Benzodiazepine Use and Dependence in Relation to Chronic Pain Intensity and Pain Catastrophizing

Abstract

Benzodiazepines (BZDs), a class of sedative-hypnotic medications, generated concern as their popularity grew, with particular alarm regarding elevated rates of BZD use among chronic pain populations. Consistent with negative reinforcement/motivational models of substance use, desire for pain alleviation may motivate BZD use. Yet, little is known about relations between pain and addiction-relevant BZD use processes. This cross-sectional survey study aimed to: a) test associations between pain intensity and clinically relevant BZD use patterns, and b) examine the role of pain catastrophizing in hypothesized pain-BZD relations. Participants included 306 adults with chronic musculoskeletal pain and a current BZD prescription who completed an online survey study (M_age = 38.7, 38.9% female). Results indicated that pain intensity was positively associated with past-month BZD use frequency, BZD dependence severity, and likelihood of endorsing BZD misuse behaviors (ps < .05). Pain catastrophizing was positively associated with BZD dependence/likelihood of BZD misuse, covarying for pain intensity (p < .05). These findings build upon an emerging literature by highlighting positive covariation of pain intensity and pain catastrophizing with addiction-relevant BZD use behaviors. Results underscore the need to further investigate high-risk BZD use among individuals with chronic pain, with and without concurrent opioid use, to inform prevention/intervention efforts.

Perspective:

This article presents findings on cross-sectional associations of pain intensity and pain catastrophizing with clinically relevant benzodiazepine (BZD) use outcomes, including dependence and misuse, among individuals with chronic pain. Findings help elucidate the higher burden of BZD misuse/dependence in chronic pain populations and suggest that pain relief may be a common, yet underrecognized, self-reported motivation for taking BZDs.

Benzodiazepines (BZDs) are sedative-hypnotic drugs³¹ and one of the most commonly prescribed medication classes, with up to 13% of adults in the United States reporting past-year BZD use.^{9, 56, 69} While insomnia and anxiety remain the most common indications for BZD prescription,³⁴ clinical guidelines now advise only short-term use for insomnia⁷⁷ and recommend alternative first-line treatments for anxiety,⁴ with changes to guidelines driven by growing recognition of the risks and abuse potential of BZDs.^{1, 94}

Addictions experts have recognized a recent period of growth in BZD use and associated harms,^{49, 51, 76} marked by quadrupling of overdose deaths from the mid-1990s to 2013.³ During that approximate period, physician visits involving BZD prescription doubled¹, refill prescription rates increased,⁴² and average dosages tripled.³

Researchers have highlighted the need to explicate biopsychosocial factors contributing to the maintenance/escalation of BZD use among groups at elevated risk of deleterious BZD-related outcomes, including chronic pain populations.^{34, 73} The ~50 million U.S. adults with chronic pain (i.e., pain persisting beyond expected healing time or > 3 months)^{40, 90} may experience higher risk of BZD-related problems for several reasons. First, chronic pain patients may be 2–3 times as likely as the general population to report BZD use (over 30% versus 5 – 15%)^{1, 18, 67} and to meet criteria for BZD-related substance use disorder (SUD).⁵⁰ Among chronic pain samples, greater pain intensity has been associated with greater likelihood of endorsing past-year,⁸⁹ past-month,⁶⁷ and current¹⁸ BZD use. Second, although BZDs are frequently prescribed in pain management contexts,^{1, 21, 43} we lack evidence for BZD efficacy in most pain conditions,⁹⁸ and use of BZDs to manage chronic pain is thought to promote long-term, higher-risk use.⁷³ Third, co-use of BZDs with opioid analgesics is prevalent³³ and associated with higher overdose fatality¹⁹ and all-cause mortality¹⁰⁰ rates than use of opioids alone.

To address the disproportionate, negative impact that BZD use may have on chronic pain populations, critical next steps include examining relations between pain and granular, clinically relevant BZD use patterns. Such patterns include use frequency,^{49, 59, 94} indices of BZD misuse (i.e., use without a prescription or otherwise not as directed by a physician),^{57, 95} and use consistent with BZD-related SUDs.⁷⁹ Among adults reporting any past-year BZD use, 17% endorse BZD misuse and 1.5% meet criteria for a BZD-related SUD.⁹ Misuse encompasses behaviors that increase risk for overdose and death, including consuming BZDs with contraindicated substances (e.g., alcohol, opioids)^{41, 94} or at larger doses than prescribed. BZD-related SUD^{2, 9} is characterized by clinically significant distress and/or impairment along with other (e.g., dose escalation).

