Nicotine use is Associated With Elevated Risk of Opioid use Disorder Following Total/Hemi Hip Arthroplasty Procedures

Abstract

Introduction

Opioids are often necessary for pain control after surgery; however, it is challenging to predict the risk of developing opioid use disorder (OUD) after surgical interventions. Nicotine impacts opioid metabolism, which is associated with the risk of OUD. We aimed to determine whether perioperative nicotine use is associated with incident OUD following hip arthroplasties.

Aims and Methods

We performed a retrospective cohort study using a national de-identified database to identify subjects who underwent total/hemi hip arthroplasty in the United States from 2013 to 2018 and received postoperative opioid treatment within two weeks of the procedure. The matched cohorts consisted of nicotine-dependent (N = 10 464) versus non-nicotine-dependent individuals (N = 10 464). Subjects were matched on known confounders: age, sex, race, ethnicity, alcohol use disorder, sedative-hypnotic/anxiolytic disorders, congestive heart failure, chronic obstructive pulmonary disease, depressive disorders, and anxiety disorders. We assessed for development of OUD, and though IVDU was not directly measured due to data limitations, we assessed for secondary outcomes of OUD associated with IVDU: human immunodeficiency virus (HIV), Hepatitis C Virus (HCV), and Hepatitis B Virus (HBV) at multiple timepoints (1 month, 3 months, 6 months, 1 year, 3 years) after surgery utilizing logistic regression analysis.

Results

Nicotine dependence was associated with increased risk of developing OUD at 3 months postsurgery (OR = 2.36, p < .001) and thereafter (OR = 2.37–2.42, p < .001). Peri-operative nicotine dependence was also associated with HIV (OR = 1.26–1.92, p < .05) and HCV (OR = 1.47–1.60, p < .02) at all timepoints.

Conclusions

Considering nicotine use when determining OUD risk following total/hemi hip arthroplasty procedures is warranted.

Implications

Orthopedic surgeons should monitor individuals who use nicotine products with vigilance for signs and symptoms of OUD postoperatively. The opioid risk tool should be evaluated to determine if the inclusion of nicotine products enhances the predictive model. Future research efforts may focus on the impact of presurgical smoking cessation on the development of OUD.

Introduction

Combustible nicotine products, used by more than 40 million people in the United States, are associated with chronic pain as well as increased cartilage loss that may lead to orthopedic surgery.^1–4 Following total hip arthroplasty, use of combustible nicotine products is also associated with several adverse outcomes, including increased risk of periprosthetic joint infections, cardiovascular events, thromboembolism, and implant revisions.^5–9 These complications may be attributed to the impact of these products on tissue oxygenation, the general systemic inflammatory response, and bone mineral density.^10–12 As such, those who continue to use in the peri-operative period leading up to surgery are at greater risk of experiencing complications associated with their surgery. Following a total hip arthroplasty, opioids are typically required for pain control; however, some surgeons find it challenging to predict the amount of medication patients will require after such a procedure.^13,14 It may be even more difficult to predict which patients will develop opioid use disorder (OUD) after exposure to necessary pain medication. {Delaney, 2020 #27}Previous studies that explored this question identified the following risk factors for OUD after surgery: chronic pulmonary disease, congestive heart failure, history of opioid use, female gender, younger age at time of surgery, depressive disorders, and anxiety disorders.^15–19 Use of combustible nicotine products is associated with many of these outcomes, but it has not been evaluated as an independent predictor of OUD after surgery. Moreover, alcohol, illicit drugs, gender, age, and mental illness are all considered in the opioid risk tool that providers utilize to assess the risk of OUD, but nicotine combustibles are not listed as a relevant variable. Thus, the primary objective of this study was to determine whether perioperative nicotine dependence is associated with postoperative OUD in patients undergoing total and hemi-hip arthroplasty procedures. Our second objective was to assess whether nicotine dependence is associated with known complications of OUD and intravenous (IV) drug use (ie, Human Immunodeficiency Virus (HIV), HCV, and HBV) in patients undergoing total and hemi-hip arthroplasty.

