Proton pump inhibitors are associated with increased risk of pseudoarthrosis after lumbar fusion

Abstract

Background

Pseudoarthrosis is a major complication following lumbar spine fusion, leading to revision surgery, increased disability, and substantial economic burden. Proton pump inhibitors (PPIs) are commonly used in the perioperative period, but their impact on fusion success remains poorly defined.

Methods

This retrospective cohort study utilized the TriNetX research database to identify adult patients undergoing lumbar spine fusion. Patients were divided into 2 cohorts based on documented PPI use within 30 days postoperatively and underwent 1:1 propensity score matching based on demographic and clinical characteristics. The incidence of pseudoarthrosis was assessed using diagnostic coding, and risk was compared using risk ratios and Kaplan-Meier survival analysis. An age-stratified analysis was also performed.

Results

A total of 122,584 matched pairs were analyzed. Pseudoarthrosis occurred in 7.38% of PPI users compared to 3.98% of nonusers (RR: 1.88; p < .0001). Kaplan-Meier survival analysis revealed early separation of pseudoarthrosis-free survival curves by postoperative day 60, persisting through 1-year follow-up. Age-stratified analysis showed consistently higher pseudoarthrosis risk among PPI users across all age groups, with the largest relative impact observed in patients under 55 years old.

Conclusions

Postoperative PPI use is associated with a significantly increased risk of pseudoarthrosis following lumbar spine fusion. These findings suggest careful consideration of the necessity of PPI therapy in the perioperative period, particularly in younger patients where the relative impact appears the greatest.

Introduction

Pseudoarthrosis is among the most consequential complications following spinal fusion [[1], [2], [3]], often necessitating revision surgery, prolonging disability, and increasing economic burden [[4], [5], [6]]. While surgical technique and implant choice are important, successful fusion ultimately depends on host conditions that promote osseous fusion across the intended levels [[7], [8], [9]]. Fusion imposes significant metabolic and angiogenic demands on the biologic environment, requiring tightly coordinated osteogenesis, bone remodeling, and neovascularization [10].

Various systemic factors are known to impair spinal fusion outcomes. Diabetes mellitus has been associated with impaired angiogenesis and altered osteoblast function, driving elevated rates of pseudoarthrosis [11,12]. Smoking [13], obesity [14], and chronic kidney disease [15] have also been implicated; however, findings regarding obesity remain conflicting. These conditions may contribute through mechanisms such as chronic inflammation, vascular changes, and disturbed mineral homeostasis. Additionally, disorders of bone metabolism such as osteoporosis and osteopenia may negatively impact graft integration and structural integrity of the fusion [7,16]. Despite optimization of surgical technique, these intrinsic variables continue to present major challenges to successful fusion.

Proton pump inhibitors (PPIs), widely used for gastrointestinal protection and available over the counter, are frequently administered in the postoperative setting [17,18]. Their broad accessibility and generally favorable safety profile have contributed to routine use in surgical patients, often without close scrutiny of potential effects on bone healing. By suppressing gastric acid secretion, PPIs impair calcium absorption and disrupt magnesium and vitamin B12 homeostasis [19]. Basic and translational studies have also linked PPI use to suppressed osteoclast activity, reduced bone turnover, and delayed fracture healing [20].

While the adverse skeletal effects of PPIs have been documented in the context of fracture risk [21], their impact on spinal fusion remains incompletely characterized. Prior studies in cervical spine have identified higher pseudoarthrosis rates among PPI users after anterior cervical discectomy and fusion (ACDF) [22,23], and more recent investigations have reported similar findings in lumbar fusion populations [24]. However, these analyses were limited by smaller sample sizes, single-level procedures, and lack of age-stratified assessment. Given the widespread use of PPIs and the need for large-scale evaluation across a broader lumbar fusion population, further investigation remains warranted.

In this study, we examined the association between postoperative PPI use and pseudoarthrosis following lumbar spinal surgery. We also conducted an age-stratified analysis to determine whether certain patient populations may be more susceptible to PPI-related disruption in bone healing.

Methods

Study design and database

This study was a retrospective cohort analysis utilizing the TriNetX US Collaborative Network, a standardized database that aggregates electronic health records from healthcare organizations across the United States, encompassing over 117 million patients. TriNetX adheres to HIPAA regulations for data security and confidentiality, and all datasets are fully deidentified, making this study exempt from IRB approval.

Cohort selection

Patients aged 18 years or older who underwent lumbar fusion between 2015 and 2023 were identified using ICD and CPT codes (Supplementary Table 1). Procedures included anterior, posterior, lateral, and combined interbody fusion techniques, with or without decompression. Patients were categorized based on documented PPI use within 30 days postoperatively, with the date of fusion serving as the index event.

