Association Between Hypothyroidism and Postoperative Complications Following Operative Ankle Fracture Fixation: A Retrospective Cohort Study

Abstract

Background:

Hypothyroidism has been associated with altered bone metabolism, impaired mineralization, and increased fracture risk, but its relationship with postoperative complications after operative ankle fracture fixation remains poorly defined. This study evaluated whether preexisting hypothyroidism is associated with increased complications following surgical treatment of ankle fractures.

Methods:

Adults undergoing operative ankle fracture fixation or related ankle fracture surgery in TriNetX were stratified by hypothyroidism diagnosed within 6 months before or on surgery vs no recorded hypothyroidism history. Propensity score matching was performed 1:1 using demographic factors, diagnosis-coded obesity/overweight, hemoglobin A_1c, tobacco-related variables, diabetes, chronic kidney disease, cardiovascular disease, chronic obstructive pulmonary disease, peripheral vascular disease, and osteoporosis. No single primary outcome was designated; all 6 outcomes were assessed on an equal, exploratory basis at 3 months, 6 months, and 1 year.

Results:

After matching, 10 406 patients per cohort were included in the 3- and 6-month analyses, and 10 407 per cohort were included at 1 year. With the numbers available, no significant difference could be detected for any 3-month outcome. At 6 months, nonunion/malunion (RR, 1.305; 95% CI, 1.065-1.600; RD P = .010) and hardware failure (RR, 1.302; 95% CI, 1.058-1.602; RD P = .013) were higher in hypothyroid patients. At 1 year, nonunion/malunion (RR, 1.407; 95% CI, 1.166-1.698; RD P < .001) and hardware failure (RR, 1.264; 95% CI, 1.065-1.500; RD P = .007) remained higher. Infection, hardware removal, wound dehiscence, and broad procedure-related complications were not significant at any time point.

Conclusion:

In this propensity-matched retrospective cohort study, preexisting hypothyroidism was associated with modest but statistically significant increases in delayed osseous and hardware-related complications after operative ankle fracture surgery (P < .05 for both outcomes at 6 months and 1 year). These findings should be interpreted as associative and warrant prospective validation.

Level of Evidence:

Level III, retrospective cohort study.

Introduction

Ankle fractures are some of the most common injuries treated in orthopaedics, with the mechanism of injury varying from population subtype and activity level. The most common mechanisms are falls (54.83%), followed by sports (20.76%), exercise (16.84%), jumping (4.42%), and trauma (2.84%). ¹ There is an incidence rate of 4.22/10 000 person-years, and young adults aged 21-30 years have the highest injury rates; however, in general, older individuals are more at risk for complications.^1,2 Given this epidemiologic profile, identifying modifiable and non-modifiable risk factors that predispose patients to ankle fracture remains a priority.

Thyroid hormones are crucial regulators of normal adult bone health. Elevated levels of thyroid hormones triiodothyronine (T3) and thyroxine (T4) accelerate bone resorption from baseline, resulting in a net reduction of bone density and propagation of abnormal architecture. ³ This process is typically regulated in normal adults via a classic endocrine negative feedback loop, which sustains the concentrations of thyroid hormones and thyroid-stimulating hormone as an inversely proportional relationship. Through this process, the hypothalamic-pituitary-thyroid (HPT) axis set point is defined.^4,5 Disruption of this axis alters the regulation of bone metabolism and may predispose affected individuals to skeletal fragility and increased fracture risk. ⁶

It is important to distinguish that, within the confines of what we consider hypothyroidism, there are 2 types. In primary hypothyroidism, the thyroid gland cannot produce adequate amounts of thyroid hormone. In secondary hypothyroidism, most commonly caused by autoimmune thyroiditis and iodine deficiency, the thyroid function is normal; however, the reduction in thyroid hormone results from the abnormal pituitary gland or hypothalamus function. ⁷ At the skeletal level, both subtypes result in reduced circulating T3 and T4, which slows bone turnover and impairs mineralization.^7,8 Although this distinction between primary and secondary hypothyroidism carries significant clinical implications for diagnosis and management, it is less consequential for the present study, as our analysis captures hypothyroidism as a diagnostic entity regardless of etiology, consistent with large-scale database methodology.

Large retrospective cohort studies have consistently demonstrated elevated fracture risk in hypothyroid patients across general skeletal sites. Reviews of this literature have concluded that hypothyroidism and subclinical hyperthyroidism are associated with decreased bone mineral density and increased fracture risk. ⁹ Trauma registry–based work has begun to explore this space at a clinical level: one prospective single-center registry study found that hypothyroid patients presenting with fractures were more likely to be female with low-energy trauma mechanisms, involving proximal femoral and humeral fractures. However, thyroid status was not significantly associated with post-management outcomes. ⁶ More recently, TriNetX-based database methodology has been applied to examine how hypothyroidism affects surgical outcomes in the context of knee arthroplasty, lumbar fusion, and total shoulder arthroplasty, among others.^{10 -12} Despite this growing TriNetX literature in orthopaedics, to our knowledge, no study to date has used this platform to examine the relationship between hypothyroidism and postoperative complications following operative ankle fracture fixation.

