Investigating the Relationship Between Estrogen Replacement Therapy and Postoperative Complications Following Single-Level Lumbar Fusion: A Propensity Matched Analysis Authors: Kyle B. Stump, Alec M. Giakas, Gregorio Baek, Mitchell K. Ng, Andrew P. Alvarez, Alan S. Hilibrand, Alexander R. Vaccaro, Gregory D. Schroeder, Christopher K. Kepler Published: 2026 ABSTRACT Study DesignRetrospective Cohort.ObjectivesThe objective of this study was to investigate the relationship between preoperative estrogen replacement therapy and postoperative complications following single level lumbar fusion.MethodsThe TriNetX Research Network was queried for menopausal patients aged ≥ 50 years who underwent single level lumbar fusion. Patients issued ERT within the 12-month preoperative period were separated from patients with no documented ERT use. Propensity score matching was performed 1:1 based upon demographic characteristics and medical comorbidities. Rates of medical complications were assessed within 90 days postoperatively. Secondary outcomes include pseudarthrosis and revision surgery, which were evaluated within one and two years postoperatively. P < 0.05 indicates statistical significance.ResultsA total of 1,570 matched pairs were included. The incidence of PE was more than doubled in the ERT-positive cohort (RR 2.167, 95% CI 1.097 – 4.279, P=0.022). Rates of pseudarthrosis were elevated in the ERT-positive cohort within one (12.9% versus 9.9%, RR 1.303, 95% CI 1.070 – 1.587, P=0.008) and two years postoperatively (14.8% versus 11.6%, RR 1.275, 95% CI 1.064 – 1.528, P=0.008); reoperation rates were similar at both intervals. ERT use was also associated with increased rates of ED utilization within 90 days postoperatively (15.4% versus 11.1%, RR 1.377, 95% CI 1.148 – 1.651, P=0.001).ConclusionsPatients on preoperative ERT may represent a higher-risk cohort amongst single-level lumbar fusion candidates. Further studies are warranted to explore causality and the potential mechanistic underpinnings between ERT and unsuccessful bone healing and thromboembolic complications. Introduction Often considered for the treatment of spinal stenosis or spondylolisthesis that is refractory to conservative treatment, lumbar fusion is one of the most frequently performed surgical procedures involving the spine.1-4 Lumbar fusion volumes are only projected to rise in the coming years, driven in part by an aging global population. 1 This dynamic necessitates careful candidate selection due to higher rates of medical comorbidity and polypharmacy in elderly patients. To this end, one of the most relevant characteristics known to influence postoperative outcomes following lumbar fusion is the underlying bone quality and mineralization. Age-related osteoporosis is associated with increased risk of cage subsidence, screw loosening, and pseudarthrosis amongst lumbar fusion candidates, each of which can require reoperation and additional healthcare costs.5,6 Likewise, the use of certain medications shown to compromise bone density and healing such as selective serotonin reuptake inhibitors (SSRI), proton pump inhibitors (PPI), and non-steroidal anti-inflammatory drugs (NSAID), are all associated with increased incidence of unsuccessful bony union following lumbar fusion.7-10 For these reasons, current guidelines by the Congress of Neurological Surgeons (CNS) recommend preoperative osteoporosis testing via dual-energy X-ray absorptiometry (DEXA), computerized tomography (CT) scan, and serum calcitriol. 11 Additionally, there is evidence to suggest that teriparatide may improve bone mineral density (BMD) in the preoperative period to mitigate the risk of hardware failure in these patients. 11 Likewise, given the known association between sex hormone status and BMD, the topic of hormone replacement therapy amongst lumbar fusion candidates may also be relevant to the patient selection processes. 12 Estrogen replacement therapy (ERT) is indicated for the first line treatment for both vasomotor and genitourinary symptoms in menopausal women. 