, none of these studies have examined concurrent opioid and gabapentin use in a senior population. Before this FDA warning, concurrent gabapentinoid and opioid use had become increasingly prevalent among older adults with chronic noncancer pain, increasing from 17.0% in 2011 to 23.5% in 2018²⁷. Since this FDA warning, the American Geriatrics Society has advised against this potentially lethal combination among seniors in their 2019 updated version of the Beers Criteria^® for Potentially Inappropriate Medication Use in Older Adults²⁸.

Concurrent gabapentin and opioid use remains frequent in the elderly population who suffer from noncancer chronic pain, but the extent to which the opioids affect the risk of mortality is unclear.²⁵ We conducted a retrospective cohort study to compare mortality among new users of gabapentin or of a control drug used for pain, duloxetine, that has not been associated with respiratory depression,²⁹ with and without concurrent use of opioids among Medicare beneficiaries aged 65 and older with noncancer chronic pain.

Materials and Methods:

Data source:

We used a random 20% sample of Medicare administrative claims data from 2014–2018 to assemble the study cohort. Medicare is a federal health insurance program that provides insurance coverage to people aged 65 or older and to patients with end-stage renal disease or a disability³⁰. In 2019, 86% of the 61.5 million Americans enrolled in Medicare were aged 65 or older³¹. The Medicare claims data used for this study included the following files: beneficiary summary file (enrollment and demographic information), Part D events file (filled prescription data), outpatient standard analytic file (outpatient claims by institutional providers including hospitals, home health agencies, hospices, and skilled nursing facilities), Medicare Carrier file (outpatient claims by non-institutional providers including physicians, nonphysicians, specialists, and suppliers), MEDPAR file (inpatient and skilled nursing facility data), and the hospice file.

Study Population:

This study included Medicare beneficiaries aged 65 to 89 with an indication for non-cancer chronic pain (see Table S1) who filled a prescription for gabapentin or the active control, duloxetine between January 1, 2015, and December 30, 2018. Given the concern over inaccuracy and lack of completeness in Medicare Advantage (also known as Medicare Part C) data³², study inclusion was limited to individuals who had Medicare Fee-For-Service (also known as Medicare Parts A & B) coverage in addition to Medicare Part D coverage which contains prescription drug data. Study participants had to have one year of Medicare Parts A, B, and D coverage before the initial study drug fill, during which they must have filled one prescription and filed one other claim. Beneficiaries aged 90 and older were excluded from our study given that the “Safe Harbor” method of de-identification³³ requires ages above 89 to be collapsed to 90 and above. Individuals with active cancer or other serious illnesses (see Table S2, which lists the codes that were used to define these illnesses) were excluded from this study. New users of gabapentin and duloxetine were defined as beneficiaries who had no record of the use of either study drug or any medication in the same drug class in the year before the initial study drug fill (Table S3). Individuals who were in the hospital on the initial study drug prescription fill date as well as individuals who had evidence of hospice or nursing home stay or more than 30 total days of hospitalization in the year before the initial study drug prescription fill were excluded. We did quality control of the prescriptions in this study to include only prescriptions with a capsule or tablet formation that 1) did not exceed the recommended maximum daily dose of the drug (3600 mg/day for gabapentin and 120 mg/day for duloxetine), 2) did not have a day’s supply greater than 90, and 3) had a quantity and day supply equal to non-zero integers (see Figure S1, which illustrates the study’s inclusion and exclusion criteria).

Follow-up:

Follow-up for this study began the day after the fill date of a patient’s first prescription of either gabapentin or duloxetine. We limited follow-up to days that the individual was likely currently taking the study drug, calculated from prescription days of supply. Study follow-up ended on the first date of the following: (1) the last day of Medicare enrollment in parts A, B, and D (not C); (2) 30 consecutive hospitalization days, hospice admission, or 30 cumulative days in a nursing home; (3) gap of 180 days between prescriptions; (4) prescription for the other study drug; (5) patient turns 90 years old; (6) study endpoint; (7) end of the study, December 31, 2018. Patients could not re-enter the study.