Exposure classification

PPI exposure was identified using RxNorm codes available within the TriNetX platform, which capture structured electronic health record data on medication prescriptions and administration. Patients were included in the PPI group if that had a documented order or administration of any PPI-class medication (esomeprazole, rabeprazole, dexlansoprazole, omeprazole, pantoprazole, lansoprazole) within 30 days following the date of fusion. This timeframe reflects a biologically relevant period of early fusion healing [25] and aligns with standard postoperative prescribing practices for gastroprotection [17,18].

The PPI cohort comprised patients with documented PPI use within 30 days postoperatively; continued or delayed use beyond this timeframe could not be consistently captured in TriNetX. Non-PPI patients had no documented PPI use within 30 days following surgery.

Outcomes

The primary outcome was risk of pseudoarthrosis. Pseudoarthrosis was identified using ICD codes reflecting pseudoarthrosis or failed spinal fusion, assessed up to 1 year postoperatively to capture delayed healing events [26,27]. To further evaluate differential risk across patient populations, an age-stratified analysis of pseudoarthrosis was performed using 4 groups: <50, 50–65, 66–75, and >75 years. Patients without available follow-up data during this 1-year window were not included in the outcome analysis, as TriNetX automatically limits event capture to individuals with documented postoperative encounters within the specified timeframe.

Propensity score matching

To minimize confounding, 1:1 propensity score matching was performed using TriNetX’s nearest-neighbor algorithm. Matching covariates included age, sex, race/ethnicity, comorbidities (diabetes, obesity, inflammatory polyarthropathy), bone health conditions, and medications known to influence bone metabolism (NSAIDs, Denosumab, Vitamin D, Alendronate).

Covariate balance was assessed using standardized mean differences (SMD), with values <0.1 indicating adequate balance between groups. While p-values remain statistically significant in large matched datasets due to high power, SMD provides a more appropriate measure of clinical similarity as it is independent of sample size. postmatching cohort characteristics are detailed in Table 1.

Table 1.

Baseline demographic and clinical characteristic of propensity score-matched PPI and non-PPI cohorts.

Characteristic	PPI (N=122,584)	Non-PPI (N=122,584)	p-value	SMD
Sex
Male	54,136 (44.162%)	55,066 (44.921%)	.0002	0.0153
Female	62,781 (51.215%)	62,242 (50.775%)	.0294	0.0088
Race and ethnicity
Hispanic	6,931 (45.442%)	6,083 (4.776%)	.0001	0.0303
Asian	7,005 (5.5%)	6,886 (5.407%)	.2991	0.0041
American Indian/Alaskan	363 (0.228%)	321 (0.174%)	.002	0.0121
Black	10,666 (8.375%)	10,568 (8.298%)	.4824	0.0028
Native Hawaiian	291 (0.228%)	221 (0.174%)	.0020	0.0121
White	92,801 (72.867%)	94,126 (73.908%)	.4740	0.0236
Unknown race	12,413 (9.747%)	11,808 (9.272%)	.0001	0.0162
Comorbidities
Obesity	38,035 (29.865%)	37,872 (29.737%)	.4081	0.0028
Diabetes	31,897 (25.046%)	39,855 (24.227%)	.0001	0.0190
Inflammatory polyarthropathy	26,586 (20.875%)	25,763 (20.229%)	.0085	0.0160
Disorders of bone density	25,257 (19.832%)	24,147 (18.96%)	.0001	0.0221
Medications
NSAID	32,856 (25.799%)	32,608 (25.604%)	.2608	0.0045
Vitamin D3	3,872 (3.04%)	3,242 (2.546%)	.0001	0.0300
Alendronate	3,829 (3.007%)	3,555 (2.791%)	.0012	0.0129
Denosumab	1,822 (1.431%)	1,669 (1.31%)	.0091	0.0104

Statistical analysis

Absolute risk was calculated for each cohort and expressed as event counts and corresponding percentages (Fig. 1). Risk ratios (RR) and hazard ratios (HR) were derived from Cox proportional hazard models to evaluate relative risk of pseudoarthrosis over time (Fig. 1). Kaplan-Meier survival curves were generated to visualize pseudoarthrosis-free survival after 1 year (Fig. 2). Statistical significance was assessed using log-rank test for survival analysis, chi-square, or Fisher’s exact test for categorical variables and Student t-tests for continuous variables. A significance threshold was applied, and all analyses were conducted using the TriNetX analysis platform.

Fig. 1.