Methods

Data Source and Study Design

We conducted a retrospective cohort study using TriNetX, a federated electronic health record (EHR) research network that supports analysis of de-identified patient data across participating health care organizations. The TriNetX Analytics platform contains structured information on demographics, diagnoses, procedures, medications, and laboratory values, and provides built-in tools for cohort construction, propensity score matching, and outcome analyses. All analyses were performed within TriNetX using all available historical data accessible to the network at the time of the queries, without applying an explicit calendar-year restriction. Because TriNetX is a federated system with platform-level analytic constraints, certain modules may exclude index events that fall outside allowable historical windows.

All cohorts and analyses were generated in TriNetX on March 22, 2026, using the TriNetX Global Collaborative federated research network, and results should be interpreted as reflecting the state of the network at that time (“time-stamped” analysis). Because TriNetX is a live federated platform in which contributing health care organizations refresh and update their underlying EHR data on an ongoing basis (with the majority of data sources updating on 1-, 2-, or 4-week cycles and an average refresh lag of approximately 1 month), patient counts, covariate distributions, and downstream effect estimates may change with subsequent network refreshes, retrospective documentation/coding updates, or terminology remapping. Accordingly, re-running identical cohort definitions at a later date may yield modest differences in cohort size and outcome estimates compared with the March 22, 2026, time-stamped results. Because TriNetX is a federated platform with predefined analytic outputs, certain patient-level data elements, including exact remaining risk-set counts at intermediate time points and competing-risk model outputs, were not exportable from the platform. No further patient-level risk-set granularity beyond the outcome-specific denominators and end-of-window survival summaries in Supplemental Table S5 could be exported from TriNetX.

Table 1.

Study Population and Matched Cohort Sizes by Follow-up Window. ^a

Follow-up Window	Pre-match: Hypothyroidism	Pre-match: No Hypothyroidism	Post-match: Hypothyroidism	Post-match: No Hypothyroidism
3 mo	10 476	131 345	10 406	10 406
6 mo	10 476	131 345	10 406	10 406
12 mo	10 476	131 345	10 407	10 407

^aTable 1 reports matched analysis cohort sizes by follow-up analysis. These values are not outcome-specific risk sets or confirmation that all patients were observed through each follow-up endpoint; outcome-specific analytic denominators are reported in Supplemental Table S5.

This cohort design is visually shown in Figure 1.

Figure 1.

Study design and patient selection flowchart. Patients ≥18 years undergoing ankle surgery were identified from the TriNetX Global Collaborative Network and divided into hypothyroidism (ICD-10-CM E03; n = 10 474) and non-hypothyroidism (n = 131 314) cohorts. After propensity score matching on age, race, sex, wound-healing risk factors, and diabetes-related variables, 6 postoperative complications were assessed from 1 day to 3 months, 6 months, and 1 year. The n values shown along the timeline represent the matched analytic cohort sizes for each follow-up analysis and should not be interpreted as outcome-specific numbers remaining at risk at the end of each interval. ICD-10-CM, International Classification of Diseases, Tenth Revision, Clinical Modification.

Study Population and Exposure Definition

We identified adults aged 18 years or older who underwent operative management of ankle fracture-related pathology using a predefined basket of ankle surgery procedure codes. The index event for each patient was the most recent instance of an included ankle surgery procedure. Clinically, the basket included open fixation of bimalleolar, trimalleolar, lateral malleolar, medial malleolar, posterior malleolar, pilon/tibial plafond, syndesmotic, and ankle dislocation-related injuries, as well as International Classification of Diseases, Tenth Revision, Procedure Coding System (ICD-10-PCS) ankle joint reposition with internal fixation. This intentionally broad definition was selected to capture operatively treated ankle fracture patterns represented within TriNetX; however, it also introduces heterogeneity in injury severity, soft tissue involvement, and surgical complexity. The full code list used to define the ankle surgery index event is provided in Supplemental Table S1.

Two cohorts were constructed. The hypothyroidism cohort included patients with a recorded diagnosis of hypothyroidism (International Classification of Diseases, Tenth Revision, Clinical Modification [ICD-10-CM] E03) within the 6 months preceding or concurrent with the most recent ankle surgery index event. The comparator cohort included patients meeting the same ankle surgery index definition but without any recorded diagnosis of hypothyroidism in their available TriNetX history.

