Impact of Preoperative Chronic Opioid Use on Postoperative Outcomes Following Lumbar Fusion: A Nationwide Analysis

Abstract

Study Design

Retrospective cohort study.

Objectives

Chronic opioid use is common among patients undergoing lumbar fusion, and preoperative use has been linked to worse postoperative outcomes. However, much of this evidence derives from administrative datasets with limited ability to adjust for mental health. We examined associations between chronic preoperative opioid use and a panel of postoperative outcomes, utilizing linked survey data to enhance control of confounding.

Methods

Retrospective cohort study within a diverse national research program. Adults undergoing lumbar fusion were classified as chronic opioid users (≥3 opioid prescriptions spanning ≥90 days in the preoperative year) or non-chronic users. Outcomes were persistent postoperative opioid use (prescriptions 90-180 days after surgery), reoperation, readmission, major medical complications, wound complications, pseudarthrosis, and death.

Results

Among 3424 patients, chronic use was strongly associated with persistent postoperative opioid use ([aOR] 5.40; 95% CI, 4.52-6.46). In contrast, crude associations with surgical outcomes were substantially attenuated after adjustment and were not statistically significant, although confidence intervals for several outcomes did not exclude clinically meaningful effects: reoperation (aOR 0.95; 95% CI, 0.76-1.19), readmission (1.10; 0.92-1.33), major complications (1.03; 0.78-1.34), wound complications (1.36; 0.97-1.88), pseudarthrosis (1.33; 0.86-2.01), and all-cause mortality (1.17; 0.74-1.82).

Conclusion

Chronic preoperative opioid use predicted persistent postoperative use but showed no statistically significant independent association with adverse surgical outcomes after adjustment, though the surgical models were underpowered and though confidence intervals did not exclude clinically meaningful effects for pseudarthrosis and wound complications. The attenuation of crude associations suggests that surgical risks attributed to chronic opioid use may in part reflect confounding by mental health and comorbidity rather than a direct effect.

1. Introduction

Opioids are prescribed to a large proportion of patients before spine surgery; back pain is among the most common indications for opioid therapy, and studies report that up to 70% of spine surgery patients use opioids preoperatively. ¹ Lumbar fusion, performed for degenerative disease in an aging and increasingly comorbid population, sits at the center of this concern. Preoperative opioid exposure has been associated in numerous studies with prolonged postoperative use, greater healthcare utilization, and worse patient-reported outcomes.^2-4

The most consistently reported consequence is persistent postoperative opioid use. Systematic reviews describe persistence in roughly 35% to 77% of preoperative opioid users after major surgery, compared with under 1% to 26% of opioid-naive patients, with the duration of preoperative use being among the strongest predictors.^5,6 A meta-analysis spanning 60 cohort studies and more than 13 million participants concluded that perioperative opioid use adversely affects a range of outcomes following spinal surgery. ⁷ Beyond persistence, preoperative use has been variably linked to reoperation, readmission, complications, and prolonged length of stay.^7,8

A crucial and underappreciated methodological problem complicates this literature: mental illness is a powerful shared cause of both chronic opioid use and poor surgical outcomes. Depression and anxiety are strongly associated with preoperative opioid consumption and independently predict persistent postoperative use, greater complications, and worse recovery.^9-11

We therefore examined the association between chronic preoperative opioid use and postoperative outcomes after lumbar fusion in a large, diverse national research program. This resource permitted a combined diagnostic-and-survey measure of depression and anxiety, enabling more complete adjustment for mental health than administrative databases allow. We evaluated a balanced panel of outcomes, persistent opioid use alongside surgical endpoints, hypothesizing that comprehensive confounding control would clarify which associations reflect a relationship and which are attributable to the psychiatric and medical burden accompanying chronic opioid use.

2. Methods and Materials

2.1. Study design and Data source

We conducted a retrospective cohort study following STROBE and RECORD guidance. Data derived from the All of Us Research Program Registered Tier Dataset v8, with data available through October 1, 2023, a national longitudinal cohort integrating electronic health records, participant surveys, and physical measurements from a demographically diverse United States population enriched for groups underrepresented in biomedical research, harmonized to the OMOP Common Data Model. All participants had an observation period beginning before the index procedure. Analyses were performed within the program’s secure environment against a version-bound Registered Tier release. Per the Data and Statistics Dissemination Policy, no cell representing fewer than 20 participants is reported.

