Keywords
Thyroidectomy; Hypocalcemia; Parathyroid; Vitamin D; Calcium
Post-thyroidectomy hypocalcemia remains a significant surgical complication with varying reported incidence rates. While traditional focus has centered on surgical factors, the role of metabolic predictors in calcium homeostasis disturbances remains incompletely understood.
We conducted a single-center retrospective analysis of 1,760 thyroidectomy patients based on data obtained from King Abdullah International Medical Research Center (KAIMRC), Saudi Arabia, evaluating metabolic predictors and long-term calcium homeostasis patterns. Preoperative assessment included thyroid function tests, glycemic control, vitamin D status, and anthropometric measurements. We have followed the TRIPOD + AI statement which provide updated guidance for reporting clinical prediction models that use regression or machine learning methods, We developed a risk stratification model using multivariate logistic regression and implemented mixed-effects modeling for longitudinal analysis. Recovery patterns were evaluated through standardized follow-up at immediate postoperative, one-month, three-month, six-month, and one-year timepoints.
Post-thyroidectomy hypocalcemia occurred in 585 patients (33.2%, 95% CI: 31.0-35.4%). Preoperative thyroid stimulating hormone (TSH) levels appeared to be the strongest predictor (OR 2.24, 95% CI 1.71-2.93, p<0.0001), followed by T3 (OR 1.34, 95% CI 1.19-1.51) and glycated hemoglobin (Hemoglobin A1C), (OR 1.18, 95% CI 1.07-1.30). Our risk stratification system, including TSH, vitamin D, and immediate postoperative calcium, achieved superior discrimination (AUC=0.921) and effectively categorized patients into low (6.5%), intermediate (14.6%), and high-risk (83.2%) groups. Recovery patterns revealed a non-linear trajectory, with an unexpected decline at six months (79.71%) before maximal improvement at one year (88.61%).
Our findings highlight and demonstrate the significant role of metabolic factors in post-thyroidectomy hypocalcemia risk and challenge the current standard recovery assumptions, suggesting the need for extended monitoring protocols and comprehensive metabolic assessment in perioperative management.
Thyroidectomy; Hypocalcemia; Parathyroid; Vitamin D; Calcium
Post-thyroidectomy hypocalcemia remains one of the most challenging complications in endocrine surgery, significantly impacting patient outcomes, healthcare resource utilization, and quality of life. Despite advances in surgical techniques and perioperative management, reported incidence rates continue to vary ranging from 19% to 38% for transient hypocalcemia and 0.8% to 3.6% for permanent cases.1,2
This variability underscores the variant and heterogenous pathophysiological considerations and multifactorial nature of post-thyroidectomy calcium homeostasis disturbances and the nature of systemic regulation. The clinical significance of this complication extends beyond immediate postoperative care. While traditional focus has centered on surgical technique and parathyroid glands preservation, the recent evidence suggests a more systematic, multifactorial and integrated interplay between pre-existing metabolic factors and postoperative outcomes that differ between individuals.3 The previous studies have highlighted the possible role and the influence of preoperative metabolic status on calcium homeostasis, however there are no complete nor comprehensive evaluation and analysis of these relationships in the literature. Furthermore, current risk stratification models, primarily based on surgical factors and parathyroid hormone levels, may inadequately capture the full spectrum of predictive variables in disease pathophysiology and regulations.4
The optimization of post-thyroidectomy care faces several critical challenges. First, the timing and frequency of calcium monitoring vary significantly across institutions, lacking standardization based on patient-specific risk factors.4 Second, while several computational and predictive models exist, their clinical utility is often limited by complexity or insufficient integration of metabolic factors in their proposed structure. Third, the long-term trajectory of calcium homeostasis recovery remains inadequately characterized, especially when it comes to the concerns about impact of preoperative metabolic status on recovery conditions.5 Previous studies have focused on immediate postoperative outcomes and surgical techniques. However, these studies often overlooked the impact and contribution of pre-existing metabolic conditions and metabolic factors of individuals on the surgical outcomes.3 The relationship between thyroid gland functions, glycemic control, and post-thyroidectomy calcium homeostasis represents an understudied area. While some studies have suggested associations between preoperative thyroid function and hypocalcemia risk, focused analysis of these relationships, especially in the context of other metabolic factors, remains a missing point that needs to be studied.6