Covariates and Propensity Score Matching

To address confounding variables, baseline characteristics were selected based on their clinical relevance to fracture healing, postoperative risk, and infection or complication susceptibility. Matching variables included age, sex, race or ethnicity, overweight/obesity/hyperalimentation as an ICD-10-CM diagnosis-code flag (E65-E68), and hemoglobin A_1c. Continuous body mass index (BMI) was intentionally not used for matching because BMI values in TriNetX were inconsistently available and variably timed relative to surgery; therefore, obesity-related body habitus was handled as a diagnosis-code covariate rather than a continuous measured BMI variable. Hemoglobin A_1c was included as a TriNetX curated numeric variable using the most recent value available within the platform. Comorbidities included diabetes mellitus, chronic kidney disease, chronic ischemic heart disease, heart failure, chronic obstructive pulmonary disease, atherosclerosis of native arteries of the extremities, peripheral vascular disease, and osteoporosis without current pathologic fracture. Tobacco exposure was captured through nicotine dependence, tobacco use, and personal history of nicotine dependence. These variables were used for 1:1 propensity score matching within TriNetX. Matching was performed using the platform’s nearest-neighbor greedy algorithm without replacement with the default 0.10-SD caliper; exact matching was not applied. ¹³ A 1:1 ratio was selected to prioritize covariate balance and conservative comparability rather than maximizing the size of the substantially larger comparator cohort. Cohort balance after matching was evaluated using standardized mean differences from the platform’s balance diagnostics, with SMD values <0.10 interpreted as acceptable balance. The full covariate list is provided in Supplemental Table S3.

Post-matching balance is summarized in Supplemental Table S4, which reports baseline characteristics for the matched hypothyroidism and control cohorts along with standardized mean differences (SMDs). Continuous variables are presented as mean ± SD and categorical variables as n (%), as generated by TriNetX. Propensity score overlap or density plots and additional post-matching diagnostics beyond the exported balance table were not available for inclusion from the TriNetX report.

Outcomes and Statistical Analysis

Outcomes were assessed from 1 day after the index ankle surgery through 3 months, 6 months, and 1 year to reflect clinically meaningful stages of postoperative recovery and fracture healing. Outcomes included infection-related complications, hardware removal, nonunion or malunion captured using ICD-10-CM M84 as a pragmatic union-complication proxy, hardware failure or mechanical complication, wound dehiscence, and a broad complications-of-procedures category (Supplemental Table S2). No single primary outcome was designated; all 6 outcomes were prespecified and assessed on an equal, exploratory basis. For each follow-up window and each outcome, patients with documented evidence of the outcome before the beginning of that outcome window were excluded from the corresponding outcome analysis to better isolate incident postoperative events. Therefore, the matched cohort sizes in Table 1 and Figure 1 differ from the outcome-specific analytic denominators reported by TriNetX. The outcome-specific analytic denominators, event counts, and Kaplan-Meier survival probabilities at the end of each window are provided in Supplemental Table S5.

All analyses were performed within the TriNetX Analytics environment. Statistical significance was evaluated using TriNetX outputs including risk difference (RD), risk ratio (RR), log-rank test (LRT), hazard ratio (HR), and test statistic (TS) P values. The LRT was treated as the primary longitudinal comparison because it captures time-to-event differences across the study interval. No additional post-matching multivariable regression, formal proportional hazards testing, or competing-risk analysis was performed because patient-level data and these diagnostics were not exportable from the platform. TriNetX Kaplan-Meier analyses censor patients after the last fact in their record; however, the platform export did not provide patient-level numbers remaining at risk at 3, 6, and 12 months, and no further patient-level risk-set granularity beyond the outcome-specific denominators and end-of-window survival summaries in Supplemental Table S5 could be exported. TriNetX follow-up summaries demonstrated median follow-up of 90, 180, and 365 days in the matched cohorts for the 3-month, 6-month, and 1-year analyses, respectively; however, these summary values do not provide outcome-specific patient-level numbers at risk. Accordingly, these analyses should be interpreted as platform-based matched cohort comparisons rather than complete follow-up of every matched patient through each time point. Statistical significance was defined using a 2-sided alpha threshold of .05.

Results

Study Population and Matched Cohorts

Before propensity score matching, 10 476 patients with hypothyroidism and 131 345 patients without hypothyroidism met study criteria. After 1:1 matching, the 3-month and 6-month analyses each included 10 406 patients per cohort, whereas the 1-year analysis included 10 407 patients per cohort (Table 1). These values represent matched analysis cohort sizes, not the number of patients confirmed to remain under observation at the end of each interval. Because TriNetX excluded patients with a prior history of each outcome separately, the analytic denominator varied by outcome; outcome-specific denominators and event counts are reported in Supplemental Table S5. Across all follow-up windows, with the numbers available, no significant between-group differences could be detected for any prespecified outcome during the earliest postoperative period. Divergence emerged later, with hypothyroidism associated with higher rates of nonunion or malunion and hardware failure by 6 months and persisting through 1 year (Tables 3 and 4).

Table 3.

Six-Month Postoperative Outcomes After Propensity Score Matching. ^a

	P Value
Outcome	RD	LRT	HR	TS
Infection	.448	.642	.637	.698
Hardware removal	.239	.406	.226	.797
Nonunion/malunion	.010	.015	.846	.378
Hardware failure	.013	.020	.325	.313
Wound dehiscence	.665	.548	.537	.995
Complications of procedures	.997	.850	.899	.714

Abbreviations: HR, hazard ratio; LRT, log-rank test; RD, risk difference; TS, test statistic.

^aBoldface indicates significance (P < .05).