2.2. Study population

Adults aged 18 years or older undergoing lumbar or lumbosacral fusion were eligible. Procedures were identified using ICD-10-PCS codes 0SG0 and 0SG1 and relevant lumbar fusion CPT codes (22533, 22534, 22558, 22585, 22612, 22614, 22630, 22632, 22633, and 22634). The index date was the first qualifying procedure. We required a clinical encounter within the year before and after the index date and excluded prior lumbar fusion to restrict to incident procedures.

2.3. Exposure definition

Chronic preoperative opioid use, the primary exposure, was defined as three or more opioid prescriptions spanning at least 90 days during the year before surgery, excluding the final 7 preoperative days, using a standard ingredient list (oxycodone, hydrocodone, morphine, hydromorphone, oxymorphone, fentanyl, tramadol, codeine, and tapentadol; methadone and buprenorphine were excluded to avoid conflating analgesic use with opioid use disorder treatment). Because the days-supply field was incompletely populated, a prescription-count-and-span definition was specified as primary, with a stricter definition (≥90 cumulative days of supply) reserved for sensitivity analysis. Non-chronic users comprised all remaining patients.

2.4. Covariates and outcomes

Covariates comprised age, sex, race, ethnicity, body mass index, Charlson Comorbidity Index (Quan ICD-10 adaptation, 730-day look-back), current/former/never smoking, and diagnostic flags for opioid use disorder, other substance use, tobacco use disorder, chronic pain, obesity, and diabetes. To address surgical and structural confounding, we additionally captured the number of levels fused (single vs multilevel, derived from add-on procedure codes), surgical approach (anterior, posterior, combined, or unspecified), osteoporosis diagnosis (ICD-10-CM M80/M81) and anti-osteoporosis therapy, and the structural surgical indication (spondylolisthesis, spinal stenosis, and degenerative disc disease). Mental health was captured by a combined measure: positive if a depression or anxiety diagnosis (ICD-10-CM F32/F33 or F40/F41) was recorded in the 730-day pre-index window, OR if the participant screened positive on the two-item depression and anxiety screens embedded in the program’s survey instruments (items corresponding to depressed mood, anhedonia, nervousness, or uncontrolled worry, each rated on the standard four-point frequency scale), with a positive screen defined as a response of “more than half the days” or “nearly every day” on any item. This measure reflects a history of diagnosed depression or anxiety together with recent symptom burden rather than a validated point-in-time perioperative psychiatric assessment; survey instruments were administered at program enrollment, whose timing relative to surgery varied across participants. The prespecified balanced outcome panel comprised persistent postoperative opioid use (any opioid prescription 90-180 days after surgery), reoperation (repeat fusion >30 days postoperatively), 90-day readmission, major medical complications (acute renal failure, respiratory failure, heart failure, pneumonia, or pulmonary embolism within 30 days of the index procedure), wound complications and surgical site infection (ascertained within 90 days), pseudarthrosis (ascertained from diagnostic codes 90-730 days postoperatively), and mortality (both 90-day and any-time all-cause mortality).

2.5. Statistical Analysis

Baseline balance was assessed by standardized mean differences (SMDs). Propensity scores were estimated by logistic regression of chronic use on all 20 covariates. Because outcomes were binary events within fixed windows, propensity-score-adjusted logistic regression estimated odds ratios with 95% confidence intervals; reoperation was additionally modelled as time-to-event using Cox regression. Covariate balance was verified using inverse-probability-of-treatment weights truncated at the 99th percentile, and overlap was examined via the region of common propensity-score support. Prespecified subgroup analyses evaluated the primary outcome within mental-health strata and, consistent with program policy on responsible reporting, among racial and ethnic groups large enough to meet the reporting threshold (White, Black, and Hispanic participants). Missing body mass index was imputed to the median. Analyses used R.

2.6. Ethical considerations

The present analysis used de-identified data within the secure analytic environment and adhered to the programme’s data-use and dissemination policies; separate ethics approval was not required for this secondary analysis.