Methods

Study Design and Data Source

This is a retrospective cohort study that utilized the TriNetX Network database Research Network. This research network compiles data from 57 health systems across the United States. Our cohorts consisted of individuals 21 years of age and older who underwent a total/hemi hip arthroplasty captured by the research network (established in 2013) prior to November 1, 2018, and received postoperative opioid treatment within 2 weeks of the procedure. Individuals were followed for 3 years after their index date of hip surgery for outcomes of interest. We identified the exposure and outcomes of interest, with ICD-10 and current procedural terminology codes (Appendix A). Then we divided individuals into two cohorts based on the presence of nicotine dependence, which we used as a proxy for combustible nicotine exposure (nicotine dependent and non-nicotine dependent). Because TriNetX data is de-identified, this study was exempt from Institutional Review Board approval.

Cohort Comparison

We chose to focus on individuals who received total/hemi hip arthroplasty because this surgery often requires postoperative pain management with an opioid. We clustered these groups given recent meta-analysis that demonstrated no difference in complications associated with chronic pain, such as reoperation or infection, with these procedures.²⁰ However, given potential differences in indications for hemiarthroplasty and total hip arthroplasty, our study controlled for preoperative health status by propensity score matching.^21,22 The 1:1 propensity score matching method was performed using logistic regression and nearest neighbor algorithms with a caliper width of 0.1 pooled standard deviation, ensuring that matched pairs have similar baseline characteristics. The matching was based on risk factors identified in previously published studies, including age, sex, race, ethnicity, congestive heart failure, chronic obstructive pulmonary disease, depressive disorders, and anxiety disorders.²³ Additional substance use disorders associated with tobacco and OUD (alcohol and sedative-hypnotic/anxiolytic disorders) were added to the model to mitigate the risk of confounding.^24–27 Sedative-hypnotic/anxiolytic disorders are inclusive of disorders involving benzodiazepines, barbiturates, and Z-drugs. Both alcohol and sedative-hypnotic/anxiolytics are independently associated with nicotine products, potentiate the effect of opioids, and are associated with adverse consequences when utilized with opioids.^26,28–32 All diagnoses were determined by ICD-10 codes. Postoperative outcomes, including OUD, HIV, hepatitis B virus (HBV), hepatitis C virus (HCV), were assessed at five time points up to 3 years after surgery (1 month, 3 months, 6 months, 1 year, 3 years). HIV, HCV, and HBV were included as secondary outcomes as they are known complications of OUD and intravenous drug use (IVDU).^33–35 Due to the diagnostic complexity of HIV, HBC, and HCV, we also included positive test results from serum studies for these outcomes. We excluded individuals if they developed an outcome of interest (OUD, HIV, HBV, HCV) prior to their surgery. We calculated odds ratios with 95% confidence intervals (95% CI) to assess the likelihood that each cohort experienced the primary outcome (development of OUD) and secondary outcomes (HIV, HCV, HBV) at the time-points of interest. We did not use omnibus outcomes, so the alpha value was set at 0.05 for all tests of statistical significance.³¹ We analyzed all data through the TriNetX software, which utilizes JAVA, R, and Python programming languages.

Results

We were able to identify 85 436 individuals who underwent total/hemi hip arthroplasty procedures and received opioid treatment during the recovery period. There were 11 342 subjects classified in the nicotine-dependent cohort and 74 094 subjects in the non-nicotine-dependent cohort. After matching, both cohorts had 10 464 subjects. The cohorts were similar in all demographic variables and relevant comorbidities after matching (Table 1). In the nicotine-dependent group, there were a total of 274 cases of OUD, 304 cases of HIV, 245 cases of HBV, and 407 cases of HCV at 3 years. By comparison, there were a total of 122 cases of OUD, 230 cases of HIV, 210 cases of HBV, and 298 cases of HCV in the non-nicotine exposed group at 3 years (Figure 1). Compared to the non-nicotine dependent group, nicotine dependent individuals were significantly more likely to develop OUD at 3 months (OR = 2.36, p < .001), 6 months (OR = 2.42, p < .001), 1 year (OR = 2.42, p < .001), and 3 years (OR = 2.37, p < .001) but not at 1 month (OR = 1.81, p = .064; Table 2). The nicotine dependent group was also more likely to develop HIV at all timepoints (OR = 1.26–1.92, p < .05) and HCV at all timepoints (OR = 1.47–1.60, p < .02; Table 2). We did not detect a difference between our cohorts for the development of HBV at any timepoint.

Table 1.