13 Further, while the prescription of ERT for osteoporosis in the absence of menopausal symptoms is not advised, exogenous estrogen with or without progesterone offers robust improvement in BMD that confers 34% and 29% reductions in vertebral and hip fractures, respectively.12,14 In spite of this clinical benefit, utilization of ERT remains low and has actually demonstrated a downward trend in recent years, likely driven by safety concerns.15-17 ERT has been associated with elevated risk of both venous thromboembolism and cardiovascular disease.15,16 However, the Food and Drug Administration (FDA) recently approved a label change in February 2026 that clarified these safety considerations. 18 Thus, rates of ERT are expected to trend higher in the coming years and it is crucial for practitioners to understand the implications of its use particularly amongst elective orthopaedic surgery candidates. While a paucity of literature exists regarding the association between ERT and postoperative complications following lumbar fusion, initial investigations of its use in the total joint arthroplasty (TJA) population demonstrate conflicting results. For example, a retrospective database by McCormick et al found no association between hormone replacement therapy (HRT) and venous thromboembolic risk in female patients who underwent total hip arthroplasty (THA). 19 This result was replicated by a similarly designed database study investigating ERT specifically and postoperative complications following THA. 20 In contrast, a more recent investigation by Mousavi et al reported increased deep vein thrombosis (DVT) risk amongst ERT users within 90 days following total shoulder arthroplasty (TSA). 21 In addition, preoperative ERT use is associated with increased rates of aseptic prosthesis loosening, other mechanical complications, and revision surgery in both TSA and THA populations.19-21 With this in mind, it is essential to investigate whether ERT use confers increased VTE risk following lumbar fusion, which represents a fundamentally distinct procedure from TJA. While these implant-related complications associated with ERT are somewhat paradoxical given the known benefit to BMD but is nonetheless incredibly relevant to outline the effect of ERT on lumbar fusion complications regardless of the direction of the association. 17 Therefore, the objective of this study was to investigate the relationship between preoperative ERT and postoperative outcomes following single-level lumbar fusion. We hypothesize that patients issued ERT in the one year prior to lumbar fusion will experience greater risk of thromboembolic events in the near-term whereas the likelihood of pseudarthrosis and revision surgery will be lower than that compared to matched controls. Materials and Methods Data Source This retrospective cohort analysis utilized the TriNetX Research Network database (Cambridge, Massachusetts), a real-time healthcare data platform that aggregates de-identified electronic medical record information from over 120 healthcare organizations (HCOs) worldwide. 22 Participating HCOs authorize the use of their medical record data in exchange for permission to use the platform’s research capabilities, which include clinical trial site analysis, database query, and drug safety evaluation. Ethical Approval The TriNetX Research Network is compliant with the Health Insurance Portability and Accountability Act of 1996 (HIPAA) Security Rule and certified according to the ISO 27001:2013 standard. For these reasons, this study was deemed exempt from institutional review board approval and informed consent at our institutions. The analysis was performed and is reported in accordance with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Cohort Selection All data included in this analysis were obtained on April 5, 2026, from the TriNetX Research Network database within the ten-year period beginning April 4, 2014, and ending April 4, 2024. This interval was specifically selected to ensure that at least two years of follow-up data would be available for each patient included in the analysis. Inclusion criteria, exclusion criteria, propensity score matching criteria, and study endpoints utilized classifications outlined by the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM), International