3. Results

3.1. Cohort characteristics

Of 4238 patients undergoing lumbar fusion, 3424 met eligibility criteria after restriction to procedures performed in 2013-2023: 839 chronic opioid users (24.5%) and 2585 non-chronic users. Chronic users carried a heavier burden across nearly every measured domain (Table 1): higher prevalence of combined mental illness (55.9% vs 36.1%), opioid use disorder (11.2% vs 3.1%), other substance use (23.5% vs 11.6%), chronic pain (68.4% vs 46.8%), current smoking (16.8% vs 11.2%), and higher mean Charlson index (2.49 vs 1.56). The largest baseline imbalances were in chronic pain (SMD 0.45), combined mental illness (0.40), Charlson index (0.38), and opioid use disorder (0.32). The combined mental-health measure identified 177 patients with a positive symptom screen but no diagnostic code, who would have been misclassified using diagnostic data alone.

Table 1.

Selected baseline characteristics by exposure group

Characteristic	Non-chronic (n=2585)	Chronic (n=839)	SMD
Age, mean (SD), y	62.1 (11.7)	60.3 (11.0)	0.15
Female, %	1399 (54.1)	497 (59.2)	0.10
White, %	1742 (67.4)	576 (68.7)	0.14*
Black, n (%)	312 (12.1)	133 (15.9)	0.11
Hispanic, %	330 (12.8)	74 (8.8)	0.13*
Body mass index, mean (SD)	30.0 (5.2)	31.5 (5.9)	0.24
Charlson index, mean (SD)	1.56 (1.99)	2.49 (2.70)	0.38
Combined mental illness, %	932 (36.1)	469 (55.9)	0.40
Opioid use disorder, %	79 (3.1)	94 (11.2)	0.32
Other substance use, %	301 (11.6)	197 (23.5)	0.32
Chronic pain diagnosis, %	1209 (46.8)	574 (68.4)	0.45
Current smoker, %	289 (11.2)	141 (16.8)	0.16
Multilevel fusion, n (%)	1334 (51.6)	390 (46.5)	0.10
Posterior approach, n (%)	1594 (61.7)	473 (56.4)	0.11*
Osteoporosis diagnosis, n (%)	625 (24.2)	239 (28.5)	0.10
Spondylolisthesis, n (%)	1630 (63.1)	548 (65.3)	0.05
Spinal stenosis, n (%)	2250 (87.0)	783 (93.3)	0.21

SMD, standardized mean difference. Percentages are aggregate rates; cells <20 are suppressed per policy.

*SMD reflects the full multi-category distribution.

Multilevel procedures (46.5% vs 51.6%) and posterior approaches (56.4% vs 61.7%) were modestly less common among chronic users, indicating that their higher crude event rates were not attributable to greater surgical complexity. A truly opioid-naive subgroup (no preoperative opioid prescription; n=1597) had a persistent-use rate of 13.9%.

Propensity-score weighting achieved excellent balance: all weighted SMDs fell below 0.07. Only approximately 1.5% of patients lay outside the region of common support, indicating strong overlap; the full cohort was therefore retained.

3.2. Postoperative outcomes

Unadjusted outcome rates were higher among chronic users for every endpoint (Table 2), most strikingly for persistent postoperative opioid use (60.8% vs 18.5%).

Table 2.

Postoperative outcome rates by exposure group

Outcome	Non-chronic (n=2585), n (%)	Chronic (n=839), n (%)
Persistent opioid use (90–180 d)	479 (18.5)	510 (60.8)
Reoperation	386 (14.9)	146 (17.4)
Readmission (90 d)	644 (24.9)	250 (29.8)
Major medical complication	223 (8.6)	105 (12.5)
Wound complication/SSI	117 (4.5)	74 (8.8)
Pseudarthrosis	79 (3.1)	40 (4.8)
90-day mortality	0 (0.0)	0 (0.0)
All-cause mortality	70 (2.7)	35 (4.2)

SSI, surgical site infection.