Bar plot illustrating 1-year incidence of pseudoarthrosis in propensity score-matched cohorts (n = 122,584 per group). Patients receiving PPIs are compared to nonusers. Error bars represent 95% confidence intervals; **** indicates p < .0001.

Fig. 2.

Kaplan-Meier analysis of pseudoarthrosis-free survival in patients who received PPI within 30 days of lumbar spinal fusion compared to propensity score-matched nonusers.

Results

Demographics

A total of 245,168 patients undergoing lumbar spine fusion were identified, including 122,584 in the PPI cohort and 122,584 in the non-PPI cohort. Following 1:1 propensity score matching, each group was balanced across baseline characteristics including age, sex, race/ethnicity, and relevant clinical comorbidities (Table 1). Standardized mean differences (SMDs) for all covariates were below 0.1, indicating successful covariate balance.

Incidence of pseudoarthrosis

Pseudoarthrosis occurred significantly more frequently among patients in the PPI cohort, with 7.378% of PPI users diagnosed compared to 3.933% of nonusers (p < .0001) (Fig. 1). Cox proportional hazards modeling confirmed a strong association, with a risk ratio of 1.876 indicating a substantially increased risk of fusion failure among PPI users. This association remained significant after adjusting for relevant clinical and demographic covariates, suggesting that postoperative PPI exposure may be an independent risk factor for impaired fusion healing.

Kaplan-Meier analysis

Kaplan-Meier survival analysis (Fig. 2) demonstrated a clear divergence in pseudoarthrosis free survival between cohorts. As early as 60 days postoperatively, patients in the PPI group exhibited a lower probability of remaining pseudoarthrosis-free compared to nonusers. This gap continued to widen over time, with 1-year pseudoarthrosis-free survival rates of 91.85% in the PPI group compared to 95.62% in the non-PPI group. These results indicate a persistent and time-dependent difference in fusion outcomes associated with postoperative PPI exposure.

Age-stratified analysis

To evaluate whether the association between PPI use and pseudoarthrosis varied by age; patients were stratified into 4 age groups: <55, 56–65, 66–75, and 76 years or older. Across all strata, PPI use was consistently associated with a significantly higher rate of pseudoarthrosis compared to nonuse (Fig. 3A and B). Among patients under 55 years old, the pseudoarthrosis rate was 5.19% in the PPI group versus 2.69% in the non-PPI group (RR:1.93, 95% CI 1.76–2.10, p < .0001). In the 56–65 group, rates were 7.37% versus 3.82% (RR: 1.92, 05% CI: 1.80–2.07, p < .0001). For those aged 66-75, pseudoarthrosis occurred in 7.62% of PPI users and 4.27% nonusers (RR: 1.79, 05% CI 1.69–1.89 p < .0001). Finally, in patients aged 76 and older, rates were 6.35% in the PPI group compared to 3.84% in the non-PPI group (RR: 1.66, 95% CI 1.56–1.76, p < .0001).

Fig. 3.

(A) Absolute pseudoarthrosis rates by age group in PPI users versus nonusers. (B) Relative risk of pseudoarthrosis in PPI users compared to nonusers across age cohorts. (C) Relative risk of pseudoarthrosis within the PPI cohort, comparing older cohorts to those under 55 (reference group). * Error bars represent 95% confidence intervals.

Although the absolute rate of pseudoarthrosis increased with age, the relative risk conferred by PPI exposure was highest in younger patients. To further explore this, we compared pseudoarthrosis rates between age groups within the PPI cohort (Fig. 3C). Compared to PPI users under 55, those aged 55-65 had a significantly higher pseudoarthrosis rate of 6.97% versus 5.36% (RR: 1.30, 95% CI 1.21–1.40, p < .0001). Similarly, those aged 66-75 had a rate of 6.53% versus 5.36 (RR: 1.21, 95% CI 1.13–1.30, p < .0001), and patients 76 and older had a rate of 5.91% versus 5.36% (RR:1.09, 95% CI 1.01–1.18, p = .0317). These findings suggest that while increasing age is associated with higher absolute rates of fusion failure, the relative impact of PPI use remains significant across all age groups and appears most pronounced in younger, lower-risk individuals, where even modest absolute increases in pseudoarthrosis translate to a disproportionately elevated risk.

Discussion

Our findings demonstrate a significant association between postoperative PPI use and increased risk of pseudoarthrosis following lumbar spinal fusion. Patients with documented PPI prescription within 30 days of surgery exhibited nearly double the rate of pseudoarthrosis compared to control patients without such documentation. Kaplan-Meier survival analysis revealed an early and sustained divergence in pseudoarthrosis-free survival between groups, with separation emerging by postoperative day 60 and widening throughout the 1-year follow-up. These findings suggest that PPI exposure during the early postoperative period may negatively impact spinal fusion healing.