Outcome:

The primary study outcome was all-cause mortality, defined as any death that occurred during study follow-up. The secondary outcome was out-of-hospital mortality which was defined as a death that occurred on a date during which the patient was not hospitalized for at least 2 consecutive inpatient days including admission and discharge dates. Both the primary and secondary outcomes identified death dates using the Medicare beneficiary summary file.

Covariates:

Covariates included in this study were selected for inclusion in the propensity score based on past studies and literature review (see Table S4, which lists covariates by category). These 335 covariates included demographic variables, history of opioid use, chronic pain indication, other pain-related medications, cardiovascular diagnoses and medication use, medical care utilization, musculoskeletal diseases, psychiatric diagnoses, and other medication use, as well as other comorbidities and medications.

Concurrent Opioid Use:

To consider the extent to which concurrent opioid use may affect the potential association between gabapentin use and mortality, we calculated the daily morphine milligram equivalents (MME) based on opioid prescriptions. MME was calculated by multiplying the dose of an opioid by a conversion factor³⁴. Daily MME was calculated by summing the total MME for each opioid record for that day. We grouped MME status into no, low, and high dose categories. The CDC recommends that clinicians prescribing opioids for chronic pain “carefully reassess” prescribing ≥ 50 MME/day to individuals as it can increase the risk of overdose⁷. Based on this recommendation, we defined high MME status as ≥50 MME/day, low MME as 0<MME<50/day, and no MME as 0 MME/day.

Statistical Analysis:

The primary analysis of this study was Cox proportional hazards regressions of the time-dependent inverse probability of treatment weighted (IPTW) propensity score sample, corrected for weighting-induced dependencies with Huber White Sandwich estimates^{35 36} and stratified by concurrent MME category. We also performed a sensitivity analysis using propensity score matching based on this propensity score.

We used propensity scores, defined as the probability of treatment assignment conditional on the observed covariates³⁷, for adjustment. Since we were interested in both the differences in the distribution of covariates between the study drugs at the time of each prescription and how concurrent MME dosage at the time of each prescription affected these differences, we calculated propensity scores stratified by concurrent MME use category on the first day of each prescription of gabapentin or duloxetine. We trimmed extreme values (above the 99^th percentile and below the 1^st percentile) to avoid including prescriptions where treatment was effectively predetermined by these characteristics. Calculating propensity score at the time of prescription, similar to models in previous literature³⁸, allows us to account for time-dependent covariates and achieve a balanced distribution of the covariates in the treatment and active control groups at each prescription³⁹. Upon developing the propensity score, we compared the standardized difference of each covariate to the generally accepted threshold of 10% to determine if the propensity score was successful in balancing the covariate between the weighted study drug groups⁴⁰. An IPT weighted table of real or potential confounding variables is given in Table S6

To account for the concurrent MME category as a time-varying covariate in our stratified Cox models while also using the propensity scores to weight by the inverse probability of treatment, we summed the consecutive days within each prescription where an individual fell into the same concurrent MME category. We then used the total of each of these numbers of consecutive days in each MME category as our unit of time measurement in our Cox proportional hazards regression models stratified by MME category group. Using this method, the baseline hazard function was reset to 0 after every MME group change within a prescription.

A secondary analysis of this study was conducted where the propensity score was calculated based on concurrent MME category at the time of the first prescription, rather than at each prescription. The remainder of this secondary analysis followed the methods highlighted in our primary analysis. This secondary analysis was intended to mimic an intention-to-treat analysis, wherein in a randomized controlled trial, participants are analyzed based on the group to which they were originally assigned. By calculating the propensity score by stratum of concurrent MME at the time of prescription, we were able to study how concurrent MME category at the time of initial prescription fill impacted our results.

All analyses were performed using STATA 16.0 (StataCorp LP, College Station, TX)⁴¹. All calculated p-values are two-sided. Under the Center for Medicare & Medicaid Services (CMS) policy, we do not report data on fewer than 11 individuals in any group.