3.3. Adjusted associations

After propensity-score adjustment, chronic opioid use remained overwhelmingly associated with persistent postoperative opioid use (aOR 5.40; 95% CI, 4.52-6.46) (Table 3). In contrast, associations with surgical outcomes attenuated markedly from their crude values and none reached statistical significance: reoperation (crude OR 1.20 to aOR 0.95; 95% CI, 0.76-1.19; Cox HR 1.15, 0.96-1.38), readmission (1.28 to 1.10; 0.92-1.33), major medical complications (1.52 to 1.03; 0.78-1.34), wound complications (2.04 to 1.36; 0.97-1.88), pseudarthrosis (1.59 to 1.33; 0.86-2.01), and all-cause mortality (1.56 to 1.17; 0.74-1.82). No deaths occurred within 90 days in either group.

Table 3.

Crude and propensity-score-adjusted odds ratios

Outcome	Events, n (non/chr)	Crude OR	Adjusted OR	95% CI
Persistent opioid use	479/510	6.82	5.40	4.52-6.46
Reoperation	386/146	1.20	0.95	0.76-1.19
Readmission	644/250	1.28	1.10	0.92-1.33
Major medical complication	223/105	1.52	1.03	0.78-1.34
Wound complication/SSI	117/74	2.04	1.36	0.97-1.88
Pseudarthrosis	79/40	1.59	1.33	0.86-2.01
All-cause mortality	70/35	1.56	1.17	0.74-1.82

OR, odds ratio; CI, confidence interval. Events shown as non-chronic/chronic. Reference group is non-chronic use. Adjusted models include the propensity score (20 covariates: age, sex, race, ethnicity, body mass index, Charlson index, combined mental health, opioid use disorder, other substance use, tobacco use, obesity, smoking, surgery year, number of levels fused, surgical approach, osteoporosis diagnosis and therapy, spondylolisthesis, spinal stenosis, disc degeneration).

3.4. Sensitivity and subgroup analyses

The persistent-use association was robust. Requiring at least two prescriptions in the 90-180 day window yielded an aOR of 4.95 (4.11-5.96), and defining persistence as any prescription 180-365 days postoperatively yielded an aOR of 4.29 (3.60-5.11); the association was therefore not an artifact of a permissive outcome definition. The association was consistent across racial and ethnic groups (White, Black, and Hispanic participants) and across mental-health strata (present and absent), indicating that the effect was not explained by psychiatric comorbidity or confined to any demographic subgroup. To address the possibility that adjusting for variables on the causal pathway (opioid use disorder, other substance use, and chronic pain) biased the surgical estimates toward the null, we refit the propensity model excluding these pathway variables. The surgical estimates were essentially unchanged (reoperation aOR 0.95, readmission 1.12, major complications 1.05, pseudarthrosis 1.30, all-cause mortality 1.15), and the primary persistent-use association was identical (aOR 5.40); the wound-complication estimate shifted slightly and became marginally significant (aOR 1.43; 95% CI, 1.02-1.97). For most surgical outcomes, the attenuation of the estimates was therefore unlikely to reflect over-adjustment. Wound complications were an exception: the association crossed the conventional threshold for statistical significance after pathway variables were removed, suggesting that this finding is sensitive to the adjustment strategy and cannot be clearly distinguished from covariates that may lie on the causal pathway.

4. Discussion

In this nationwide cohort of 3424 lumbar fusion patients, chronic preoperative opioid use was consistently associated with persistent postoperative opioid use, yet showed no statistically significant independent association with adverse surgical outcomes after comprehensive adjustment. These paired findings are complementary and, together, refine the prevailing narrative, chronic opioid use is a marker of continued opioid reliance, but the surgical harms often attributed to it appear, at least in part, confounded by the mental-health and comorbidity burden that accompanies chronic use. The magnitude and consistency of the persistent-use association are notable. An adjusted odds ratio of 5.40, which remained between 4.29 and 4.95 under stricter definitions of persistent use, accords closely with prior estimates: preoperative opioid use has repeatedly been shown to increase the odds of prolonged postoperative use several-fold, and duration of preoperative use ranks among the strongest predictors.^5,6,13 Our observed rates (60.8% among chronic users vs 18.5% among non-chronic) fall squarely within published ranges of 35-77% and under 1-26%, respectively. ⁵ That the association held uniformly across White, Black, and Hispanic participants and across mental-health strata strengthens its credibility and argues against explanation by residual psychiatric confounding alone.