Table 4.

One-Year Postoperative Outcomes After Propensity Score Matching. ^a

	P Value
Outcome	RD	LRT	HR	TS
Infection	.466	.795	.721	.859
Hardware removal	.310	.745	.950	.845
Nonunion/malunion	<.001	<.001	.502	.504
Hardware failure	.007	.017	.511	.573
Wound dehiscence	.486	.631	.960	.607
Complications of procedures	.609	.826	.242	.980

Abbreviations: HR, hazard ratio; LRT, log-rank test; RD, risk difference; TS, test statistic.

^aBoldface indicates significance (P < .05).

Three-Month Outcomes

At 3 months, with the numbers available, no significant differences could be detected between cohorts for infection, hardware removal, nonunion or malunion, hardware failure, wound dehiscence, or broad complications of procedures (Table 2). The closest signals to significance were observed for nonunion or malunion (RD P = .062; LRT P = .072) and hardware failure (RD P = .159; LRT P = .177), but neither met the prespecified threshold for significance. These findings suggest that any association between hypothyroidism and adverse postoperative recovery may not be apparent during the earliest phase of follow-up, as illustrated in the forest plot of 3-month postoperative outcomes in Figure 2.

Table 2.

Three-Month Postoperative Outcomes After Propensity Score Matching.

	P Value
Outcome	RD	LRT	HR	TS
Infection	.637	.752	.202	.336
Hardware removal	.370	.446	.750	.974
Nonunion/malunion	.062	.072	.932	.382
Hardware failure	.159	.177	.613	.896
Wound dehiscence	.566	.495	.641	.805
Complications of procedures	.919	.829	.402	.315

Abbreviations: HR, hazard ratio; LRT, log-rank test; RD, risk difference; TS, test statistic.

Figure 2.

Forest plot of 3-month postoperative outcomes in hypothyroid vs euthyroid patients following operative ankle fracture fixation. No outcomes reached statistical significance.

Six-Month Outcomes

By 6 months, 2 outcome domains diverged between groups. Nonunion/malunion was more common in the hypothyroidism cohort, with significance identified by both RD P = .010 and LRT P = .015; the corresponding RD 95% CI excluded zero (.001 to .009), and the log-rank χ² was 5.891 (Table 3). Hardware failure also reached significance at 6 months, with RD P = .013 and LRT P = .020; the corresponding RD 95% CI excluded zero (.001 to .008), and the log-rank χ² was 5.392 (Table 3). The corresponding outcome-specific event counts were 209 of 9989 vs 161 of 10 044 for nonunion/malunion and 201 of 10 284 vs 155 of 10 323 for hardware failure in the hypothyroidism and comparison cohorts, respectively (Supplemental Table S5). No significant differences were observed for infection, hardware removal, wound dehiscence, or broad procedural complications. Notably, these significant findings were driven by cumulative risk differences, reflecting the total proportion of patients who experienced each complication by 6 months. Time-to-event analysis further supported these differences, accounting for when during follow-up events occurred rather than simply whether they occurred. Hazard ratio P values, which capture the instantaneous rate of events at any given moment, remained non-significant, suggesting the between-group differences emerged gradually over the follow-up window rather than at a discrete point in time. These patterns are illustrated in the forest plot of 6-month postoperative outcomes in Figure 3.

Figure 3.

Forest plot of 6-month postoperative outcomes in hypothyroid vs euthyroid patients following operative ankle fracture fixation. Nonunion/malunion and hardware failure reached statistical significance at this time (RR 1.305, P = .010; RR 1.302, P = .013, respectively).

One-Year Outcomes

At 1 year, the association between hypothyroidism and impaired osseous or mechanical recovery persisted. Nonunion/malunion demonstrated the strongest signal in the study, with RD P < .001 and LRT P < .001; the RD 95% CI remained above zero (.003 to .011), and the log-rank χ² increased to 11.392 (Table 4). Hardware failure also remained significantly more common in the hypothyroidism cohort at 1 year, with RD P = .007 and LRT P = .017; the RD 95% CI was .002 to .010, and the log-rank χ² value was 5.723 (Table 4). The corresponding outcome-specific event counts were 256 of 9991 vs 183 of 10 050 for nonunion/malunion and 289 of 10 286 vs 229 of 10 302 for hardware failure in the hypothyroidism and comparison cohorts, respectively (Supplemental Table S5). In contrast, infection, hardware removal, wound dehiscence, and broad complications of procedures continued to show no statistically significant differences between groups. Taken together, the 1-year results suggest a complication profile centered more on delayed healing and fixation integrity than on generalized wound or infectious morbidity, as illustrated in the forest plot of 1-year postoperative outcomes in Figure 4.

Figure 4.

Forest plot of 1-year postoperative outcomes in hypothyroid vs euthyroid patients following operative ankle fracture fixation. Nonunion/malunion and hardware failure remained statistically significant (RR 1.407, P < .001; RR 1.264, P = .007, respectively).