Prior investigations have begun to explore the potential relationship between PPI use and impaired spinal fusion, though primarily in smaller and anatomically limited cohorts. Mangan et al. first identified an association between PPI use and increased pseudoarthrosis rates following anterior cervical discectomy and fusion (ACDF) [22] in a single-center analysis and a subsequent study by Chang et al. confirmed this relationship in a larger database cohort of single-level ACDF patients [23]. More recently, another study by Chang et al. expanded PPI exposure in the context of single level lumbar fusion and likewise reported elevated nonunion risk among PPI users [24]. Our study complements and expands upon these findings by analyzing a substantially larger national cohort (over 120,000 PPI users and match controls) and incorporating age-stratified analysis that highlight how the relative effect of PPI exposure varies across patient age group. In contrast to the prior single-level design, our work captures outcomes within a more heterogenous real-world lumbar fusion population, offering broader generalizability of this association.

The mechanisms underlying this association are supported by prior preclinical and translational research. PPIs have been shown to impair calcium and magnesium absorption [28] suppress osteoclast activity [29], and inhibit angiogenesis [30]. Successful spinal fusion is dependent on coordinated osteoblast-mediated bone formation, osteoclast-driven remodeling of graft, and angiogenesis to support nutrient delivery across the fusion interface [31]. Because spinal fusion depends on both mechanical stability [30] and biologic integration of the graft, including revascularization and remodeling, even subtle disruptions in osteogenic processes may increase the risk for pseudoarthrosis.

Age-stratified analysis further demonstrates that the relative effect of PPI use on pseudoarthrosis was most pronounced in younger patients. While absolute pseudoarthrosis rates increased with age, the highest relative risk was observed in patients under 55, where baseline healing potential is typically considered more robust [32]. In this group, PPI exposure nearly doubled the risk of pseudoarthrosis compared to nonusers. In contrast, older patients demonstrated higher absolute rates of pseudoarthrosis but a smaller relative impact from PPI exposure, likely reflecting the contribution of broader age-related declines in bone turnover, vascularity, and healing potential that may obscure the effect of a single pharmacologic factor [32]. The disproportionate effect observed in younger patients may reflect the greater reliance on tightly coupled osteoblast-osteoclast activity in this population, where baseline risk for fusion failure is otherwise low and even modest disruptions to bone remodeling may carry outsized consequences.

Several limitations inherent to retrospective database studies apply to this analysis. PPI exposure was defined using electronic health record (EHR) data on prescriptions and medication administration, which may not accurately reflect true patient adherence. Additionally, because PPIs are available over the counter, nonprescription may be underreported unless documented during clinical encounters, potentially leading to an underestimation of total PPI exposure. The PPI cohort comprised patients with documented PPI use within 30 days postoperatively; continued or delayed use beyond this timeframe could not be consistently captured in TriNetX. Non-PPI patients had no recorded PPI prescriptions within this 30-day window, though some may have initiated use later in follow-up. This limitation likely biases the observed association toward, rather than away from, the null. While 1:1 propensity score matching was used to minimize confounding, residual confounders, indications for PPI use including bone quality or surgical technique may still influence the observed associations.

Stratification by specific surgical approach or concomitant decompression was not performed, as this information is inconsistently captured within TriNetX and such heterogeneity in CPT and operative coding is a commonly noted limitation in large database studies. Further subdivision by these procedural factors markedly reduced subgroup sizes, limiting interpretability and statistical power. Pseudoarthrosis was identified through diagnostic coding, which may vary in accuracy across institutions, though this method aligns with established practices in large-scale database studies. Finally, we were unable to assess PPI dose or duration due to TriNetX limitations. Prospective studies with more granular clinical, imaging, and medication data are needed to clarify the nature and magnitude of this association.

Conclusion

Our findings reveal a significant association between postoperative PPI use and increased risk of pseudoarthrosis following lumbar spinal fusion, with the greatest relative impact observed in younger, lower-risk patients. Kaplan-Meier analysis demonstrated an early and sustained divergence in pseudoarthrosis-free survival between PPI users and nonusers, suggesting a potential time-dependent relationship between PPI exposure and impaired fusion healing. While further prospective studies are needed to explore underlying mechanisms, these findings underscore the importance of carefully weighing the risks and benefits of routine PPI use in the perioperative spine population.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of competing interest

One or more of the authors declare financial or professional relationships on ICMJE-NASSJ disclosure forms.