The surgical-outcome findings are where our study most informs the literature. Unadjusted associations pointed toward elevated risk of reoperation, readmission, complications, and wound problems, directionally consistent with meta-analytic reports that perioperative opioid use worsens spine surgery outcomes. ⁷ However, these associations attenuated substantially after adjustment and no longer reached statistical significance. We acknowledge that some surgical outcomes were relatively infrequent (e.g., 40 cases of pseudarthrosis and 74 wound events among chronic users), which left these surgical models underpowered. Accordingly, these models are best interpreted as demonstrating the attenuation of crude associations after adjustment rather than as negative studies capable of excluding a potential effect; their non-significant results should not be read as evidence of no effect. Several point estimates continued to trend toward increased risk, most notably for pseudarthrosis (aOR, 1.33; 95% CI, 0.86-2.01) and wound complications (aOR, 1.36; 95% CI, 0.97-1.88). Moreover, the relatively wide confidence intervals indicate uncertainty and do not exclude the possibility of a clinically meaningful independent effect for these outcomes. This consideration is particularly important for wound complications. Although the primary adjusted estimate was not statistically significant (aOR, 1.36; 95% CI, 0.97–1.88), the association reached conventional statistical significance after opioid use disorder, other substance use, and chronic pain were removed from the propensity model (aOR, 1.43; 95% CI, 1.02–1.97). Thus, the apparent association between chronic opioid use and wound complications is sensitive to the adjustment strategy, depending in part on whether these variables are regarded as confounders or as potential mediators on the causal pathway. Because the available data cannot establish their causal roles, the primary null finding should be interpreted cautiously rather than considered definitive.

This pattern is consistent with confounding by indication and by shared causes, chronic opioid users in our cohort had markedly higher burdens of depression and anxiety, substance use, chronic pain, and comorbidity, each an established driver of adverse surgical outcomes.^9-11 With adjustment for these factors, the apparent independent effect of opioids on surgical endpoints diminishes.

Our findings resonate with an emerging reappraisal in the spine literature. Recent work suggests that preoperative opioid misuse, rather than opioid use itself, may be the operative risk factor for adverse opioid trajectories, implying that poorly measured behavioral and psychological dimensions carry much of the signal. ¹⁴ Investigators studying adult spinal deformity have likewise cautioned against treating preoperative opioid consumption as a straightforward causal factor, noting weak correlations between opioid use and measured mental health in some cohorts and emphasizing the difficulty of disentangling these intertwined exposures. ¹⁵ Our results extend this reappraisal by demonstrating, in a large and diverse sample with enhanced confounder measurement, that the surgical-outcome associations are fragile to adjustment while the persistence association is not.

These observations carry practical implications. First, the persistence finding reinforces the value of preoperative identification of chronic opioid users and of perioperative pathways, multimodal analgesia, opioid-sparing protocols, and structured tapering, designed to reduce long-term reliance, an outcome that is both patient-centered and directly modifiable.^16,17 Second, the attenuation of surgical-outcome associations should temper causal claims: counseling that frames chronic opioid use as a direct cause of surgical complications may overstate the evidence. The more defensible message is that chronic opioid users constitute a higher-risk group whose elevated crude event rates reflect a broader constellation of vulnerability, psychiatric, behavioral, and medical, that itself warrants optimization. ¹⁸ Third, the incremental value of survey-based mental-health measurement highlights a limitation of claims-only research and supports the integration of patient-reported psychological data into surgical risk assessment. This advantage should be considered in the context of how the survey data were collected. The survey screens were administered at program enrollment rather than at a standardized interval before surgery, and the time between assessment and surgery varied across participants. Thus, the mental health measure reflects previously diagnosed depression or anxiety and symptom burden at enrollment rather than a specific perioperative assessment. The extent to which it adds information beyond claims-based measures of mental health during the perioperative period remains uncertain.