Discussion

This retrospective propensity-matched TriNetX study identified a reproducible association between preexisting hypothyroidism and later postoperative complications after operative ankle fracture surgery. Specifically, hypothyroidism was associated with higher rates of nonunion or malunion and hardware failure by 6 months, and both associations persisted through 1 year. In contrast, no significant relationship was observed for infection, wound dehiscence, hardware removal, or the broader complications-of-procedures code. As summarized in Tables 3 and 4, the pattern that emerged was therefore one of delayed osseous and mechanical compromise rather than generalized perioperative morbidity.

Although the relative differences were statistically significant, the absolute event differences were modest. At 1 year, nonunion/malunion occurred in 256 of 9991 patients in the hypothyroidism cohort compared with 183 of 10 050 in the comparison cohort, and hardware failure occurred in 289 of 10 286 compared with 229 of 10 302, respectively (Supplemental Table S5). Thus, these findings are best interpreted as a reproducible population-level signal that may inform risk stratification and follow-up intensity rather than as evidence that hypothyroidism alone should determine postoperative management.

Potential interaction between hypothyroidism and other metabolic or vascular risk factors should also be considered. Although diabetes, chronic kidney disease, obesity/overweight, tobacco exposure, peripheral vascular disease, and osteoporosis were included in propensity score matching and demonstrated acceptable post-matching balance, these conditions may still modify the relationship between hypothyroidism and fracture healing. For example, diabetes and chronic kidney disease may compound thyroid-related effects through impaired microvascular function, altered bone turnover, and disrupted mineral metabolism. The present TriNetX analysis was not designed to test interaction terms or stratified effect modification; future studies should evaluate whether the risk associated with hypothyroidism is amplified in patients with these comorbidities.

The time course of the findings is clinically meaningful. At 3 months, no outcome met the threshold for statistical significance, whereas by 6 months, both union-related complications and hardware failure had become significant. This delayed separation is biologically plausible. Hypothyroidism is associated with reduced thyroid hormone activity, slowed bone turnover, and impaired mineralization.^3,7,8 In the context of operatively treated ankle fractures, such alterations may not necessarily produce early wound problems but may instead interfere with the gradual process of osseous consolidation and the restoration of stable load sharing across fixation constructs. The stronger 1-year signal for nonunion or malunion, particularly the highly significant LRT result at that interval, is consistent with this interpretation.

These findings also align directionally with the broader orthopaedic literature on hypothyroidism. Prior database analyses in total knee arthroplasty, lumbar fusion, and total shoulder arthroplasty have reported increased postoperative complications, including late mechanical or implant-related events, among patients with hypothyroidism.^{10 -12} Our ankle-specific analysis extends that pattern into a population that is especially dependent on timely fracture healing and maintenance of fixation.

At the same time, our results differ somewhat from prior general fracture registry work that did not detect a clear relationship between hypothyroidism and post-management outcomes. Several explanations are possible. The prior study by Ahmad et al ⁶ evaluated a more heterogeneous fracture population and was not restricted to a single surgically treated anatomic region. By contrast, the current analysis focused on operative ankle fracture surgery, where delayed union and hardware-related failure may be easier to detect as clinically coded postoperative endpoints. It is also possible that hypothyroidism exerts a relatively modest effect that becomes more visible when the cohort, treatment context, and outcome definitions are more narrowly focused.

One notable feature of this study is the absence of a significant signal for infection or wound dehiscence at any point. This distinction is important because it suggests that hypothyroidism in this setting may be more closely related to bone-healing biology and fixation durability than to soft tissue breakdown or infectious susceptibility. Hardware removal was likewise not significantly different between cohorts, which may reflect the fact that hardware removal is influenced by many factors beyond construct failure alone, including implant prominence, patient symptoms, surgeon preference, and differential follow-up. The lack of significant HR P values despite significant RD and LRT findings further suggests that the observed differences may represent cumulative divergence over time rather than a sharp early hazard spike.

From a practical standpoint, these findings support consideration of hypothyroidism during preoperative risk assessment and postoperative surveillance for ankle fracture surgery. However, the absolute risk differences were modest, and this study cannot determine whether optimization of thyroid status or changes in postoperative weight-bearing protocols would reduce complications. Therefore, the findings should be interpreted as supporting closer longitudinal attention to radiographic healing and fixation integrity rather than as evidence to mandate a specific management change. Future studies should evaluate whether laboratory-confirmed thyroid control, thyroid hormone replacement status, or disease severity modifies this risk profile.

Limitations

This study has several limitations inherent to retrospective database research. First, as a de-identified TriNetX analysis, it can establish association but not causation. Residual confounding is possible despite propensity score matching, particularly because important fracture- and surgery-specific variables are not reliably captured in structured EHR data. These include fracture severity, open vs closed injury status, soft tissue injury, syndesmotic instability severity, surgical technique, fixation construct, surgeon and hospital variability, and postoperative rehabilitation or weight-bearing protocol.