Demographics and Comorbidities of Cohort for Propensity Score Matching

	Unmatched					Matched
Characteristic name	Hip replacement + Nicotine	% Cohort	Hip replacement + no nicotine	% Cohort	p-value	Hip replacement + nicotine	% Cohort	Hip replacement + no nicotine	% Cohort	p-value
	N = 11 342		N = 74 094			N = 10 464		N = 10 464
Age mean at index (SD)	61.1 (13.2)		69.0 (13.9)			61.5 (13.2)		60.8 (14.9)
Male	5878	51.83%	30274	40.86%	<.0001	5328	50.92%	5396	51.57%	.347
Female	5464	48.18%	43812	59.13%	<.0001	5136	49.08%	5067	48.42%	.340
White	8545	75.34%	58219	78.58%	<.0001	7898	75.48%	8024	76.68%	.041
Black or African American	1874	16.52%	7958	10.74%	<.0001	1693	16.18%	1663	15.89%	.572
Asian	25	0.22%	643	0.87%	<.0001	25	0.24%	16	0.15%	.159
American Indian or Alaska Native	37	0.33%	336	0.45%	.056	35	0.33%	27	0.26%	.309
Native Hawaiian or Other Pacific Islander	10	0.09%	45	0.06%	.283	10	0.10%	10	0.10%	1.000
Unknown ethnicity	2438	21.50%	19007	25.65%	<.0001	2301	21.99%	2232	21.33%	.247
Not Hispanic or Latino	8669	76.43%	53487	72.19%	<.0001	7944	75.92%	7996	76.41%	.399
Hispanic or Latino	235	2.07%	1600	2.16%	.550	219	2.09%	236	2.26%	.420
Unknown race	856	7.55%	6893	9.30%	<.0001	809	7.73%	733	7.01%	.044
Depressive disorders	3192	28.14%	9676	13.06%	<.0001	2601	24.86%	2491	23.81%	.076
COPD	3135	27.64%	5048	6.81%	<.0001	2486	23.76%	2393	22.87%	.128
Anxiety disorders	2798	24.67%	8404	11.34%	<.0001	2288	21.87%	2229	21.30%	.322
HF	1358	11.97%	6935	9.36%	<.0001	1183	11.31%	1172	11.20%	.810
Alcohol related disorders	1722	15.18%	1369	1.85%	<.0001	1047	10.01%	992	9.48%	.200
Sedative-hypnotic/anxiolytic disorders	132	1.16%	97	0.13%	<.0001	76	0.73%	44	0.42%	.003

HF = congestive heart failure, COPD = chronic obstructive pulmonary disease.

Figure 1.

Line Graph Displaying the Trend of Developing Opioid Use Disorder, Human Immunodeficiency Virus, Hepatitis B Virus, and Hepatitis C Virus after Hip Surgery.

Table 2.

Comparison of Each Matched Outcome in Nicotine Dependent Versus Non-nicotine Dependent Cohorts

		Odds ratio	95% CI	p-value
OUD	1 Month	1.81	(0.958, 3.419)	.064
	3 Months	2.36	(1.502, 3.723)	<.001
	6 Months	2.42	(1.685, 3.483)	<.001
	1 Year	2.42	(1.797, 3.269)	<.001
	3 Years	2.37	(1.918, 2.949)	<.001
HIV	1 Month	1.87	(1.215, 2.887)	.004
	3 Months	1.92	(1.371, 2.714)	<.001
	6 Months	1.88	(1.388, 2.550)	<.001
	1 Year	1.26	(1.007, 1.600)	.043
	3 Years	1.37	(1.155, 1.635)	<.001
HBV	1 Month	1.09	(0.697, 1.706)	.706
	3 Months	1.28	(0.891, 1.858)	.178
	6 Months	1.12	(0.827, 1.540)	.445
	1 Year	1.08	(0.845, 1.396)	.52
	3 Years	1.18	(0.980, 1.423)	.08
HCV	1 Month	1.47	(1.064, 2.036)	.009
	3 Months	1.59	(1.211, 2.088)	.001
	6 Months	1.54	(1.215, 1.972)	<.001
	1 Year	1.60	(1.309, 1.977)	<.001
	3 Years	1.45	(1.250, 1.695)	<.001

OUD = opioid use disorder; HIV = human immunodeficiency virus; HBV = hepatitis B virus; HCV = hepatits C virus.