Mechanistically, several pathways may link chronic opioid use to persistent postoperative use independent of surgical outcome. Opioid-induced hyperalgesia, tolerance, physical dependence, and the reinforcing interplay between chronic pain and mood disorders can sustain opioid consumption across the surgical episode regardless of technical surgical success.^19,20 That persistence was undiminished among patients without diagnosed mental illness may suggest neuroadaptive and pain-related mechanisms operate even absent overt psychiatric comorbidity. Conversely, the absence of an independent effect on structural outcomes such as pseudarthrosis, though the point estimate (aOR 1.33) leaves residual uncertainty given limited events, argues against a strong direct biological effect of opioids on fusion healing in these data, notwithstanding experimental concerns about opioid effects on bone. ²¹

This study has important strengths. It is among the larger analyses of this question and, critically, draws on a diverse population enriched for underrepresented groups, addressing the predominantly homogeneous samples of much prior work. The linkage of electronic health records with validated survey instruments enabled a combined mental-health measure that captured symptomatic patients invisible to diagnostic codes. The prespecified balanced outcome panel, dual exposure definitions, and consistent subgroup findings lend robustness.

Limitations warrant emphasis. First, despite enhanced confounder measurement, residual and unmeasured confounding remains plausible. Opioid use disorder, included as a covariate, overlaps conceptually with the exposure and may induce over-adjustment, while unmeasured pain severity and functional status could bias estimates. Baseline pain severity and disability measures, such as visual analog scale and Oswestry Disability Index scores, were unavailable in this database; therefore, we cannot fully distinguish chronic opioid use as a behavioral exposure from its role as a marker of more severe structural or nociceptive disease, despite adjustment for the underlying surgical indication. Mental-health measures were collected at enrollment rather than uniformly in relation to the index surgery, and the interval between these assessments and surgery may have varied. Consequently, although these measures enhanced adjustment for psychological confounding, they may not have fully captured the patient’s perioperative psychological state. Similarly, bone mineral density (DEXA T-scores), muscle mass, and sarcopenia were not captured in structured form due to unavailability in the database. Osteoporosis diagnosis and anti-osteoporosis therapy therefore served only as partial proxies for bone quality, which is particularly relevant to interpretation of the pseudarthrosis outcome.

Second, the persistent-use outcome is partly definitional: patients using opioids chronically before surgery are, by disposition, more likely to continue, so this association should be interpreted as confirmation and quantification rather than discovery of a novel effect. Third, opioid exposure was classified categorically, and morphine milligram equivalent dose could not be calculated because dose was unavailable in the database leaving “chronic use” a binary construct. Exposure and outcomes were also ascertained from prescription and diagnostic records, which capture dispensing rather than actual consumption and may undercount events occurring outside contributing institutions. Finally, surgical-outcome analyses, although adequately powered to detect large effects, cannot exclude modest true associations given the width of some confidence intervals, and residual confounding by perioperative psychological state and other unmeasured factors remains possible.

5. Conclusion

Among adults undergoing lumbar fusion, chronic preoperative opioid use was strongly and robustly associated with persistent postoperative opioid use but showed no statistically significant independent association with adverse surgical outcomes after comprehensive adjustment incorporating survey-based mental-health data, although wide confidence intervals for pseudarthrosis and wound complications do not exclude a clinically meaningful effect. The surgical models were, moreover, underpowered, so their non-significant results should be read as attenuation of the crude associations after adjustment rather than as evidence of no effect. These findings suggest that surgical risks attributed to chronic opioid use may in part reflect confounding by accompanying psychiatric. Preoperative identification of chronic opioid users remains valuable, chiefly to target strategies that reduce long-term opioid reliance, but causal framing of opioids as a direct driver of surgical complications should be approached with caution pending confirmation in studies with granular pain, functional, and behavioral data.

ORCID iDs

Atef F. Hulliel https://orcid.org/0009-0000-0121-6397

Dana Saleh https://orcid.org/0009-0005-1894-5373

Ahmad Zayd Alkadri https://orcid.org/0009-0008-6625-216X

Srujan Kopparapu https://orcid.org/0000-0002-5862-7063

Mohsen Rostami https://orcid.org/0000-0003-1296-9751

Data Availability Statement

Data from the All of Us Research Program are accessible only through the Researcher Workbench as stipulated in the informed consent of participants in the program.*