Second, hypothyroidism was defined by diagnosis coding rather than by laboratory-confirmed thyroid status. Accordingly, the database could not reliably distinguish primary from secondary hypothyroidism, disease severity, chronicity, medication adherence, levothyroxine treatment status, perioperative thyroid hormone levels, or whether a patient with a diagnosis of hypothyroidism was euthyroid at the time of surgery. Vitamin D status, nutritional factors, calcium and phosphate metabolism, and other metabolic bone conditions were also not consistently available for matching or stratified analysis.

Third, the ankle surgery cohort was intentionally broad and included a basket of open ankle fracture-related procedures rather than a single narrowly defined operation. Although this approach improved cohort size and generalizability, it likely introduced procedural heterogeneity with respect to injury severity, surgical complexity, and baseline risk of postoperative complications. This heterogeneity is particularly relevant for pilon fractures, syndesmotic procedures, and ankle dislocation-related procedures, which may carry complication risks that differ from isolated malleolar fracture fixation.

Fourth, outcome ascertainment relied on diagnosis and procedure coding. In particular, nonunion or malunion was captured using ICD-10-CM M84 as a pragmatic proxy for disorder of continuity of bone, which may incompletely reflect the full clinical spectrum of union complications. Coding heterogeneity across institutions may also have affected capture of hardware-related and infectious endpoints.

Fifth, continuous BMI was not used for propensity score matching because of inconsistent availability and variable timing in TriNetX; overweight/obesity/hyperalimentation was instead captured as an ICD-10-CM diagnosis-code flag (E65-E68). This diagnosis-based approach may undercapture patients with elevated measured BMI but no corresponding obesity-related diagnosis code. Hemoglobin A_1c was included using the most recent TriNetX curated numeric values available within the platform. Because this variable may change over time and may not have been measured uniformly near the surgical date, residual confounding from time-varying metabolic risk remains possible.

Finally, no minimum follow-up requirement was imposed. TriNetX Kaplan-Meier analyses censor patients after the last fact in their record, but the available export did not provide exact patient-level numbers remaining at risk at each requested interval, and no further patient-level risk-set granularity could be exported beyond Supplemental Table S5. The platform also did not permit formal competing-risk analysis. Death, severe medical illness, transfer of care outside the network, or differential follow-up intensity could prevent complications from being observed or recorded, particularly in older patients and those with chronic kidney disease, cardiovascular disease, or diabetes. Although matching included age and several major comorbidities, competing events and incomplete outcome capture may still have influenced the reported complication rates and may have affected the hypothyroidism and comparison cohorts differently.

Conclusion

In this propensity-matched TriNetX cohort study, preexisting hypothyroidism was associated with modest but statistically significant increases in nonunion or malunion and hardware failure after operative ankle fracture surgery at 6 months and 1 year. No significant associations were identified for infection, wound dehiscence, hardware removal, or broad procedure-related complications. Overall, the observed signal appears to be focused on delayed osseous healing and mechanical construct integrity rather than generalized postoperative morbidity. Because this study was retrospective and subject to residual confounding, incomplete follow-up, and platform-level analytic constraints, these findings should be interpreted as associative rather than causal. Further prospective research is needed to determine whether thyroid optimization or other perioperative interventions can reduce delayed healing or hardware-related complications in this population.

Supplemental Table 1.

(Cohort Definitions and Code Sets).

Concept Set	Code System	TriNetX Concept ID	Term / Description
Ankle surgery procedures
Ankle surgery	CPT	UMLS:CPT:27814	Open treatment of bimalleolar ankle fracture (eg, lateral and medial malleoli, or lateral and posterior malleoli, or medial and posterior malleoli), includes internal fixation, when performed
Ankle surgery	CPT	UMLS:CPT:27822	Open treatment of trimalleolar ankle fracture, includes internal fixation, when performed, medial and/or lateral malleolus; without fixation of posterior lip
Ankle surgery	CPT	UMLS:CPT:27823	Open treatment of trimalleolar ankle fracture, includes internal fixation, when performed, medial and/or lateral malleolus; with fixation of posterior lip
Ankle surgery	CPT	UMLS:CPT:27792	Open treatment of distal fibular fracture (lateral malleolus), includes internal fixation, when performed
Ankle surgery	CPT	UMLS:CPT:27766	Open treatment of medial malleolus fracture, includes internal fixation, when performed
Ankle surgery	CPT	UMLS:CPT:27769	Open treatment of posterior malleolus fracture, includes internal fixation, when performed
Ankle surgery	CPT	UMLS:CPT:27828	Open treatment of fracture of weight bearing articular surface/portion of distal tibia (eg, pilon or tibial plafond), with internal fixation, when performed; of both tibia and fibula
Ankle surgery	CPT	UMLS:CPT:27826	Open treatment of fracture of weight bearing articular surface/portion of distal tibia (eg, pilon or tibial plafond), with internal fixation, when performed; of fibula only
Ankle surgery	CPT	UMLS:CPT:27827	Open treatment of fracture of weight bearing articular surface/portion of distal tibia (eg, pilon or tibial plafond), with internal fixation, when performed; of tibia only
Ankle surgery	CPT	UMLS:CPT:27829	Open treatment of distal tibiofibular joint (syndesmosis) disruption, includes internal fixation, when performed
Ankle surgery	CPT	UMLS:CPT:1005278	Open treatment of fracture of weight bearing articular surface/portion of distal tibia (eg, pilon or tibial plafond), with internal fixation, when performed
Ankle surgery	CPT	UMLS:CPT:1005290	Open treatment of ankle dislocation, with or without percutaneous skeletal fixation
Ankle surgery	CPT	UMLS:CPT:1005272	Open treatment of trimalleolar ankle fracture, includes internal fixation, when performed, medial and/or lateral malleolus
Ankle surgery	ICD-10-PCS	UMLS:ICD10PCS:0SSG04Z	Reposition left ankle joint with internal fixation device, open approach
Ankle surgery	ICD-10-PCS	UMLS:ICD10PCS:0SSF04Z	Reposition right ankle joint with internal fixation device, open approach
Hypothyroidism exposure (diagnosis)
Hypothyroidism	ICD-10-CM	UMLS:ICD10CM:E03	Other hypothyroidism