These statistical analyses were matched (N = 10 464 in each cohort) on age, sex, race, ethnicity, depressive disorders, anxiety disorders, alcohol use disorder, sedative-hypnotic/anxiolytic disorders, congestive heart failure, and chronic obstructive pulmonary disease.

Discussion

Following total/hemi hip arthroplasty procedures, patients with nicotine dependence had more than double the risk of OUD 3 years postoperatively, and they experienced increased risk of OUD-related and IV drug use-related infectious diseases (HIV and HCV) after surgery. There are multiple potential mechanisms that might explain the relationship between nicotine dependence and the development of OUD following surgery. First, studies have demonstrated that variants in opioid metabolism may increase the risk of OUD,³⁶ and combustible nicotine products can impact enzymes critical for opioid metabolism.^37–40 This can enhance the euphoric effects associated with opioid use, decrease the duration of opioid effectiveness, and increase the need for higher levels of medication.⁴¹ All these factors have the potential to increase the risk of OUD.⁴² In addition, combustible nicotine products are associated with systemic inflammation, poor blood flow, delayed wound healing, and adverse outcomes after surgery, which all increase the risk of prolonged postoperative opioid exposure.^43–46 An elevated dose requirement, more frequent dosing, and potentially longer treatment courses due to postoperative complications or delayed healing may all contribute to the elevated risk of OUD following surgery.

The risk of developing HIV, HCV, and HBV (conditions strongly associated with OUD and IVDU) postoperatively also bears discussion. Tolerance to opioids and the relative efficiency (potency/cost) of injection are the primary reasons people transition from oral opioids to IVDU.⁴⁷ Tolerance is a consequence of opioid exposure (dose/duration) that may be perpetuated by surgical complications and poorly controlled chronic pain in a postoperative cohort.^47,48 Perioperative nicotine use is directly associated with each of these intermediates that promote tolerance, OUD, and IVDU.^23,49

Our study found that there was no significant increase in risk for HBV between patients who used and did not use nicotine perioperatively. This may be attributed to the broad deployment of HBV vaccines and the relatively low prevalence of HBV in the United States.⁵¹ However, our data do suggest an increased risk of HIV and HCV after total/hemi hip arthroplasty procedures in our nicotine-dependent cohort. While challenging to establish causality in a retrospective study, some cases of HIV and HCV in our cohorts may be direct sequelae of IVDU and other high-risk behaviors associated with OUD. This inference is based on existing literature that estimates 75 percent of people newly infected with HCV have a history of IVDU and opioid misuse.⁵² HIV is more often transmitted sexually than through IVDU, but 10% of new HIV cases are attributable to IVDU.⁵³ There are currently 6.7–7.6 million people in the United States with an OUD, and 3.7 million people who inject drugs.^54-55 Similar to a previous estimate that 0.9% of people advance to OUD after an opioid prescription, 1.1% of people in our non-nicotine cohort were subsequently diagnosed with OUD.⁵⁶ By comparison, 2.6% of our cohort with peri-operative nicotine dependence were later diagnosed with OUD.

While most who undergo surgery and receive an opioid do not progress to OUD or IVDU, there are 40 million people who undergo surgery each year in the United States.⁵⁷ Surgery is a risk factor for OUD, and each new case of OUD, HIV, or HCV may have devastating physical, psychological, and social consequences.^58-60 Cumulatively, OUD costs the United States more than 700 billion dollars annually, and preventing a single case translates to 2.2 million dollars in savings.⁶¹ Continued emphasis on avoidance of harmful perioperative nicotine products and providing increased attention to opioid prescriptions for patients with perioperative combustible nicotine use may offer improved postsurgical outcomes as well as significant and multi-faceted public health benefits.^62-64