A waiver of informed consent was granted by the Vanderbilt University Medical Center Institutional Review Board (IRB#182117).

Results

Patient Characteristics:

This cohort of Medicare beneficiaries aged 65 and older with non-cancer chronic pain included 233,060 new users of gabapentin and 34,009 new users of duloxetine. Beneficiaries were predominately female (65.8%) and non-Hispanic White (80.2%). Back pain and other musculoskeletal conditions were the most common indications of chronic pain. Gabapentin and duloxetine users had a similar prevalence of many chronic pain diagnoses (Table 1)⁴², though the ones that differed at baseline were related to specific drug indications. For example, a higher prevalence of neuropathic pain was found among gabapentin users, an expected result given gabapentin is commonly prescribed to treat this condition. Similarly, fibromyalgia was more prevalent among duloxetine users as it is one of its major indications. At baseline, a history of long-acting opioid use was more prevalent among duloxetine users along with higher median days of short-acting opioid use in the past year (see Table S5, which is an extended version of Table 1).

Table 1.

Selected Baseline Characteristics for New Users of Gabapentin or Duloxetine

Variable	Gabapentin (N=233,060)	Duloxetine (N=34,009)	Standardized Difference
	N (%)	N (%)
Demographics
Age, median (IQR), years	73 [69,79]	73 [69,78]	−0.07
Male	82619 (35.45)	8646 (25.42)	0.22
Race, Non-Hispanic White	184,729 (79.26)	29,530 (86.83)	−0.2
Race, Black (or African-American)	18,225 (7.82)	1,625 (4.78)	0.13
Race, Other than White or Black	30,106 (12.92)	2,854 (8.39)	0.15
Chronic Pain Indicators
Fibromyalgia	28546 (12.25)	7649 (22.49)	−0.27
Neuropathic pain	141938 (60.90)	14734 (43.32)	0.36
Back pain/degenerative back disorders	156719 (67.24)	22686 (66.71)	0.01
Arthralgia	110842 (47.56)	17331 (50.96)	−0.07
Inflammatory pain	51433 (22.07)	8384 (24.65)	−0.06
Other musculoskeletal/soft tissue pain	165374 (70.96)	25923 (76.22)	−0.12
Opioids
Long-Acting Opioid Analgesics Use	5359 (2.30)	1535 (4.51)	−0.12
Short-Acting Opioid Analgesics Use	118933 (51.03)	18670 (54.90)	−0.08
Opioid Replacement Therapy Use	1000 (0.43)	352 (1.04)	−0.07
Daily long-acting opioid MME dose (average), median (IQR)	58.5 [30,90]	60 [30,110.37]	0.11
Daily short-acting opioid MME dose (average), median (IQR)	30 [20,56.77]	33.45 [20,60]	0.08
Daily methadone MME dose (average), median (IQR)	87.85 [45,90]	86.33 [45,90]	−0.04
Non-opioid pain drugs
NSAIDs, non-selective	81435 (34.94)	11580 (34.05)	0.02
Systemic oral corticosteroids	71215 (30.56)	10233 (30.09)	0.01
Other skeletal muscle relaxants	21891 (9.39)	3508 (10.31)	−0.03
Analgesics, other	4728 (2.03)	769 (2.26)	−0.02
Cardiovascular diagnoses
Diabetes (neuropathy)	28636 (12.29)	2850 (8.38)	0.13
Hypertension, benign or unspecified	187776 (80.57)	26809 (78.83)	0.04
Hyperlipidemia	174656 (74.94)	25070 (73.72)	0.03
Myocardial infarction	11060 (4.75)	1400 (4.12)	0.03
Respiratory diagnoses and drugs
Chronic obstructive pulmonary disease	32094 (13.77)	5065 (14.89)	−0.03
Asthma	23585 (10.12)	3894 (11.45)	−0.04
Beta-agonists	36717 (15.75)	5754 (16.92)	−0.03
Bronchodilators, other	17341 (7.44)	2708 (7.96)	−0.02
Inhaled corticosteroids	48880 (20.97)	7652 (22.50)	−0.04

The differences between these study drug groups at baseline and at the time of each prescription fill were accounted for using propensity scores in that the prescriptions included in the analysis had comparable characteristics as indicated by the standardized differences (see Tables S6, S7, and S8) which were <0.10 which is considered by many to be negligible⁴⁰.