Understanding these associations could change our approach to preoperative optimization and postoperative monitoring. Given that, our study aims to address multiple knowledge gaps from the current point of view from literature. At first, we look to develop a risk stratification model incorporating both traditional surgical factors, metabolic factors and conditions. Secondly, we aim to characterize the long-term trajectory of calcium homeostasis recovery, with special focus to the influence of preoperative metabolic status. In addition to that, we intended to quantify the relative contributions of various metabolic predictors to post-thyroidectomy hypocalcemia risk.5
We conducted a single-center retrospective analysis based on the data obtained from King Abdullah International Medical Research Center (KAIMRC) after obtaining the necessary approvals and documentations, our study aims focus investigating calcium homeostasis patterns and metabolic risk factors in patients who underwent thyroid surgery. Our study utilized electronic health records collected data from consecutive patients who underwent thyroidectomy based on KAIMRC database. We have followed the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis Or Diagnosis (TRIPOD) + AI statement which provide updated guidance for reporting clinical prediction models that use regression or machine learning methods.7
Preoperative assessment included metabolic profiling, including thyroid function tests (thyroid stimulating hormone (TSH), T3, and T4), glycated hemoglobin levels glycemic control (HBA1C), vitamin D status, and anthropometric measurements including body mass index (BMI). Serum calcium levels were measured using standard automated analyzers, with post-thyroidectomy hypocalcemia defined as serum calcium below 8.5 mmol/L in the immediate postoperative period. We have utilized a standardized follow-up protocol with scheduled calcium measurements at immediate postoperative, one-month, three-month, six-month, and one-year timepoints to record the recovery effects over one-year duration.
Our statistical analysis strategy has integrated multiple statistical and computational methods aiming to achieve best possible precision and evaluation of outcomes and risk factors. We used multivariate logistic regression modeling to identify predictors of post-thyroidectomy hypocalcemia, with all continuous variables standardized prior to analysis. Multicollinearity was assessed using variance inflation factors, and model performance was evaluated through Akaike Information Criterion and pseudo-R2 statistics. To account for the longitudinal nature of calcium measurements, we implemented mixed-effects modeling with both fixed and random effects, incorporating time as a scaled continuous variable.
To minimize selection bias and confounding factors, we performed propensity score matching using a caliper width of 0.2 standard deviations. The matching algorithm included key preoperative variables as, BMI, vitamin D levels, thyroid function laboratory values, and glycemic status. We developed a clinical risk stratification system based on identified predictors, with cut-off values determined through receiver operating characteristic analysis and weighted according to their strength of association in multivariate analysis.
Recovery pattern analysis utilized time-to-event methodology, with recovery defined as achievement of calcium levels ≥2.1 mmol/L. We calculated recovery rates at predefined follow-up intervals and constructed recovery curves with 95% confidence intervals. Missing data were handled through median imputation for predictive modeling,
We developed our predictive model using an integrative multiphasic approach, beginning with univariate analysis of possible risk factors followed by multivariate modeling of significant predictors. The model was assessed through internal validation using cross-validation techniques, with performance assessed through area under the receiver operating characteristic curve (AUC) analysis. We evaluated model calibration using Hosmer-Lemeshow testing and assessed discriminative ability through sensitivity and specificity analyses at optimal cut-off points.
All statistical analyses were performed using Python-based statistical packages, including scikit-learn for machine learning techniques, statsmodels for statistical modeling, and scipy for statistical testing. We set statistical significance at p<0.05 and calculated 95% confidence intervals where appropriate. Our reporting adheres to The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational studies and the TRIPOD statement for prediction model development and validation.