Future development of interventions for individuals using BZDs in the context of chronic pain may be aided by identification of modifiable cognitive-affective transdiagnostic factors (i.e., underlying factors common to two or more disorders)⁴⁶ involved in both pain experience and BZD use.²³ One leading candidate is pain catastrophizing, understood as a set of exaggerated, negative, cognitive-affective response styles (e.g., rumination) related to pain.^{25, 74} Empirically, pain catastrophizing has been implicated in pain-driven urge to use tobacco,⁴⁵ use of cannabis for pain-coping,⁸² opioid misuse,^{53, 54} and emerging evidence links pain catastrophizing to greater likelihood of BZD use.¹⁸

The goal of the current study was to test associations of pain intensity with BZD misuse, severity of BZD dependence, and past-month BZD use frequency among individuals reporting a current BZD prescription and chronic musculoskeletal pain. The secondary goal was to examine the role of pain catastrophizing as a potential moderator of hypothesized pain-BZD use relations. A final goal was to conduct preliminary analyses among participants co-prescribed opioids, given the potential for distinct experiences and elevated BZD-related risks in this subgroup.¹⁹

Method

Participants and Recruitment

Potential participants were screened for the following inclusion criteria: ≥ 18 years of age, current prescription for one or more BZD medications (i.e., prescription sedatives or hypnotics),⁷⁵ current chronic musculoskeletal pain, residence in the United States, and ability to read English. Participants were recruited during March and April 2021 using the crowdsourcing platform Amazon Mechanical Turk (mTurk), a growing source of health sciences and addictions research data.⁸⁶ A human intelligence task (HIT) was posted briefly describing survey content (“You will be asked to answer questions about your health and behaviors.”), duration, and compensation. Consistent with prior work, access was limited to respondents with U.S. residence.^{17, 84, 85} Informed consent was obtained from all participants who were eligible and interested. Participants then completed a brief (~25-minute) survey. All participants who completed the survey were compensated regardless of subsequent data exclusions. Consistent with prior work^{7, 14, 47, 83} and estimated survey completion time, survey respondents were compensated $3.00–4.00 (compensation was raised due to concerns about recruitment feasibility). All study activities were approved by the Syracuse University Institutional Review Board.

The desired sample size was determined via power analysis using the G*Power program,²⁷ with significance criterion α = .05 and power 1 - β = 0.8. Small effect sizes (f²) have been observed previously when testing contributions of pain catastrophizing to variance in prescription drug use outcomes, accounting for pain severity.⁵⁴ Sample size calculations were thus based on a conservative assumption for small sample sizes, f² = 0.05. Based upon these assumptions and a hierarchical multiple regression (F test) model, a sample size of 262 was determined to yield power exceeding 0.95. Estimating ~10% loss, we planned to recruit 300 participants.

Steps to Promote Data Quality

Steps were taken at the screening stage to minimize deceptive responding (e.g., disingenuous endorsement of behaviors, often in order to gain access to a study).⁸⁶ Accumulating evidence^{15, 38, 78} suggests that rates of deceptive responding decrease substantially (i.e., from 89% to 5%) when eligibility criteria are made unknown to participants.⁷⁸ Thus, best practices include posting HITs that conceal the population being sampled.^{24, 97} and utilizing a separate screening questionnaire that masks eligibility requirements through inclusion of additional questions unrelated to study eligibility.^{14, 38, 78} A two-part screening process was also implemented to reduce response bias; Screener I asked participants to indicate general medication classes (selecting “sleeping medications” and/or “anxiety medications” made a participant provisionally eligible), while Screener II asked participants to select from among specific medications including BZD and non-BZD anxiolytics/hypnotics.

Additionally, inconsistency methods were used to exclude participants who failed to respond consistently when invariant items (i.e., marital status, past-year cigarette use) were administered at two separate points.^{14, 86} Improbability methods, consistent with recommendations, were used to exclude participants endorsing use of a number of substances that is considered biologically implausible (e.g., past-year non-medical use of > 35 substances).⁸ While co-prescription of 2 or more BZDs is relatively common (e.g., nearly 15% in a pain treatment sample),⁷¹ the modal number of prescriptions was 1 and median was 2 in the present sample. Based upon score distributions, participants endorsing ≥ 5 BZD prescriptions were considered to display probable disingenuous/careless responding and were excluded from analyses.