Classification of Diseases, Tenth Revision, Procedural Coding System (ICD-10-PCS), Current Procedural Terminology (CPT), and Anatomical Therapeutic Chemical (ATC) medical billing codes. Patient cohorts were first established by querying the TriNetX database for female patients aged at least 50 years who underwent single-level lumbar fusion of any operative approach (CPT 22558, 22630, 22633, or 22612; ICD-10-PCS 0SG00AJ, 0SG007J, 0SG00KJ, 0SG00JJ, 0SG03KJ, or 0SG037J). We did not differentiate patients by surgical indication for lumbar fusion, as administrative claims databases are well recognized to inaccurately capture and code clinical indications, which would increase the likelihood of misclassification bias. 23 Additional inclusion criteria was the presence of at least one ICD-10-CM billing code indicating the onset of menopause (ICD-10-CM N95, E28.31, or Z78.0). Patients who underwent multi-level lumbar fusion (CPT 22585, 22634 or 22632) or had a previous history of primary bone malignancy or secondary bone metastasis were excluded. Estrogen is known to influence tumor biology, and thus, guidelines for the use of hormone therapy in oncologic patients differ substantially from those for the general population which may confound results.24-26 Once the inclusion and exclusion criteria were applied, patients who received estrogen replacement therapy (ATC G03C) within the one year prior to surgery, denoting the experimental group, were separated from controls without documented ERT at any time prior to or up to two years following the index procedure (Figure 1). We deliberately excluded postoperative ERT exposure in the control group to mitigate the risk of misclassification bias as the assignment of patients with postoperative ERT initiation to the control group may dilute or confound any potential findings. The ATC G03C classification encompassed natural and semisynthetic estrogens (estradiol, estetrol, estriol, conjugated estrogens, ethinyl estradiol, estrone), synthetic estrogens (dienestrol, diethylstillbestrol), and tibolone. Selective estrogen receptor modulators (SERMs), progesterone, and other menopausal pharmacotherapies were not included under this designation. Information regarding dose, duration of therapy, and route of administration was unable to be obtained. Of the 12,484 single-level lumbar fusion candidates meeting inclusion criteria, 1,680 patients comprised the initial experimental cohort. There were 10,804 patients in the control group.Figure 1.Patient Selection and Cohort Design. Legend – N = number of patients. Abbreviations – ERT, estrogen replacement therapy, PSM, propensity-score matching; BMI, body mass index; HbA1c, hemoglobin A1c Patient Selection and Cohort Design. Legend – N = number of patients. Abbreviations – ERT, estrogen replacement therapy, PSM, propensity-score matching; BMI, body mass index; HbA1c, hemoglobin A1c Propensity Score Matching Next, the two cohorts were prepared for 1:1 nearest-neighbor propensity score matching to limit the potential for residual confounding. The characteristics chosen for matching included age at the time of procedure, race, ethnicity, body mass index (BMI), hemoglobin A1c (HbA1c), essential hypertension, nicotine dependence, diabetes mellitus, heart failure, hyperlipidemia, osteoporosis, long-term systemic steroid use, and coagulation defects, including both thrombophilia and hemorrhagic disorders. Successful cohort matching was denoted by a standard difference (std diff) less than or equal to 0.1 for each of the selected covariates. Outcomes of Interest Primary outcomes included the incidence of DVT and pulmonary embolism within 90 days following single-level lumbar fusion, as well as pseudarthrosis (ICD-10-CM M96.0) which was assessed within one and two years postoperatively. Additional secondary outcomes included 90-day rates of acute renal failure, pneumonia, sepsis, emergency department (ED) utilization, and all-cause inpatient hospitalization (assessed from postoperative day five) in addition to the incidence of reoperation which was evaluated at one and two year postoperative intervals. Statistical Analysis Baseline characteristics