Our cohort included a total of 1,760 thyroidectomy patients, with a prominent female predominance (79.5%, n=1,399). Preoperative metabolic profiling has shown a significant heterogeneity in baseline demographics. Mean BMI was 32.20 ± 18.19 kg/m2 (n=1,466), with 65.9% of patients classified as obese. Baseline thyroid function parameters demonstrated considerable variability: TSH (3.92 ± 13.58 mIU/L, n=538), T3 (4.58 ± 1.61 pmol/L, n=346), and T4 (11.91 ± 2.44 pmol/L, n=443). Vitamin D status assessment resulted with mean levels of 61.51 ± 28.78 nmol/L (n=245), while glycemic control evaluation showed mean HBA1C of 6.48 ± 1.54% (n=308) ( Table 1).
We performed an evaluation of data completeness check that have revealed differential capture rates across our included key variables from the dataset. Anthropometric measurements showed the highest completeness (83.3% for BMI), while metabolic parameters showed varying availability: calcium (30.3%), TSH (28.2%), T4 (25.1%), vitamin D (13.8%), and HBA1C (17.5%). This variability within data completeness required proper statistical approaches for handling missing data ( Figure 1).
Post-thyroidectomy hypocalcemia occurred in 585 patients, with an incidence rate of 33.2% (95% CI: 31.0-35.4%). Multivariate risk factor analysis identified significant predictors. Preoperative TSH was the most significant predictor (OR 2.24, 95% CI 1.71-2.93, p-value < 0.0001), followed by T3 (OR 1.34, 95% CI 1.19-1.51, p-value < 0.0001), and HBA1C (OR 1.18, 95% CI 1.07-1.30, p-value < 0.0001). The results of our predictive model have resulted in a good performance with significant discrimination values, through a cross-validation AUC of 0.683 ± 0.004 and test set performance of 0.457 ( Figure 2 and Figure 3).
Initial recovery rates in the immediate postoperative period reached 81.49% (n=471/578), followed by peak recovery at one month (86.92%, n=372/428). A subsequent plateau phase was observed at three months (84.58%, n=362/428), followed by a temporary decline at six months (79.71%, n=330/414), before maximal improvement observed at one year (88.61%, n=451/509) ( Table 2). Mixed-effects modeling quantified these variances, highlighting significant time-dependent effects (coefficient -0.030, SE 0.009, p=0.001) and BMI influence (coefficient -0.013, SE 0.006, p=0.034).
Time point | Group | Number of individuals | Mean ± SD | Normal Ca2 (%)† | P-value‡ |
---|---|---|---|---|---|
Immediate Postoperative | All Patients | 585 | 2.20 ± 0.15 | 80.7 | 0.960 |
Hypocalcemia | 585 | 2.20 ± 0.15 | 80.7 | ||
One-Month Postoperative | All Patients | 436 | 2.27 ± 0.18 | 86.2 | |
Hypocalcemia | 319 | 2.27 ± 0.18 | 83.7 | ||
No Hypocalcemia | 117 | 2.27 ± 0.18 | 93.2 | ||
Three-Months Postoperative | All Patients | 436 | 2.21 ± 0.15 | 83.0 | 0.028 * |
Hypocalcemia | 308 | 2.22 ± 0.14 | 84.1 | ||
No Hypocalcemia | 128 | 2.19 ± 0.16 | 80.5 | ||
Six-Months Postoperative | All Patients | 421 | 2.19 ± 0.14 | 78.4 | 0.864 |
Hypocalcemia | 252 | 2.19 ± 0.14 | 78.6 | ||
No Hypocalcemia | 169 | 2.19 ± 0.15 | 78.1 | ||
One-Year Postoperative | All Patients | 524 | 2.20 ± 0.14 | 86.1 | 0.064 |
Hypocalcemia | 256 | 2.21 ± 0.14 | 85.9 | ||
No Hypocalcemia | 268 | 2.19 ± 0.14 | 86.2 |
Propensity score matching paired a total of 322 cases, with aim to achieve balance across preoperative characteristics (standardized mean differences <0.1 for all variables). The matched cohort resulted with mean recovery rates of 84.3% (SD= 3.3%). Density distribution analysis of propensity scores demonstrated optimal overlap between treated and control groups ( Figure 4).