Measures

Sociodemographic Characteristics

Participants were asked to report on a range of sociodemographic characteristics, including age, sex, racial and ethnic identities, marital status, educational attainment, and household income. Given previously observed associations with both BZD use (versus non-use)^{18, 56, 69} and pain,^{28, 62} age and sex were identified as a priori covariates for all analyses.

BZD Prescription

Participants were asked to indicate which type(s) of BZD they were currently prescribed by selecting from among the BZDs assessed in the National Survey on Drug Use and Health⁷⁵: alprazolam products (Xanax, Xanax XR, Alprazolam, Extended-Release Alprazolam), lorazepam products (Ativan, Lorazepam), clonazepam products (Klonopin, Clonazepam), diazepam products (Valium, Diazepam), flurazepam (also known as Dalmane), temazepam products (Restoril, Temazepam), triazolam products (Halcion, Triazolam), estazolam products (Prosom), other benzodiazepine tranquilizers or sedatives. For participants endorsing multiple BZD prescriptions, the primary BZD was defined as that with the maximum number of past-month days of use.

BZD Dependence

The Severity of Dependence Scale (SDS) is a 5-item measure that has been validated as a tool for assessing BZD dependence.^{12, 20} Items assess: the extent to which individuals feel their BZD use is out of control, worry about missing a dose, worry about their BZD use, wish they could cease BZD use; and how difficult they feel it would be to go without BZDs. Items are rated on a 4-point scale (e.g., 0 ‘never/almost never’ to 4 ‘always/almost always’) and summed to produce a total score (range: 0 to 20), with higher total scores reflecting greater severity of BZD dependence symptoms (i.e., ‘dependence severity’). The SDS has demonstrated diagnostic utility for BZD dependence and use disorders,^{16, 20} including sensitivity of 98% and specificity of 94% relative to DSM-IV diagnostic criteria for BZD dependence (cutoff score of 6).²⁰ Internal consistency in the present sample was acceptable (Cronbach’s α = .77).

BZD Misuse

Past-12-month BZD misuse was assessed in a manner consistent with current NSDUH definitions. That is, misuse constituted a) use without own prescription, b) use in greater amounts, more often, or over a longer period than prescribed, and/or c) use in another way not as directed by a physician.⁷⁵ Participants endorsing one or more of these behaviors were considered positive for misuse.⁶⁶

BZD Use Frequency

Frequency of use was indicated by the self-reported number of days using BZDs over the past month.⁸¹ Past-month days of use has reliably been used to index current BZD use frequency^{61, 81} and identify individuals with heavier current BZD use for clinical purposes.⁸¹

Pain Characteristics

Chronic pain intensity was assessed via the Graded Chronic Pain Scale (GCPS),⁹³ a widely used and reliable measure.⁹² The characteristic pain intensity score represents the sum of three numerical rating scale items (rated 0–10) assessing current, past-3-month worst, and past-3-month average pain intensity. Pain intensity demonstrated acceptable internal consistency (α = .72). A categorical classification of pain severity can also be generated and ranges from Grade 1 (low intensity, low interference) to Grade 4 (severe interference). Participants were also asked to mark all locations with past-3-month pain and a primary pain location, from among 15 common bodily pain locations.

Pain Catastrophizing Scale

The Pain Catastrophizing Scale (PCS)⁸⁷ is a 13-item measure assessing the extent to which negative thoughts/emotions are experienced in response to pain, and validated in chronic pain populations.^{70, 87} Items are rated from 0 (‘not at all’) to 4 (‘all the time’) and summed for a total score (range: 0 to 52). The PCS has demonstrated criterion-related and predictive validity (e.g., prediction of postoperative pain).³⁰ Internal consistency was excellent (α = .92).

Other Substance Use and Clinical Characteristics

Participants were asked to report on their use of prescription opioid medications⁷⁵ and to complete the Generalized Anxiety Disorder −7 (GAD – 7),⁸⁰ a measure of generalized anxiety symptoms validated in the general population,⁵² and the Insomnia Severity Index (ISI),⁵ a measure of insomnia severity/functional impact validated in community samples.⁶³ GAD-7 and ISI internal consistency were good and acceptable, respectively (α = .82; α = .79).