were evaluated using descriptive statistics. Binary outcomes for each endpoint were assessed with relative risk (RR) and 95% confidence intervals (CI). All statistical procedures were automatically performed within the TriNetX research platform. A two-sided P-value < 0.05 was considered to be statistically significant. Results A total of 1,570 matched pairs were included in the final analysis. The mean age at the time of index procedure was 64.3 ± 9.1 years in the ERT-use cohort compared with 63.9 ± 9.4 years in the ERT-naïve group (std diff = 0.042). More than four-fifths of all included patients were White (81.9%), whereas approximately 8.8% identified as Black or African American and 3.0% as Hispanic or Latino (Table 1). After propensity score matching, rates of all medical comorbidities were balanced between the two groups. The mean BMI was 29.5 ± 6.2 kg/m2 in the ERT-use cohort compared with 29.6 ± 6.3 kg/m2 in matched controls (std diff = 0.002). The mean HbA1c was approximately 6.0% in both groups (std diff = 0.024). Successful balance was also achieved across all individual BMI and HbA1c strata (Table 2).Table 1.Baseline Demographics and Medical ComorbiditiesCharacteristicBefore MatchingAfter MatchingERT (+)ERT (-)P-ValueStd diffERT (+)ERT (-)P-ValueStd diff Demographics Age at Index64.2 (9.2)65.3 (9.4)<0.0010.12064.3 (9.1)63.9 (9.4)0.2410.042Black or African American141 (8.4)1,352 (12.5)<0.0010.135138 (8.8)137 (8.7)0.9500.002White1,381 (82.2)8,004 (74.1)<0.0010.1971,276 (81.3)1,296 (82.5)0.3540.033Hispanic or Latino48 (2.9)423 (3.9)0.0340.05947 (3.0)47 (3.0)1<0.001 Medical Comorbidities Essential Hypertension966 (57.5)6,265 (58.0)0.7060.010904 (57.6)869 (55.4)0.2080.045Nicotine Dependence138 (8.2)1,139 (10.5)0.0030.080136 (8.7)134 (8.5)0.8990.005Diabetes Mellitus324 (19.3)2,674 (24.8)<0.0010.132316 (20.1)290 (18.5)0.2400.042Chronic Kidney Disease145 (8.6)1,037 (9.6)0.2080.034137 (8.7)145 (9.2)0.6180.018Heart Failure64 (3.8)478 (4.4)0.2500.03163 (4.0)68 (4.3)0.6550.016Coagulation Defects42 (2.5)234 (2.2)0.3860.02239 (2.5)28 (1.8)0.1740.048Osteoporosis without Current Pathological Fracture243 (14.5)1,587 (14.7)0.8090.006233 (14.8)191 (12.2)0.0280.078Hyperlipidemia572 (34.0)3,565 (33.0)0.3950.022542 (34.5)520 (33.1)0.4070.030Long-term Use of Systemic Steroids67 (4.0)312 (2.9)0.0140.06058 (3.7)49 (3.1)0.3760.032Legend - Patient counts represented as n (% of cohort); std diff <0.1 after matching reflects no difference between the two groups. Abbreviations – ERT, estrogen replacement therapy; std diff, standard differenceTable 2.Body Mass Index and Hemoglobin A1c StrataCategoryBefore MatchingAfter MatchingERT (+)ERT (-)P-ValueStd diffERT (+)ERT (-)P-ValueStd diff Body Mass Index (kg/m 2 ) Mean 29.3 (6.1)30.8 (6.7)<0.0010.23529.5 (6.2)29.6 (6.3)0.9620.002<18.535 (2.1)232 (2.1)0.8660.00434 (2.2)33 (2.1)0.9020.00418.5-24.9452 (26.9)2,168 (20.1)<0.0010.162398 (25.4)402 (25.6)0.8700.00625.0-29.9655 (39.0)3,526 (32.6)<0.0010.133590 (37.6)565 (36.0)0.3550.03330.0-34.9522 (31.1)3,378 (31.3)0.8730.004490 (31.2)478 (30.4)0.6430.01735.0-39.9270 (16.1)2,262 (20.9)<0.0010.126262 (16.7)252 (16.1)0.6300.017>40.0119 (7.1)1,215 (11.2)<0.0010.145118 (7.5)114 (7.3)0.7850.010 Hemoglobin A1c (%) Mean 6.0 (1.0)6.2 (1.2)<0.0010.1886.0 (1.0)6.0 (1.3)0.6690.024<5.7306 (18.2)1,411 (13.1)<0.0010.142271 (17.3)258 (16.4)0.5350.0225.7-6.4299 (17.8)1,841 (17.0)0.4430.020281 (17.9)257 (16.4)0.2560.041>6.5173 (10.3)1,376 (12.7)0.0050.076169 (10.8)149 (9.5)0.2370.042Legend - Patient counts represented as n (% of cohort); std diff <0.1 after matching reflects no difference between the two groups. Abbreviations – ERT, estrogen replacement therapy; std diff, standard difference. Table 1.Baseline Demographics and Medical ComorbiditiesCharacteristicBefore MatchingAfter MatchingERT (+)ERT (-)P-ValueStd diffERT (+)ERT (-)P-ValueStd diff Demographics Age at Index64.2 (9.2)65.3 (9.4)<0.0010.12064.3 (9.1)63.9 (9.4)0.2410.042Black or African American141 (8.4)1,352 (12.5)<0.0010.135138 (8.8)137 (8.7)0.9500.002White1,381 (82.2)8,004 (74.1)<0.0010.1971,276 (81.3)1,296 (82.5)0.3540.033Hispanic or Latino48 (2.9)423 (3.9)0.0340.05947 (3.0)47 (3.0)1<0.001 Medical