We have developed a risk stratification scoring system based on the results from our multivariate analysis ( Table 3), utilized and integrated weighted contributions from significant predictors with optimized cut-off values. TSH formed the highest weighted contribution (+2.1 points, cut-off >2.50), followed by vitamin D (+1.7 points, cut-off >56.33), and immediate postoperative calcium (+10.5 points, cut-off >2.13). This scoring system effectively stratified patients into three risk categories: low (n=895), intermediate (n=281), and high (n=584), with corresponding event rates of 6.5%, 14.6%, and 83.2% respectively. The results of our proposed predictive model have achieved significantly good discrimination (AUC= 0.921) with high performing calibration metrics as listed in Table 4, The risk assessment consists of two scales - preoperative and postoperative. The preoperative scale categorizes risk into three levels: low (0 to 2.1 counted points), intermediate (2.2 to 4.1 counted points), and high (4.2 to 6.5 counted points). Similarly, the postoperative scale also uses three risk categories: low risk (0 to 5 points), intermediate risk (5.1 to 10.4 points), and high risk (10.5 points and above).
Variable | Univariate OR (95% CI) | P-value | Multivariate OR (95% CI) | P-value |
---|---|---|---|---|
TSH Preoperative | 1.82 (0.40-9.66) | <0.0001 * | 1.39 (1.24-1.55) | <0.0001 * |
T4 Preoperative | 0.62 (0.47-4.55) | <0.0001 * | 1.04 (0.95-1.14) | 0.375 |
Vitamin D Preoperative | 1.53 (0.49-8.16) | <0.0001 * | 1.00 (0.99-1.01) | 0.359 |
BMI Preoperative | 0.93 (0.59-1.13) | 0.433 | 1.00 (0.99-1.01) | 0.956 |
HBA1C Preoperative | 1.40 (0.50-4.64) | <0.0001 * | 1.29 (1.12-1.49) | <0.0001 * |
Our longitudinal risk and outcome model has illustrated multiple recovery phases, with significant coefficients for both fixed effects (time: -0.030, SE 0.009) and random effects (variance component: 0.003, SE 0.010). The combination of metabolic factors and variables have shown significant modulation of recovery trajectories by BMI (p=0.034) but not vitamin D levels (p=0.564).
Our analysis of post-thyroidectomy hypocalcemia reveals both confirmatory findings and novel considerations that challenge our current understanding in several key areas. The demonstrated incidence rate of 33.2% is going with previous literature statistics however, there are several points in our findings that are crucial to be discussed in broader aspects and a wider manner. Qin et al. 20218 findings have reported rates ranging from 19% to 38%, despite that their meta-analysis did not account for the interplay and association of metabolic factors we’ve identified in our results. When stratifying by preoperative TSH levels, our findings suggest significantly higher risk in specific patient subgroups, especially those with elevated TSH (OR 2.24, 95% CI 1.71-2.93, P-value < 0.0001).
While Eismontas et al. 20189 narratively discussed and mentioned surgical technique and parathyroid preservation as primary risk factors, our multivariate analysis suggests a more complicated consideration. The strong predictive value of TSH and T3 levels in our cohort indicates that the traditional focus on surgical factors only may be insufficient. This is especially relevant when considering that Seo et al. 201510 reported parathyroid hormone levels as their strongest predictor (OR 1.8, 95% CI 1.4-2.3), while our TSH-based model demonstrated superior predictive accuracy (OR 2.24, 95% CI 1.71-2.93).