Data Analytic Plan

All analyses were conducted using SPSS Version 27.³⁹ First, all variables were assessed for normality, and skewness and kurtosis values fell within an acceptable range (−2 ≤ × ≤ 2).²⁹ Second, a series of bivariate and point-biserial correlations were run to test associations between dependent variables (BZD dependence, past-month use frequency, and misuse), and participant sociodemographic characteristics. To account for the low number of participants reporting races other than White or Black (< 7%), a dichotomous white/non-white race variable was created and used in all subsequent analyses. Variables that were associated with any dependent variable were retained as covariates in addition to covariates selected a priori (age, sex).^{18, 56, 69} All regression models presented in Results are adjusted for these covariates. Third, the variance inflation factor (VIF) for each predictor was assessed in order to identify multicollinearity problems, which occur when predictor variables are highly correlated and are indicated by a VIF ≥ 10.⁶⁵ No VIF for any predictor exceeded 1.5.

Fourth, models were assessed for assumptions. Each linear regression model met assumptions: linearity of predictor-outcome relationships (assessed via scatterplots); and zero mean, equal variance, independence, and normality of residuals (assessed via residuals versus fitted values plots and residuals histograms). For the logistic regression model, tests for linearity of the relationship between predictors and log-odds of the outcome¹⁰ suggested linearity assumptions were met for all predictors except pain intensity. However, logistic regression models are considered robust to this violation of linearity at sample sizes above ~250.⁶

Fifth, separate hierarchical linear regression models were conducted to test associations between pain intensity and BZD dependence severity and past-month use frequency. Covariates were entered in the first step of each model and pain intensity in the second. Relative contributions of pain intensity to observed variance in outcome variables were assessed by examining change in R-squared statistic (ΔR²) at the second step of each model. Sixth, a hierarchical logistic regression model was conducted to test associations between pain intensity and likelihood of endorsing any past-year BZD misuse.

Seventh, we examined the interaction between pain intensity and pain catastrophizing by including an interaction term (pain intensity × pain catastrophizing) in step 3 of each model. The order of variables entered in each model was as follows: step 1 (covariates); step 2 (pain intensity, pain catastrophizing); step 3 (pain intensity × pain catastrophizing interaction). In the case of significant interaction terms, we further probed using the PROCESS Macro for SPSS.³⁶ In the case of non-significant interactions, Step 2 of the model was examined. Finally, to explore whether results differed by opioid prescription status, each model was also run among the subgroup of participants endorsing a current opioid prescription.

Results

Data Quality

A total of 2,379 individuals were screened in Qualtrics, of whom N = 390 were eligible and consented to participate. Data were excluded from analysis for any participant who failed inconsistency and/or attention checks (N = 65), did not complete measures used as covariates (N = 8), or endorsed a number of current BZD prescriptions considered likely to indicate disingenuous/inattentive responding (≥ 5 prescriptions; N = 37). Therefore, the final sample included N = 306 participants. Correlations in the expected direction based on prior literature^{37, 48, 80} were observed between pain intensity and BZD dependence severity (r = .54, p < .001), GAD - 7 scores (r = .42, p < .001), and ISI scores (r = .54, p < .001).

Participant Characteristics

The final sample included 306 adults (38.9% female; M_age = 38.7, SD = 11.1) who reported current chronic musculoskeletal pain and at least one current BZD prescription. Participant characteristics are summarized in Table 1. Respondents self-identified as predominantly White (68.6%) and Black or African American (24.5%). Over 60% of participants reported annual household income ≥ $50,000. In terms of pain, participants reported a mean of 77.6 pain days in the past six months (SD = 56.7). The mean pain intensity score was 21.1 (SD = 4.7; Range: 5 – 30). Individual pain intensity ratings were as follows: mean current pain intensity was 6.8 (SD = 2.2; Range: 0 – 10), mean past-3-month worst pain intensity was 7.4 (SD = 1.8; Range: 2 – 10), and mean past-3-month average pain intensity was 6.9 (SD = 1.9; Range 1 – 10). Approximately 90% of the sample was classified in GCPS Pain Grades III or IV, consistent with moderate to severe pain interference. Most commonly endorsed primary pain locations were: back (36.3%), head (19.6%), neck (11.1%), chest (7.2%), and lower extremities (7.1%). The mean PCS score was 30.7 (SD = 10.5), indicating high pain catastrophizing on average.⁸⁷ Over 80% of the sample (N = 268) endorsed a current opioid prescription, of which the most common types were hydrocodone products (38.4%), oxycodone products (34.7%), and tramadol products (31.3%).

Table 1.