Abbreviations: CPT, Current Procedural Terminology; ICD-10-CM, International Classification of Diseases, Tenth Revision, Clinical Modification; ICD-10-PCS, International Classification of Diseases, Tenth Revision, Procedure Coding System

Supplemental Table 2.

Postoperative Outcomes and Code Sets.

Outcome	Type	Code	Term / Description
Infection
Infection	Diagnosis	UMLS:ICD10CM:T81.4	Infection following a procedure
Infection	Diagnosis	UMLS:ICD10CM:T84.6	Infection and inflammatory reaction due to internal fixation device
Infection	Diagnosis	UMLS:ICD10CM:M86	Osteomyelitis
Infection	Diagnosis	UMLS:ICD10CM:L03	Cellulitis and acute lymphangitis
Infection	Diagnosis	UMLS:ICD10CM:L02	Cutaneous abscess, furuncle and carbuncle
Infection	Diagnosis	UMLS:ICD10CM:A41	Other sepsis
Hardware Removal
Hardware removal	Procedure	UMLS:CPT:20670	Removal of implant; superficial (eg, buried wire, pin or rod) (separate procedure)
Hardware removal	Procedure	UMLS:CPT:20680	Removal of implant; deep (eg, buried wire, pin, screw, metal band, nail, rod or plate)
Hardware removal	Procedure	UMLS:CPT:27704	Removal of ankle implant
Hardware removal	Procedure	UMLS:ICD10PCS:0SPF04Z	Removal of internal fixation device from right ankle joint, open approach
Hardware removal	Procedure	UMLS:ICD10PCS:0SPG04Z	Removal of internal fixation device from left ankle joint, open approach
Hardware removal	Diagnosis	UMLS:ICD10CM:Z47.2	Encounter for removal of internal fixation device
Non-union/Malunion
Non-union/Malunion	Diagnosis	UMLS:ICD10CM:M84	Disorder of continuity of bone
Hardware failure
Hardware failure	Diagnosis	UMLS:ICD10CM:T84.1	Mechanical complication of internal fixation device of bones of limb
Hardware failure	Diagnosis	UMLS:ICD10CM:T84.11	Breakdown (mechanical) of internal fixation device of bones of limb
Hardware failure	Diagnosis	UMLS:ICD10CM:T84.12	Displacement of internal fixation device of bones of limb
Hardware failure	Diagnosis	UMLS:ICD10CM:T84.19	Other mechanical complication of internal fixation device of bones of limb
Hardware failure	Diagnosis	UMLS:ICD10CM:T84.2	Mechanical complication of internal fixation device of other bones
Hardware failure	Diagnosis	UMLS:ICD10CM:T84.3	Mechanical complication of other bone devices, implants and grafts
Wound Dehiscence
Wound dehiscence	Diagnosis	UMLS:ICD10CM:T81.3	Disruption of wound, not elsewhere classified
Complications of Procedures
Complications of procedures	Diagnosis	UMLS:ICD10CM:T81	Complications of procedures, not elsewhere classified

Supplemental Table 3.

Covariates Used for Propensity Score Matching. ^a

Covariate	Code system	Code(s)	Notes
Age at index	TriNetX demographic field		Continuous
Sex	TriNetX demographic field		Female/Male/Unknown
Race	TriNetX demographic field		Race categories per TriNetX
Ethnicity	TriNetX demographic field		Hispanic/Latino status per TriNetX
Hemoglobin A1c	TriNetX curated lab concept (numeric)	TNX Curated: 9037	Most recent value (platform default)
Diabetes mellitus	ICD-10-CM	E08–E13	Any diabetes diagnosis
Overweight/obesity/hyperalimentation	ICD-10-CM	E65–E68	Diagnosis-code flag only (E65-E68)
Chronic kidney disease	ICD-10-CM	N18.*	Any CKD stage
Chronic ischemic heart disease	ICD-10-CM	I25.*	Chronic ischemic heart disease
Heart failure	ICD-10-CM	I50.*	Heart failure
Chronic obstructive pulmonary disease	ICD-10-CM	J44.*	Chronic obstructive pulmonary disease
Nicotine dependence	ICD-10-CM	F17.*	Current nicotine dependence
Tobacco use	ICD-10-CM	Z72.0	Tobacco use (not otherwise specified)
Personal history of nicotine dependence	ICD-10-CM	Z87.891	History of nicotine dependence
Atherosclerosis of native arteries of extremities	ICD-10-CM	I70.2	Atherosclerosis of native arteries of the extremities
Peripheral vascular disease	ICD-10-CM	I73.9	Peripheral vascular disease, unspecified
Osteoporosis (no current pathological fracture)	ICD-10-CM	M81	Osteoporosis without current pathological fracture