Primary Analysis:

During 116,707 person-years of follow-up, 1,379 total deaths occurred in the cohort. The all-cause mortality rate among gabapentin users was 12.16 per 1,000 person-years compared to 9.94 per 1,000 person-years in duloxetine users. The unadjusted incidence rates between study drug groups were comparable by concurrent MME stratum (Table 2). However, when adjusting for study covariates, the risk of all-cause mortality among gabapentin users on high doses of opioids was more than double that of duloxetine users on high doses of opioids, HR=2.03 (95% CI: 1.19, 3.46) (Table 2).

Table 2.

Primary Analysis: Stratified Propensity Score by MME Category at Each Prescription Fill

Outcome	Concurrent MME Category	Gabapentin		Duloxetine		Unadjusted Hazard Ratio (95% CI)	Adjusted* Hazard Ratio (95% CI)
		Unadjusted Rate (per 1000 person-years)	Adjusted* Rate (per 1000 person-years)	Unadjusted Rate (per 1000 person-years)	Adjusted* Rate (per 1000 person-years)
All-Cause Mortality	No MME (MME = 0)	11.15	10.93	8.97	10.75	1.25 (1.05, 1.50)	1.03 (0.80, 1.31)
	Low MME (MME >0 & MME < 50)	16.94	17.1	14.17	16.15	1.20 (0.81, 1.77)	1.06 (0.63, 1.76)
	High MME (MME ≥ 50)	15.47	15.61	14.45	7.8	1.09 (0.67, 1.78)	2.03 (1.19, 3.46)^
Out of Hospital Mortality	No MME (MME = 0)	5.89	5.83	3.90	4.55	1.50 (1.15, 1.98)	1.28 (0.89, 1.84)
	Low MME (MME >0 & MME < 50)	10.62	10.47	8.36	12.78	1.27 (0.77, 2.11)	0.82 (0.44, 1.53)
	High MME (MME ≥ 50)	9.78	9.84	9.95	5.31	0.99 (0.55, 1.80)	1.86 (0.96, 3.62)

CI=confidence interval, MME= Morphine Milligram Equivalents

^*:Propensity score-adjusted using inverse probability of treatment weighting

Duloxetine is the reference for all Hazard Ratios

^{^}p<0.05

The overall results of our analysis using out-of-hospital mortality as the outcome were consistent with those found in the all-cause mortality analysis. When analyzing out-of-hospital deaths as the outcome, 748 deaths occurred during 116,086 person-years of follow-up. Between the study drug groups, unadjusted incidence rates were comparable by concurrent MME stratum (Table 2). The risk of out-of-hospital mortality, after adjustment for study covariates, among gabapentin users taking high doses of opioids was nearly double that of duloxetine users taking high doses of opioids, though this did not achieve statistical significance, HR=1.86 (95% CI: 0.96, 3.62) (Table 2).

Secondary and Sensitivity Analysis:

The prespecified secondary analysis of this study calculated propensity score by MME category strata at the time of initial prescription. After trimming and inverse probability weighting prescriptions based on stratum-specific propensity scores, the prescriptions in this analysis were similar in terms of diagnoses and medications by study drug group, as indicated by the standardized differences (Tables S9, S10, and S11 demonstrate these characteristics based on no MME, low MME, and high MME, respectively) of these characteristics which were <0.10, indicating good balance⁴⁰. The results of this secondary analysis are consistent with that of the primary analysis which demonstrates that gabapentin users concurrently taking high doses of opioids were more than twice as likely to experience all-cause mortality than duloxetine users on high doses of opioids. In our secondary analysis gabapentin users concurrently taking high doses of opioids were also more than twice as likely to experience out-of-hospital mortality than duloxetine users on high doses of opioids (Table 3).