Participant Characteristics

	n (%)
Sex
Female	119(38.9)
Ethnicity
Hispanic/Latino	8(2.6)
Race
American Indian/Alaska Native	8(2.6)
Asian	9(2.9)
Black or African American	75(24.5)
Middle Eastern or North African	2(0.7)
White	210(68.6)
Other	2(0.7)
Marital Status
Single	38(12.4)
Married	259(84.6)
Divorced/Separated/Widowed	9(2.9)
Education
High school graduate or GED	7(2.3)
Some college	23(7.5)
Technical/Associates degree	14(4.6)
4-year college degree	174(56.9)
Some school beyond college	11(3.6)
Professional degree (e.g., JD, MD)	77(25.2)
Household Income
<$10,000	7(2.5)
$10,000–24,999	25(8.2)
$25,000–49,999	89(29.1)
$50,000-$74,999	108(35.3)
≥$75,000	77(25.2)
Anxiety Symptomsa
None to mild	80(26.1)
Moderate to severe	226(73.9)

	M (SD)	Range
Age	38.7(11.1)	18 – 77
Pain Intensityb	21.1(4.7)	5 – 30
Current	6.8(2.2)	0 – 10
Past-3-Month Worst	7.4 (1.8)	2 – 10
Past-3-Month Average	6.9(1.9)	1 – 10
Pain Catastrophizing Scorec	30.7(10.5)	0 – 52
ISId total score	16.9(4.8)	1 – 27

N = 306.

^aGeneral Anxiety Disorders – 7;

^bGraded Chronic Pain Scale – Characteristic Pain Intensity;

^cPain Catastrophizing Scale – Total Score;

^dInsomnia Severity Index.

In terms of BZD use, 43.1% of participants reported that they were first prescribed a BZD greater than 12 months ago. The most commonly reported primary BZDs (i.e., those with highest past-month days of use) were alprazolam (32%), lorazepam (17%), clonazepam (14%), and diazepam (13%). The mean SDS score of 7.7 (SD = 3.3) suggests moderate severity of BZD dependence. Mean number of days using BZDs was 14.2 (SD = 10.0) in the past month. Participants were asked to indicate all reasons for which a BZD had been prescribed to them, and the most commonly endorsed responses included anxiety (46%) and pain (45%), followed by insomnia (23%) and neurologic conditions (11%). Participants also reported all conditions/symptoms for which they seek relief by taking BZDs, and the most commonly endorsed was pain (56%), followed by anxiety (48%), depression (44%), insomnia/sleep problems (27%), and neurological conditions (11%).

Bivariate Correlations

All bivariate correlations are presented in Table 2. Significant associations were observed between BZD dependence severity and race (dichotomized white/non-white; r = .165, p = .004) and education (r = .310, p < .001); between BZD misuse and sex (r = .161, p = .005), race (r = .122, p = .033), and education (r = .334, p = < .001); and between past-month days of use and race (r = −.185, p = .001) and education (r = −.162, p = .005). Therefore, race and education were retained as covariates in all analyses.

Table 2.

Bivariate and Point-Biserial Correlations Between Sociodemographic, Pain Characteristics, Primary Predictor, and Primary Outcome Variables

Variable	1	2	3	4	5	6	7	8	9	10	11
1 Sex	-	−.08	.06	.06	.03	−.10	−.05	.10	−.05	.06	.16**
2 Age		-	−.01	−.05	.02	.26**	.18**	.06	−.04	.01	−.06
3 Race a			-	.12*	−.02	.07	.12*	.12*	−.19**	.17**	.12*
4 Education				-	.23**	.01	.16**	.16**	−.16**	.31**	.33
5 Income					-	−.17**	.09	−.04	−.10	.06	.01
6 Marital status						-	.06	−.01	−.07	.09	.06
7 Pain intensity b							-	.45**	.08	.54**	.29**
8 Pain catastrophizing d								-	−.11	.67**	.45**
9 Past-month use frequency									-	−.05	−.12*
10 BZD Dependence severity e											.57**
11 BZD Misuse f											-

Note. N = 306.

^aWhite versus non-white;

^bGraded Chronic Pain Scale – Characteristic Pain Intensity;

^cGraded Chronic Pain Scale - Disability;

^dPain Catastrophizing Scale;

^eSeverity of Dependence Scale total score;

^fNo misuse versus any misuse in past year;

^*p < .05;

^**p < .01.