The recovery trajectory patterns we observed challenge several paradigms. The non-linear recovery pattern, with initial recovery rates of 81.49%, peaking at 86.92% (one month), followed by a decline to 79.71% (six months) before reaching 88.61% (one year); contradicts the traditional assumption of steady improvement reported in Castro et al. 201811 results and findings. This finding has significant clinical implications to be highlighted, suggesting that current follow-up protocols, typically focused on the first three months post-surgery, may be inadequately structured. The temporary decline at six months, not previously reported in major studies, warrants focused attention and may indicate a need for extended monitoring protocols.
Our risk stratification system represents a significant advancement but also highlights current limitations in the field. While El-Kharadly et al. 202312 proposed risk categorization based primarily on surgical factors, our integration of metabolic parameters (TSH, vitamin D, and immediate postoperative calcium) achieved superior discrimination (AUC = 0.921). However, it’s crucial to acknowledge that our model’s complexity, requiring multiple preoperative measurements that may present practical implementation challenges in resource-limited settings.
The influence of BMI on recovery trajectories (coefficient -0.013, SE 0.006, p=0.034) has important considerations. Previous studies, including Spartalis et al. 2019,13 noted associations between obesity and surgical complications but failed to quantify the specific impact on calcium homeostasis. Our finding that 65.9% of patients were classified as obese raises important questions about the generalizability of existing management protocols, which are largely based on studies with lower obesity rates.
The variable completeness and data reporting limitations in our dataset presents both limitations and opportunities for further development and future direction for next studies. While our BMI data was relatively complete (83.3%), the lower capture rates for calcium (30.3%) and TSH (28.2%) highlight a systemic issue in preoperative assessment. This contrasts with Karunakaran et al. 202014 in which they achieved higher completion rates but with smaller cohorts. Our propensity score matching approach, while being highly strict and statistically regulated, it cannot fully mitigate this limitation. Of particular concern is our finding regarding HBA1C’s impact on outcomes (OR 1.18, 95% CI 1.07-1.30, p-value < 0.0001). This relationship, largely unexplored in previous studies, suggests a possible link and a metabolic pathway affecting calcium homeostasis that highlights the need of further investigation and studies on that point. Malki & Abouqal 201415 touched on metabolic factors but did not specifically address glycemic control’s role in post-thyroidectomy calcium regulation.
While our strengths and novel points are forming novel findings and important considerations in the literature, we have to disclose and acknowledge major limitations within our study. The limitations of our study extend beyond data completeness. The single-center design, while allowing for standardized surgical approaches, limits generalizability. Our cohort’s demographic composition, particularly regarding BMI distribution, may not reflect global populations, especially that our study is focused from Saudi Arabia only and with different racial and genetic disposition around the world, the results may differ according to multiple factors including healthcare settings and practice routines. Also, the retrospective nature of our study design and methodology, while common in many surgical studies, introduces a selection bias that could not be completely eliminated.
Looking forward, our findings suggest several critical and important areas for future studies. The various and interdisciplinary recovery patterns we observed indicate a need for prospective studies with standardized long-term follow-up protocols. The strong influence of metabolic factors suggests that preoperative optimization strategies, especially regarding thyroid function and glycemic control, deserve more focused investigation. In addition to that, the development of simplified risk assessment tools that maintain predictive accuracy while requiring fewer laboratory measurements would enhance clinical applicability.
Our analysis of 1,760 thyroidectomy patients has resulted with several significant points and conclusions into post-thyroidectomy hypocalcemia prediction and management. The development of our risk stratification model, which include both metabolic and clinical factors in consideration for risk prediction, achieved superior discriminative ability (AUC = 0.921) compared to traditional approaches. Interestingly, preoperative TSH raised and appeared as the strongest predictor (OR 2.24, 95% CI 1.71-2.93, p-value < 0.0001), challenging the standard focus on surgical factors alone. The identification of the demonstrated and associated recovery patterns, especially the unexpected decline at six months (79.71%) before final improvement at one year (88.61%), suggests the need for extended monitoring protocols. Our findings regarding BMI’s influence on recovery trajectories (coefficient -0.013, SE 0.006, p=0.034) and the significant role of glycemic control (HBA1C: OR 1.18, 95% CI 1.07-1.30, p-value < 0.0001) highlight the importance of metabolic optimization in perioperative management. The weighted risk scoring system we developed, which included TSH (+2.1 points), vitamin D (+1.7 points), and immediate postoperative calcium (+10.5 points), effectively stratified patients into risk categories with outcome profiles. This model’s strong performance suggests a possible role to be further studied into clinical utilities in different settings, though external validation is needed. These findings call for a shift and practice routine changes in post-thyroidectomy care, focusing on comprehensive metabolic assessment and individualized monitoring protocols. Future prospective studies should focus on validating these predictive models across diverse populations and investigating the mechanistic relationships between metabolic factors and calcium homeostasis disturbances.