^aCodes are listed as TriNetX concept identifiers as exported from the platform (eg, UMLS:CPT:*; UMLS:ICD10CM:*; UMLS:ICD10PCS:*; TNX Curated:*).

Supplemental Table 4.

Post-matching Balance.

Characteristic	Hypothyroidism (n = 10 407)	No Hypothyroidism (n = 10 407)	SMD
Age at index, y, mean ± SD	62.5 ± 15.1	62.8 ± 15.1	0.017
Female, n (%)	8709 (83.7)	8806 (84.6)	0.026
Male, n (%)	1695 (16.3)	1594 (15.3)	0.027
White, n (%)	9131 (87.7)	9188 (88.3)	0.017
Black or African American, n (%)	625 (6.0)	586 (5.6)	0.016
Hispanic or Latino, n (%)	414 (4.0)	386 (3.7)	0.014
Asian, n (%)	145 (1.4)	139 (1.3)	0.005
Overweight/obesity and hyperalimentation (E65-E68), n (%)	4020 (38.6)	3921 (37.7)	0.020
Hemoglobin A1c (%, mean ± SD)	6.3 ± 1.5 (n = 4979; 47.8%)	6.3 ± 1.5 (n = 4858; 46.7%)	0.001
Diabetes mellitus, n (%)	3229 (31.0)	3021 (29.0)	0.044
Chronic kidney disease, n (%)	1822 (17.5)	1563 (15.0)	0.067
Chronic ischemic heart disease, n (%)	2086 (20.0)	1887 (18.1)	0.049
Heart failure, n (%)	1436 (13.8)	1257 (12.1)	0.051
Chronic obstructive pulmonary disease, n (%)	1417 (13.6)	1258 (12.1)	0.046
Nicotine dependence, n (%)	1758 (16.9)	1705 (16.4)	0.014
Tobacco use, n (%)	575 (5.5)	545 (5.2)	0.013
History of nicotine dependence, n (%)	2496 (24.0)	2396 (23.0)	0.023
Atherosclerosis of extremities, n (%)	230 (2.2)	226 (2.2)	0.003
Peripheral vascular disease, n (%)	602 (5.8)	536 (5.2)	0.028
Osteoporosis, n (%)	1573 (15.1)	1502 (14.4)	0.019

Supplemental Table 5.

Outcome-Specific Analytic Denominators, Event Counts, and Kaplan-Meier Survival Probabilities.

Follow-up	Outcome	Hypothyroidism, n / Events	No Hypothyroidism, n / Events	KM Survival at Window End, % / %
3 mo	Infection	8306 / 441	8663 / 446	94.39 / 94.52
3 mo	Hardware removal	10 117 / 380	10 142 / 357	96.00 / 96.22
3 mo	Nonunion/malunion	9989 / 143	10 044 / 114	98.50 / 98.80
3 mo	Hardware failure	10 284 / 105	10 323 / 86	98.92 / 99.11
3 mo	Wound dehiscence	10 187 / 300	10 252 / 316	96.89 / 96.72
3 mo	Complications of procedures	9875 / 528	10 001 / 538	94.37 / 94.29
6 mo	Infection	8306 / 604	8663 / 604	92.15 / 92.37
6 mo	Hardware removal	10 117 / 690	10 142 / 650	92.53 / 92.83
6 mo	Nonunion/malunion	9989 / 209	10 044 / 161	97.75 / 98.25
6 mo	Hardware failure	10 284 / 201	10 323 / 155	97.85 / 98.33
6 mo	Wound dehiscence	10 187 / 367	10 252 / 381	96.15 / 95.98
6 mo	Complications of procedures	9875 / 637	10 001 / 645	93.11 / 93.04
12 mo	Infection	8304 / 779	8623 / 781	89.47 / 89.61
12 mo	Hardware removal	10 118 / 1059	10 136 / 1017	87.94 / 88.09
12 mo	Nonunion/malunion	9991 / 256	10 050 / 183	97.14 / 97.92
12 mo	Hardware failure	10 286 / 289	10 302 / 229	96.78 / 97.37
12 mo	Wound dehiscence	10 188 / 410	10 248 / 393	95.62 / 95.76
12 mo	Complications of procedures	9876 / 724	9982 / 713	92.00 / 92.13