Pain Intensity and BZD Dependence, Misuse, and Use Frequency

Pain intensity was positively associated with severity of BZD dependence (Table 3). Pain intensity was positively associated with past-month use frequency (Table 3). Specifically, pain intensity accounted for 24.5% and 1.8% of unique variance in dependence severity and past-month use frequency, respectively, after accounting for all other variables in the models. Greater pain intensity was associated with greater likelihood of endorsing past-year BZD misuse (Table 3), after accounting for all other variables in the model. Specifically, for every one-point increase in pain intensity score, participants were 1.2 times as likely to endorse past-year BZD misuse. Post hoc analyses of misuse subtypes further indicated that pain intensity was positively associated with use without own prescription (adjusted odds ratio [AOR] = 1.125, 95% confidence interval [CI]: 1.057 – 1.197, p < .001) and use in greater amounts than prescribed (AOR = 1.078, 95% CI: 1.022 – 1.137, p = .006).

Table 3.

Pain Intensity on BZD Dependence Severity, Past-Month BZD Use Frequency, and Likelihood of BZD Misuse

Linear Regressions
	BZD Dependence Severity			Past-Month BZD Use Frequency
Variable	β	t	p	β	t	p
Sex	.063	1.356	.176	−.031	−.559	.577
Age	−.064	−1.361	.175	−.077	−1.360	.175
Race	.073	1.554	.121	−.181	−3.220	.001**
Education	.212	4.480	< .001**	−.165	−2.902	.004**
Pain intensitya	.515	10.717	< .001**	.141	2.440	.015*
R2		.360			.076
ΔR2		.245			.018
F for ΔR2		114.863**			5.955*

Logistic Regression: Likelihood of BZD Misuse
Variable	B	SE	AOR	95% CI	p
Sex	−1.082	.383	.339	.160 – .718	.005**
Age	−.028	.016	.973	.943–.1.004	.083
Race	−.484	.453	.616	.254–1.497	.285
Education	.753	.170	2.123	1.522–2.961	< .001**
Pain intensity a	.185	.043	1.204	1.107–1.309	< .001**

Note: N = 306. Linear regression: Results shown are from the second step of each linear regression model; β = standardized beta weights. Logistic regression: Results shown are from the second step of the logistic regression model; AOR = adjusted odds ratio; Sex: Reference group = female; Race: Reference group = White;

^aGraded Chronic Pain Scale – Characteristic Pain Intensity;

^*p < .05,

^**p < .01

Pain Catastrophizing and BZD Dependence, Misuse, and Use Frequency

In linear regression models assessing pain intensity × pain catastrophizing interaction terms (Table 4), tests examining pain catastrophizing as a moderator of pain intensity-BZD relationships were non-significant (Step 3: ps > .14). However, pain catastrophizing was positively associated with dependence severity (Step 2; β = .506, p < .001) at Step 2 of the model, after accounting for pain intensity. In contrast, pain catastrophizing was negatively associated with past-month BZD use frequency (Step 2: β = −.151, p = .016) at Step 2 of the model, accounting for pain intensity.

Table 4.

Pain Intensity, Pain Catastrophizing, and BZD Use Outcomes

Linear Regression: BZD Dependence Severity
	β	t	p	ΔR 2	p for ΔR 2
Step 1				.115	<.001**
Sex	.036	.660	.510
Age	.031	.563	.574
Race	.128	2.340	.020*
Education	.294	5.369	< .001**
Step 2				.442	<.001**
Pain Intensitya	.290	6.526	< .001**
Pain Catastrophizingb	.506	11.517	< .001**
Step 3				.001	.418
Pain Intensity × Pain Catastrophizing	−.181	−.812	.418
Linear Regression: Past-Month BZD Use Frequency
	β	t	p	ΔR 2	p for ΔR 2
Step 1				.058	.001*
Sex	−.039	−.687	.493
Age	−.051	−.911	.363
Race	−.166	−2.946	.003*
Education	−.142	−2.519	.012*
Step 2				.036	003*
Pain Intensitya	.208	3.267	.001*
Pain Catastrophizingb	−.151	−2.411	.016*
Step 3				.007	.142
Pain Intensity × Pain Catastrophizing	−.469	−1.471	.142
Logistic Regression: Likelihood of BZD Misuse
Variable	B	SE	AOR	95% CI	p
Step 1
Sex	−.871	.356	.418	.208–.841	.015*
Age	−.011	.015	.989	.960–1.019	.454
Race	−.548	.435	.578	.246–1.356	.208
Education	.864	.168	2.373	1.707–3.299	< .001**
Step 2
Pain Intensitya	.107	.054	1.113	1.002–1.237	.046*
Pain Catastrophizingb	.115	.023	1.122	1.072–1.174	< .001**
Step 3
Pain Intensity × Pain Catastrophizing	.008	.005	1.008	.999–1.017	.093