Table 3.

Secondary Analysis: Propensity Score Stratified by MME on First Prescription Fill

Outcome	Concurrent MME Category	Gabapentin		Duloxetine		Unadjusted Hazard Ratio (95% CI)	Adjusted* Hazard Ratio (95% CI)
		Unadjusted Rate (per 1000 person-years)	Adjusted* Rate (per 1000 person-years)	Unadjusted Rate (per 1000 person-years)	Adjusted* Rate (per 1000 person-years)
All-Cause Mortality	No MME (MME = 0)	11.14	10.92	8.97	11.14	1.25 (1.04, 1.50)	0.99 (0.76, 1.28)
	Low MME (MME >0 & MME < 50)	17.06	17.25	14.17	17.16	1.21 (0.82, 1.78)	1.00 (0.60, 1.69)
	High MME (MME ≥ 50)	15.82	16.26	14.45	6.92	1.11 (0.68, 1.81)	2.39 (1.41, 4.06)^
Out of Hospital Mortality	No MME (MME = 0)	5.85	5.79	3.90	4.98	1.49 (1.14, 1.96)	1.16 (0.78, 1.73)
	Low MME (MME >0 & MME < 50)	10.65	10.68	8.36	13.59	1.27 (0.77, 2.11)	0.78 (0.41, 1.48)
	High MME (MME ≥ 50)	10.12	10.40	9.95	4.32	1.02 (0.56, 1.85)	2.42 (1.28, 4.58)^

CI=confidence interval, MME= Morphine Milligram Equivalents

^*:Propensity score-adjusted using inverse probability of treatment weighting

Duloxetine is the reference for all Hazard Ratios

^{^}p<0.05

The results of our prespecified sensitivity analyses using 3:1 matching based on propensity score calculated by concurrent MME stratum at the time of prescription are found in Table 4. The 3:1 gabapentin:duloxetine matched analysis resulted in a fewer number of both mortality outcomes across both study drugs and although the matched analyses were underpowered, we generally see an increased risk of mortality outcomes among gabapentin users on high doses of opioids.

Table 4.

Sensitivity Analysis: 3:1 (Gabapentin:Duloxetine) Matched Analysis

		Gabapentin (N Prescriptions: 288,531)			Duloxetine (N Prescriptions: 91,177)			Cox Proportional Hazards Analysis
Outcome	Concurrent MME	N Events	Person-Years	Rate per 1,000 person-years	N Events	Person-Years	Rate per 1,000 person-years	Hazard Ratio (95% CI)
All-Cause Mortality	No MME (MME = 0)	243	22505	10.80	83	8051	10.31	1.06 (0.83, 1.36)
	Low MME (MME >0 & MME < 50)	60	3698	16.22	20	1289	15.52	1.05 (0.63, 1.74)
	High MME (MME ≥ 50)	25	2091	11.96	<11	744	9.40	1.30 (0.56, 3.01)
Out of Hospital Mortality	No MME (MME = 0)	127	22402	5.67	34	8014	4.24	1.33 (0.91, 1.95)
	Low MME (MME >0 & MME < 50)	40	3671	10.90	13	1281	10.15	1.07 (0.57, 2.00)
	High MME (MME ≥ 50)	20	2076	9.63	<11	739	5.41	1.82 (0.62, 5.34)

CI=confidence interval, MME= Morphine Milligram Equivalents

Duloxetine is the reference for all Hazard Ratios

Discussion

This study found that the risk of all-cause mortality was higher among Medicare beneficiaries with non-cancer chronic pain who took gabapentin with high doses of opioids than the risk observed in duloxetine users taking high doses of opioids. Results of out-of-hospital mortality only reached statistical significance in our secondary analysis.