Comorbidities Essential Hypertension966 (57.5)6,265 (58.0)0.7060.010904 (57.6)869 (55.4)0.2080.045Nicotine Dependence138 (8.2)1,139 (10.5)0.0030.080136 (8.7)134 (8.5)0.8990.005Diabetes Mellitus324 (19.3)2,674 (24.8)<0.0010.132316 (20.1)290 (18.5)0.2400.042Chronic Kidney Disease145 (8.6)1,037 (9.6)0.2080.034137 (8.7)145 (9.2)0.6180.018Heart Failure64 (3.8)478 (4.4)0.2500.03163 (4.0)68 (4.3)0.6550.016Coagulation Defects42 (2.5)234 (2.2)0.3860.02239 (2.5)28 (1.8)0.1740.048Osteoporosis without Current Pathological Fracture243 (14.5)1,587 (14.7)0.8090.006233 (14.8)191 (12.2)0.0280.078Hyperlipidemia572 (34.0)3,565 (33.0)0.3950.022542 (34.5)520 (33.1)0.4070.030Long-term Use of Systemic Steroids67 (4.0)312 (2.9)0.0140.06058 (3.7)49 (3.1)0.3760.032Legend - Patient counts represented as n (% of cohort); std diff <0.1 after matching reflects no difference between the two groups. Abbreviations – ERT, estrogen replacement therapy; std diff, standard difference Baseline Demographics and Medical Comorbidities Legend - Patient counts represented as n (% of cohort); std diff <0.1 after matching reflects no difference between the two groups. Abbreviations – ERT, estrogen replacement therapy; std diff, standard difference Table 2.Body Mass Index and Hemoglobin A1c StrataCategoryBefore MatchingAfter MatchingERT (+)ERT (-)P-ValueStd diffERT (+)ERT (-)P-ValueStd diff Body Mass Index (kg/m 2 ) Mean 29.3 (6.1)30.8 (6.7)<0.0010.23529.5 (6.2)29.6 (6.3)0.9620.002<18.535 (2.1)232 (2.1)0.8660.00434 (2.2)33 (2.1)0.9020.00418.5-24.9452 (26.9)2,168 (20.1)<0.0010.162398 (25.4)402 (25.6)0.8700.00625.0-29.9655 (39.0)3,526 (32.6)<0.0010.133590 (37.6)565 (36.0)0.3550.03330.0-34.9522 (31.1)3,378 (31.3)0.8730.004490 (31.2)478 (30.4)0.6430.01735.0-39.9270 (16.1)2,262 (20.9)<0.0010.126262 (16.7)252 (16.1)0.6300.017>40.0119 (7.1)1,215 (11.2)<0.0010.145118 (7.5)114 (7.3)0.7850.010 Hemoglobin A1c (%) Mean 6.0 (1.0)6.2 (1.2)<0.0010.1886.0 (1.0)6.0 (1.3)0.6690.024<5.7306 (18.2)1,411 (13.1)<0.0010.142271 (17.3)258 (16.4)0.5350.0225.7-6.4299 (17.8)1,841 (17.0)0.4430.020281 (17.9)257 (16.4)0.2560.041>6.5173 (10.3)1,376 (12.7)0.0050.076169 (10.8)149 (9.5)0.2370.042Legend - Patient counts represented as n (% of cohort); std diff <0.1 after matching reflects no difference between the two groups. Abbreviations – ERT, estrogen replacement therapy; std diff, standard difference. Body Mass Index and Hemoglobin A1c Strata Legend - Patient counts represented as n (% of cohort); std diff <0.1 after matching reflects no difference between the two groups. Abbreviations – ERT, estrogen replacement therapy; std diff, standard difference. Both primary and secondary outcomes are contained in Table 3. The incidence of PE was 1.7% amongst ERT-users compared with 0.8% in the ERT-naïve group (RR 2.167, 95% CI 1.097 – 4.279, P=0.022). The rate of DVT was 2.4% and 2.3% in the experimental and control groups, respectively (RR 1.028, 95% CI 0.653 – 1.617, P=0.906). In contrast, the rate of ED utilization was significantly elevated amongst ERT-users (15.4% versus 11.1%, RR 1.377, 95% CI 1.148 – 1.651, P=0.001). Frequencies of all other 90-day complications were similar between the two groups.Table 3.Primary and Secondary OutcomesOutcomesRisk in ERT (+)Risk in ERT (-)Risk DifferenceRisk Ratio (95% CI)P-value 90 Days Pulmonary Embolism26 (1.7%)12 (0.8%)0.9%2.167 (1.097 – 4.279) 0.022 Deep Vein Thrombosis37 (2.4%)36 (2.3%)0.1%1.028 (0.653 – 1.617)0.906 Acute Renal Failure41 (2.6%)34 (2.2%)0.4%1.206 (0.770 – 1.890)0.413 Pneumonia24 (1.5%)22 (1.4%)0.1%1.091 (0.614 – 1.937)0.766 Sepsis18 (1.1%)26 (1.7%)-0.6%0.692 (0.381 – 1.258)0.225 ED Utilization241 (15.4%)175 (11.1%)4.2%1.377 (1.148 – 1.651) 0.001 Inpatient Hospitalization211 (13.7%)190 (12.4%)1.3%1.111 (0.925 – 1.333)0.261 One Year Pseudarthrosis202 (12.9%)155 (9.9%)3.0%1.303 (1.070 – 1.587) 0.008 Reoperation149 (9.5%)141 (9.0%)0.5%1.057 (0.849 – 1.316)0.622 Two Years Pseudarthrosis232 (14.8%)182 (11.6%)3.2%1.275 (1.064 – 1.528) 0.008 Reoperation208 (13.2%)183 (11.7%)1.6%1.137 (0.944 – 1.369)0.177Legend – risk is the number of patients with the outcome (proportion of cohort). Bold-face type indicates statistical significance. Abbreviations – ERT, estrogen