Note: N = 306. Linear regression: β = standardized beta weights. Logistic regression: Results shown are from the second step of the logistic regression model. Sex: Reference group = female; Race: Reference group = White;

^aGraded Chronic Pain Scale – Characteristic Pain Intensity;

^bPain Catastrophizing Scale total score;

^*p < .05;

^**p < .001.

Finally, in the logistic regression model assessing the pain intensity × pain catastrophizing interaction term (Table 4), the test examining pain catastrophizing as a moderator was non-significant (Step 3: p = .093). However, pain catastrophizing was positively associated with likelihood of endorsing BZD misuse at Step 2 of the model (Step 2: AOR = 1.122, 95% CI: 1.072–1.174, p < .001). Specifically, for every one-point increase in pain catastrophizing score, participants were approximately 1.1 times as likely to endorse past-year BZD misuse after accounting for pain intensity.

Opioid Subsample: Pain Intensity, Pain Catastrophizing, and BZD Dependence, Misuse, and Use Frequency

Results of linear regression models were similar among the N = 268 participants who endorsed a current opioid prescription. Linear regression models suggested that greater pain intensity was associated with greater BZD dependence severity (F[1,262] = 71.277, p < .001, ΔR² = .201) and with past-month BZD use frequency (F[1,262] = 4.468, p = .035, ΔR² = .015), after accounting for all other variables in the models. In contrast, in the logistic regression model, there was no significant relationship between pain intensity and likelihood of endorsing BZD misuse (AOR = 1.096, 95% CI: .970 – 1.237, p = .141) among participants reporting an opioid prescription.

Similar to the overall sample, in linear regression models assessing pain intensity × pain catastrophizing interaction terms, neither interaction term was significant (Step 3: ps > .23). However, pain catastrophizing was positively associated with BZD dependence severity (Step 2: β = .496, p < .001) and negatively associated with past-month BZD use frequency (Step 2: β = −.127, p = .048), each at step 2 of the model and accounting for pain intensity. Finally, pain catastrophizing did not significantly moderate associations between pain intensity and likelihood of BZD misuse (Step 3: p = .548). However, pain catastrophizing was positively associated with likelihood of BZD misuse at Step 2 of the model (Step 2: AOR = 1.115, 95% CI: 1.046 – 1.189, p = .001), among the subgroup of participants prescribed opioids.

Discussion

The current study is the first to examine pain intensity and pain catastrophizing in relation to clinically relevant patterns of BZD use (i.e., dependence, misuse) in a chronic pain population. Among a sample of adults with chronic musculoskeletal pain who report currently holding prescriptions for BZD anxiolytics and/or hypnotics, pain intensity was positively associated with severity of BZD dependence symptoms and likelihood of BZD misuse. Notably, pain intensity accounted for approximately 25% of unique variance in BZD dependence scores after controlling for relevant sociodemographic factors. Individuals with greater pain intensity also reported greater past-month days of BZD use, an indicator of heavier BZD use. Results further indicated that pain catastrophizing scores accounted for variance in BZD use outcomes above and beyond that attributable to pain intensity. It is also notable that over half of all participants endorsed pain as a symptom for which they seek relief by taking BZDs. Overall, these findings add to an emerging literature examining pain-BZD use relations by highlighting positive covariation between pain intensity, pain catastrophizing, clinically relevant measures of BZD use, including indices of dependence and misuse.

These findings may also be viewed in light of an established reciprocal model, which posits that pain and substance use interact in the manner of a positive feedback loop that leads to the exacerbation of both conditions over time.²³ Consistent with this perspective, individuals with chronic pain who report greater pain and more frequent use of BZDs may be at elevated risk for escalation of both pain and BZD misuse/dependence. Indeed, pain has been shown to motivate use of other substances (e.g., tobacco, alcohol),^{22, 64} and research on BZD use motives specifically has shown that the desire to alleviate negative affect and enhance acute analgesia can function as primary motives for BZD use.^{11, 35, 58, 91} Taken together, this prior literature suggests that pain may be a salient motivator of BZD use. Future experimental and longitudinal research is needed to test whether pain may motivate BZD use and elucidate mechanisms in pain-BZD relations. Conversely, it is also possible that heavy, regular BZD use worsens pain through mechanisms including allostatic load, that is, maladaptive changes in a physiological system due to accumulated stressors.^{26, 60} Finally, a key next step is investigating potential moderating or mediating roles of anxiety and insomnia in these relationships.