To our knowledge, this is the first study to assess the extent to which concurrent opioid dosage may impact the risk of mortality among new users of gabapentin who are enrolled in Medicare and have a diagnosis of non-cancer chronic pain. Our findings warn of the clinical consequences of the interaction between the concurrent use of gabapentin and high doses of opioids. Not only are results of our study biologically plausible, they are consistent with the FDA warning against the combination of gabapentinoids and opioids¹⁹ as well as the CDC recommendation to reassess prescribing ≥ 50 MME/day of opioids as it can increase the risk of respiratory depression⁸.

Gomes et al. conducted a nested case-control study among opioid users and reported the association between concurrent use of opioids and gabapentin with increased opioid-related mortality when compared to opioid use alone. Most of their patients (~94%) were less than 65 years of age.²⁵ In contrast to this report, our study has several distinct characteristics that expand upon the findings of Gomes et al. First, we conducted a retrospective cohort study among Medicare beneficiaries aged 65 and older with noncancer chronic pain – an important study population given their increased vulnerability.⁴³ Second, we used an active comparator drug, as we compared mortality associated with the concurrent use of opioids and gabapentin with the mortality associated with the active comparator duloxetine, used concurrently with opioids. Third, our study focused on differences based on escalating opioid dose whereas their study examined escalating gabapentin dose. Finally, to increase the sensitivity of death due to respiratory depression, our outcomes were all-cause and out-of-hospital mortality rather than opioid-related deaths.

The strengths of this study include a new-user study design, a large number of covariates considered in a well-balanced propensity score, and an active comparator with a similar indication. We also included a large population of individuals who are often excluded from clinical trials^{44 45} due to comorbidities but who suffer from chronic pain disproportionately compared to the rest of the U.S. population. Additionally, our dataset allowed us to observe the daily change in opioid MME dosage to account for daily MME dose rather than assuming that the opioid dose remained the same throughout a prescription.

There are some limitations of this study. First, because it relied on claims data to assess prescription drug use, it is possible that individuals may not have taken the study drug as prescribed; however, we would expect that if this were to have occurred, it would impact both study drug groups at similar rates, resulting in nondifferential bias. Second, our study may not be generalizable to all Medicare beneficiaries as it excluded individuals with Medicare Advantage (Medicare Part C) due to the concerns regarding lack of completeness and inaccuracy of the data³². Third, though we included over 300 covariates in our propensity score calculation, covariates may have gone unmeasured in our study, including ones that are not available in Medicare files. Fourth, while our results are significant for all-cause mortality, they do not reach statistical significance for out of hospital mortality. A plausible explanation is the fewer number of events with less power to detect differences. While we were able to define mortality based on death dates from the Medicare beneficiary summary file, we lacked details on the mode of death which could have been helpful to propose a mechanism of death. Additionally, our study excluded prescriptions where an individual was initially prescribed more than the maximum daily dose for either of the study drugs, though, we did not examine the extent to which the dose of gabapentin may have impacted the results when combined with the dose category of MME. It is possible that non-prescription opioid use, which is not captured in Medicare data, occurs in this population, however, it is more likely that the problem in this age group is that elderly individuals could misuse their prescribed opioids⁴⁶. Lastly, our study did not aim to address the effectiveness of gabapentin or duloxetine and concurrent opioid use, a concern that should be addressed in future research⁵.

In conclusion, the risk of all-cause mortality among Medicare beneficiaries aged 65 or older with non-cancer chronic pain who were new users of gabapentin and took high doses of opioids was more than double that of new duloxetine users who took high doses of opioids.

Study Highlights.

What is the current knowledge on the topic?

The FDA has warned that concurrent gabapentin and opioid use may lead to respiratory depression.

What question did this study address?

This study addresses the extent to which concurrent opioid use among new-users of gabapentin is associated with increased risk of mortality compared to concurrent use with an active control that has not been associated with respiratory depression.

What does this study add to our knowledge?

This study warns of a potential interaction between concurrent use of gabapentin and high doses of opioids that can lead to all-cause mortality.

How might this change clinical pharmacology or translational science?

Gabapentin and opioids are often co-prescribed to individuals suffering from pain. The results of our study indicate that clinicians should reconsider prescribing high doses of opioids concurrently with gabapentin.