replacement therapy; ED, emergency department; CI, confidence interval. Table 3.Primary and Secondary OutcomesOutcomesRisk in ERT (+)Risk in ERT (-)Risk DifferenceRisk Ratio (95% CI)P-value 90 Days Pulmonary Embolism26 (1.7%)12 (0.8%)0.9%2.167 (1.097 – 4.279) 0.022 Deep Vein Thrombosis37 (2.4%)36 (2.3%)0.1%1.028 (0.653 – 1.617)0.906 Acute Renal Failure41 (2.6%)34 (2.2%)0.4%1.206 (0.770 – 1.890)0.413 Pneumonia24 (1.5%)22 (1.4%)0.1%1.091 (0.614 – 1.937)0.766 Sepsis18 (1.1%)26 (1.7%)-0.6%0.692 (0.381 – 1.258)0.225 ED Utilization241 (15.4%)175 (11.1%)4.2%1.377 (1.148 – 1.651) 0.001 Inpatient Hospitalization211 (13.7%)190 (12.4%)1.3%1.111 (0.925 – 1.333)0.261 One Year Pseudarthrosis202 (12.9%)155 (9.9%)3.0%1.303 (1.070 – 1.587) 0.008 Reoperation149 (9.5%)141 (9.0%)0.5%1.057 (0.849 – 1.316)0.622 Two Years Pseudarthrosis232 (14.8%)182 (11.6%)3.2%1.275 (1.064 – 1.528) 0.008 Reoperation208 (13.2%)183 (11.7%)1.6%1.137 (0.944 – 1.369)0.177Legend – risk is the number of patients with the outcome (proportion of cohort). Bold-face type indicates statistical significance. Abbreviations – ERT, estrogen replacement therapy; ED, emergency department; CI, confidence interval. Primary and Secondary Outcomes Legend – risk is the number of patients with the outcome (proportion of cohort). Bold-face type indicates statistical significance. Abbreviations – ERT, estrogen replacement therapy; ED, emergency department; CI, confidence interval. The incidence of pseudarthrosis was 12.9% amongst ERT users compared with 9.9% in the ERT-naïve group at the one-year postoperative interval (12.9% versus 9.9%, RR 1.303, 95% CI 1.070 – 1.587, P=0.008). Conversely reoperation rates were similar at the one-year interval (9.5% in ERT-use cohort versus 9.0% in ERT-naïve cohort, RR 1.057, 95% CI 0.849 – 1.316, P=0.622). A similar trend was observed within two years postoperatively (14.8% versus 11.6%, RR 1.275, 95% CI 1.064 – 1.528, P=0.008) whereas no difference in reoperation was observed (13.2% in ERT-use cohort versus 11.7%, RR 1.137, 95% CI 0.944 – 1.369, P=0.177). Discussion Given the conflicting evidence concerning hazards of ERT use in the TJA population, we sought to investigate the association between preoperative ERT and postoperative complications following single-level lumbar fusion. We hypothesized that ERT would be associated with increased risk of venous thromboembolism within the 90 day postoperative interval, which was partially upheld by our data. The incidence of PE was more than doubled amongst ERT users, although in absolute terms, the associated absolute risk elevation was 0.9%, (1.7% versus 0.8%), corresponding to approximately 111 patients on a number-needed-to-harm basis. Further, no such association was observed between ERT and postoperative DVT. ERT was also associated with mildly elevated risk of pseudarthrosis within one and two years postoperatively alongside similar incidence of reoperation. Lastly, we found that preoperative ERT was associated with increased rates of ED utilization, although the specific reason for presentation cannot be ascertained and may be unrelated to the index fusion procedure. The findings of the primary analysis further divide the existing literature regarding ERT use prior to orthopaedic procedures. No previous investigation of ERT use amongst primary TJA candidates demonstrated increased risk of PE within the short-term postoperative interval, and only one identified ERT as a risk factor for postoperative DVT following total shoulder arthroplasty (TSA).19-21 Furthermore, a recent propensity-matched database study found that ERT use was associated with significantly reduced incidence of both DVT and PE following TJA, although this may be confounded by greater perioperative VTE prophylaxis due to heightened hypercoagulability concerns. 27 The validity of our findings is further complicated by the inconsistency between PE and DVT risk elevation. Given the relative infrequency of coded PE in this population and the absence of adjustment for multiple comparison due to inherent constraints of the TriNetX database, it is possible that the observed PE risk elevation indicates noise as opposed to true biological signal. Nonetheless, oral contraceptives are a well-known independent risk factor for VTE following lumbar spine surgery, and the mechanism by which this occurs is well understood. 