Institutional Review Board (IRB) Approval: The responsible IRB at King Abdullah International Medical Research Center (KAIMRC) has approved our research protocol for the study under protocol number: NRC23R/748/10. The proposed primary/preliminary protocol was approved at 29/10/2023.
This study was approved by the Institutional Review Board (IRB) at King Abdullah International Medical Research Center (KAIMRC) under protocol number NRC23R/748/10. The requirement for individual patient consent was waived by the IRB due to the retrospective nature of the study, as it involved analysis of existing medical records data only. All patient data were collected and handled in accordance with the ethical standards of the institutional research committee and the 1964 Helsinki declaration and its later amendments. Patient confidentiality was maintained through complete de-identification of all collected data.
Figshare: DATA REPOSITORY FOR: Metabolic Predictors and Risk Stratification of Post-Thyroidectomy Hypocalcemia: A Retrospective Analysis and Predictive Machinery Model. Doi: https://doi.org/10.6084/m9.figshare.28200563.v1.16
The project contains the following underlying data:
• THYROID SURGERY CALCIUM HOMEOSTASIS ANALYSIS REPORT.txt
• THYROID SURGERY OUTCOMES ANALYSIS REPORT.txt
• sheet_deidentified.csv
Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).
Figshare: EXTENDED DATA FOR: Metabolic Predictors and Risk Stratification of Post-Thyroidectomy Hypocalcemia: A Retrospective Analysis and Predictive Machinery Model. Doi: https://doi.org/10.6084/m9.figshare.28200617.v1.17
This project contains the following extended data:
• Table 2.docx
• Table 3.docx
• Table 4.docx
• Figure 1.png
• Figure 2.png
• Figure 3.png
• Figure 4.png
• Table 1. docx
Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).
We acknowledge King Abdullah International Medical Research Center (KAIMRC) for providing open access funding coverage.
Views | Downloads | |
---|---|---|
F1000Research | - | - |
PubMed Central
Data from PMC are received and updated monthly.
|
- | - |
Is the work clearly and accurately presented and does it cite the current literature?
Partly
Is the study design appropriate and is the work technically sound?
No
Are sufficient details of methods and analysis provided to allow replication by others?
No
If applicable, is the statistical analysis and its interpretation appropriate?
No
Are all the source data underlying the results available to ensure full reproducibility?
Yes
Are the conclusions drawn adequately supported by the results?
No
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: I am an endocrine surgeon
Alongside their report, reviewers assign a status to the article:
Invited Reviewers | |
---|---|
1 | |
Version 1 04 Mar 25 |
read |
Provide sufficient details of any financial or non-financial competing interests to enable users to assess whether your comments might lead a reasonable person to question your impartiality. Consider the following examples, but note that this is not an exhaustive list:
Sign up for content alerts and receive a weekly or monthly email with all newly published articles
Already registered? Sign in
The email address should be the one you originally registered with F1000.
You registered with F1000 via Google, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Google account password, please click here.
You registered with F1000 via Facebook, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Facebook account password, please click here.
If your email address is registered with us, we will email you instructions to reset your password.
If you think you should have received this email but it has not arrived, please check your spam filters and/or contact for further assistance.
Comments on this article Comments (0)