Limitations

Though we feel these data offer a critical perspective regarding a national epidemic, there are several limitations to consider. First, there are inherent limitations associated with electronic health record (EHR) studies, such as inaccurate and incomplete coding, which are well-documented and unavoidable. For example, substance use disorders are often under-reported by patients and thus may not be accurately coded by providers.⁶³ The TriNetX database that we utilized is a research network that spans the United States and provides access to de-identified aggregated EHR data. The aggregate nature of these data and the lack of patient-level identifiable information available for analysis preclude assessments for missingness or misclassification. In addition, we were not able to determine the level of nicotine use within our nicotine-dependent cohort, so we could not assess for a dose–response relationship. Furthermore, we matched cohorts on key variables, but matching is an imperfect process, and the variables uncovered from previously published literature do not exclude the potential for unmeasured confounders. For example, our dataset is limited to clinical diagnoses coded in the EHR, so key behavioral health variables such as high-risk sexual behaviors and IVDU that contribute to HIV and HCV could not be included in the models. Although individuals with a baseline diagnosis of OUD (which accounts for the majority of IVDU) were excluded from the study, some risk of bias persists from unmeasured confounders. Among the matched variables, there remained a residual between-group difference in white race (1.2%), unknown race (0.72%), and sedative-hypnotic/anxiolytic disorders (0.3%) even after matching. Additional post-matching multi-variable analysis was not possible due to limitations with our cloud computing software. Finally, the TriNetX platform may not represent the general United States population, but rather people receiving clinical care by hospital systems within its network.

Given the limitations inherent in EHR research, a single retrospective study is not sufficient to determine causation. Nevertheless, these findings are relevant to scientists, health care providers, public health officials, and patients for several reasons. First, the proxy variable of patient self-report with tobacco use has been demonstrated to approximate the gold standard of cotinine with 97.6% accuracy in the peri-operative setting.⁶⁴ Second, were nicotine screening was not performed (which is less likely peri-operatively) or nicotine use was under-reported, this would yield higher levels of nicotine-exposed individuals in the non-nicotine cohort. We would expect under-reporting and under-diagnosis to increase similarity between groups, which would mute a between-group difference and bias results towards the null. The imbalance in sedative-hypnotic/anxiolytic disorders after matching introduces an anti-conservative bias, but the magnitude of the imbalance (0.3%) is unlikely to shift inferences.

Although EHR data and matching processes have known imperfections, as a health care and scientific community, we have very limited information regarding the risk of OUD after surgery. Identification of risk factors are necessary to structure interventions that may mitigate preventable morbidity, mortality, and health care expenditure. In addition, while we cannot establish a biological gradient in this sample and there are inherent limitations with EHR data, the association described in this study meets many of the Bradford Hill criteria for causal inference, including strength of association, temporality, biological plausibility, and coherence. Given that opioid overdoses continue to surge across the United States, additional research is required to prevent the development of OUD postoperatively while still providing thoughtful and appropriate pain management. These data offer a foundation for future inquiry. For example, the current opioid risk tool may benefit from considering combustible nicotine products, such as tobacco use, as an additional variable in its model. Finally, this study provides yet another reason for patients and providers to engage in smoking cessation efforts prior to orthopedic surgery.

Conclusion

While opioids are often necessary for pain control after surgery, surgeons may benefit from guidance to help determine who may be at elevated risk for OUD. The scientific literature and opioid risk tool are the most common resources clinicians utilize to determine the risk of OUD, but combustible nicotine products have not been mentioned in either of these. Here we demonstrate that nicotine dependence may be an independent predictor of OUD after total and hemi-hip arthroplasty that should be considered by researchers, patients, and providers when determining postoperative use of opioids for pain management.

References 51-64 are provided in the Supplementary Material

Funding

The project described was supported by the National Center for Advancing Translational Sciences, National Institutes of Health, through Grant K12 TR004529. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Declaration of Interest

The authors have no conflicts of interests to disclose.

Author Contributions

Zachary Freedman (Data curation [lead], Formal analysis [lead], Writing - original draft [lead]), Andrew Kim (Conceptualization [supporting], Project administration [lead], Writing - original draft [equal]), Nicholas Graziane (Conceptualization [supporting], Methodology [supporting], Writing - review & editing [equal]), Matthew Silvis (Conceptualization [equal], Methodology [equal], Supervision [equal], Validation [equal], Writing - review & editing [equal]), Elise Marino (Conceptualization [supporting], Writing - review & editing [equal]), B.S Brandon Fross (Project administration [equal], Writing - review & editing [supporting]), Ducel Jean-Berluche (Project administration [supporting], Writing - review & editing [supporting]), and Curtis Bone (Conceptualization [lead], Data curation [supporting], Formal analysis [supporting], Investigation [lead], Methodology [lead], Project administration [supporting], Supervision [lead], Validation [lead], Visualization [supporting], Writing - original draft [Supporting], Writing - review & editing [lead])

Data Availability

Data are not publicly available.