28 Exogenous estrogen is known to decrease levels of inhibitors of coagulation, such as protein S and antithrombin, while increasing fibrinogen, prothrombin, and Factors VII, VIII, and X. 29 Estrogen also induces activated Protein C resistance in a dose-dependent manner, which results in elevated Factor V activity and further downstream coagulative activity. 29 Likewise, estradiol has been shown to promote hypercoagulability by increasing the quantity and morphology of fibrin crosslinks while initiating the release of platelet-derived nitric oxide which is believed to increase platelet activity and adhesion.30,31 Lastly, exogenous estrogen stimulates the release of von Willebrand factor and interleukin-8 from vascular endothelial cells via MAP kinase pathway activation, therefore contributing to greater vascular inflammation and procoagulatory milieu. 32 Importantly, the clinical impact of this hypercoagulable state appears to be conferred only by oral estrogen formulations with no such association with transdermal administration. 16 Contrary to our original hypothesis, preoperative ERT use was associated with elevated pseudarthrosis risk within both one (RR 1.30) and two years (RR 1.28) postoperatively. While the effect size is relatively modest, it is largely consistent with generally accepted risk factors for incomplete fusion in this patient population, notably osteoporosis (1.40), smoking (1.38), age (1.06), and involvement of adjacent vertebral segments (1.35).6,33,34 This finding is paradoxical given the known efficacy of ERT in restoring BMD and may be explained by several factors. First, this result may reflect confounding by indication, as current guidelines include high baseline risk of fracture as one of the key criteria for the prescription of ERT in menopausal women with osteoporosis. 14 Therefore, despite successful matching for the diagnosis of age-related osteoporosis, it is possible that patients in the ERT cohort still possessed compromised bone mineralization, particularly as more objective measures such DEXA or CT Hounsfield units data were not available. This possibility is further compounded by the relatively narrow preoperative exposure interval. Subgroup analysis of the Framingham study found the threshold for a significant improvement in BMD following initiation of ERT to be seven years; patients issued ERT for a shorter duration demonstrated similar BMD to that of control subjects. 35 As a result, it is possible that the exposure interval chosen was insufficient for patients in the experimental cohort to yield the clinical benefit conferred by ERT. It is also possible that patients who received ERT experienced more prominent menopausal symptoms or reflect other unmeasured clinical characteristics that contribute to increased risk of postoperative pseudarthrosis. Nonetheless, the existing literature corroborates our findings, as preoperative ERT is associated with increased rates of implant failure and reoperation following TJA.19-21 Additionally, a meta-analysis by Sinha et al found menopausal HRT correlates with reduced osseointegration of dental implants. 36 The mechanism for these findings is unclear. There is emerging data that suggests the existence of differing bone remodeling phenotypes, with estrogen demonstrating deleterious effects on the sclerotic subtype specifically. 37 Therefore, at best, our findings represent a signal that requires radiographic confirmation and additional investigation are more clearly elucidate the mechanism by which ERT may contribute to these adverse events. In spite of these concerns, there is insufficient evidence regarding the necessity of HRT cessation prior to undergoing lumbar fusion. Whereas combined hormonal contraceptives are contraindicated in patients scheduled for major surgery, particularly those that require extended immobilization, no such recommendations yet exist for ERT. 38 Likewise, a recent review by Abouharb et al found no reference to cessation of HRT prior to TJA despite frequent acknowledgement of exogenous estrogen as a risk factor for postoperative VTE. 39 While this topic is ultimately outside the scope of the present investigation, we recommend that the decision of preoperative maintenance or cessation of ERT be made on an individualized basis, particularly in the context of competing risk factors for complication following lumbar fusion. 