The current findings further suggest that pain catastrophizing may warrant consideration as a modifiable cognitive-affective mechanistic factor in the context of integrated treatment for comorbid pain and BZD use disorders. Prior work suggests that pain catastrophizing may be an important target in interventions designed to reduce medication misuse among individuals seeking treatment for chronic pain.³² In the present study, pain catastrophizing scores showed positive associations with BZD dependence scores and likelihood of BZD misuse, but negative associations of small magnitude with past-month days of BZD use. Some prior literature suggests that catastrophizing may be more relevant to outcomes that are closely connected with dependence symptoms than with quantity/frequency of consumption.^{44, 68} For example, Nieto and colleagues (2021) found that pain intensity and catastrophizing were both positively associated with alcohol dependence and alcohol craving, while only pain intensity, and not pain catastrophizing, was associated with alcohol consumption (drinks/day). Pain catastrophizing may be a target for intervention in pain-BZD relations to the extent that situational catastrophizing¹³ is shown to predict BZD self-administration or mediate pain-BZD use relations in future work.

That over 80% of respondents with chronic pain and current BZD prescriptions also reported current use of prescription opioid medications underscores the need for further research on concurrent prescription BZD and opioid use, with an emphasis on mitigating the potentially fatal consequences associated with such patterns of use. High prevalence of BZD-opioid and BZD-alcohol co-use have previously been reported.^{33, 41} Our findings are broadly consistent in showing high prevalence of opioid prescription and past-month alcohol use among this sample of individuals prescribed BZDs. However, the current data does not reflect the extent to which individuals may be using BZDs in the same general timeframe (e.g., same-day use, use within two hours) as they take their opioid medications or consume alcohol, behaviors more likely to produce negative health outcomes.⁵⁵ Investigation of specific, higher-risk co-use patterns among individuals with chronic pain is a key next step in this line of research.

Results should be interpreted in light of several limitations. First, cross-sectional analyses preclude causal interpretations and inferences regarding directionality. Second, these data were collected via online survey, and it was not possible to verify self-reported medication use or chronic pain status. The present study did, however, utilize recommended screening methods to disguise inclusion criteria and thereby reduce deceptive responding⁷⁸. Third, generalizability may be limited in that mTurk samples tend to display demographic differences from nationally representative samples (e.g., higher average educational attainment),^{72, 96} as well as higher lifetime prevalence of use of illicit substances.⁸⁶ Fourth, although participants were excluded if reporting primary neuropathic pain, some participants may have experienced additional pain types beyond musculoskeletal pain. Fifth, we did not assess use of Z-drugs, a separate medication class approved for insomnia that, like BZDs, confers health risks when used concurrently with opioids.^{88, 100}

In summary, this is the first study to examine pain intensity and catastrophizing in relation to clinically relevant BZD use outcomes, including dependence and misuse, among individuals with chronic pain. These initial findings implicate pain and pain-related cognitive-affective processes in higher-risk BZD use, and suggest that pain relief is a common, yet underrecognized, self-reported motivation for taking BZDs. Limited prior research has examined the role of pain in relation to problematic patterns of BZD use, despite growing concern related to heavy BZD use, misuse, and dependence in chronic pain populations. Future research would likely benefit from elucidating mechanisms and temporal ordering in pain-BZD use relations and examining the extent to which these findings replicate among treatment-seeking pain samples with and without concurrent prescription opioid use.

Funding

This work was funded by National Institute on Alcohol Abuse and Alcoholism grants (R01AA028639, R01AA02484) awarded to Joseph W. Ditre.

Highlights.

• Findings extend limited prior work examining pain in relation to addiction-relevant BZD use patterns

• Pain intensity was positively associated with BZD dependence severity and likelihood of BZD misuse.

• Pain catastrophizing accounted for variance in BZD use beyond that accounted for by pain intensity.

• 56% of participants endorsed pain as a symptom for which they seek relief by taking BZDs.

• Future research should elucidate mechanisms and temporal ordering in pain-BZD use relations.