40 This retrospective analysis possesses several inherent limitations that require explicit acknowledgement. First, the study is purely observational and precludes causal claims. Second, although propensity-score matching was performed, there is a high likelihood of additional confounding from unmeasured variables that may influence postoperative complication rates such as compliance with surgeon and rehabilitative protocols, baseline functional and nutritional status, living environment, and the exact segmental location at which the fusion occurred. Additionally, while it is implied that patients were issued some variety of perioperative VTE prophylaxis, we are unable to identify the specific medications used nor when therapy was initiated and terminated. Definition of the index lumbar fusion procedure using a heterogenous mix of surgical approaches that limits more granular investigation of one specific approach. We also cannot account for individual variations in surgical technique, such as the incorporation of adjunctive materials, use of interbody cages, bone grafting, application of external bone stimulation, or surgeon experience. The selected ERT exposure interval, while specifically chosen to mitigate misclassification error, may alternatively introduce bias as patients started on ERT to promote bone healing in the postoperative setting were excluded. This may artificially inflate the observed association between ERT and pseudarthrosis, with control subjects at heightened risk of this complication being removed from analysis. Moreover, the TriNetX platform is wholly dependent upon medical billing codes, which render it subject to misclassification bias and limit data granularity. For this reason, we chose not to separate patients according to procedural indication, which likely contributes additional heterogeneity to the study population. It is possible that exclusive definition of pseudarthrosis using its ICD-10-CM code underestimates its true incidence, as the diagnosis would require identification according to radiographic or clinical parameters followed by addition of the appropriate medical billing code. The definition of pseudarthrosis in this manner also impairs applicability of our results, as the exact diagnostic criteria could not be extracted, and may have been used to inappropriately classify related outcomes including delayed or malunion. Inherent limitations of the TriNetX platform preclude more rigorous statistical analyses that adjust for multiple comparisons; therefore, our significant findings may reflect Type I error. Finally, by nature this de-identified database lacks information regarding the specific dose, duration, route of administration, and patient adherence with ERT, all of which may undermine our results. The absence of therapy duration is perhaps most notable, as the effects of exogenous estrogen on bone mineralization and hypercoagulability are both dose dependent. 35 Conclusion Preoperative estrogen replacement therapy is associated with increased incidence of pulmonary embolism, healthcare resource utilization, and incidence of pseudarthrosis following single-level lumbar fusion; of note, no such associations were observed with incidence of deep vein thrombosis or reoperation. At best, these findings represent hypothesize-generating evidence that post-menopausal patients who use estrogen replacement therapy may represent a higher-risk cohort in comparison to the broader lumbar fusion population. However, our results primarily necessitate future studies to clarify causality, particularly in light of the limitations of the present investigation, as well as potential mechanistic relationships between exogenous estrogen therapy and the bone healing and thromboembolic complications observed herein.