This observational study was based on the French national cancer cohort, which includes all patients diagnosed or treated for cancer since 2010. This cohort, described in detail elsewhere (17), is extracted from the large-scale French National Health Data System (SNDS) (21).

Briefly, SNDS is a national database collecting healthcare consumption and reimbursement claims covering 99% for the French population (i.e., 67 million people). It includes demographic data (e.g., sex, date of birth, date of death if applicable, health insurance scheme), hospitalization data (diagnoses, medical procedures, expensive drugs and medical devices), outpatient care data (drugs dispensed, lab tests, procedures and services). Sick leave allowances and disability allowances are also entered in SNDS.

The Diagnosis codes were recorded based on the International Classification of Diseases – 10^th revision, ICD-10 (World Health Organization) (22). Procedures were recorded with the CCAM classification (classification commune des Actes Médicaux) (23). Medicinal products were identified with the Anatomical Therapeutic Classification (ATC) code (24). Laboratory assays were identified with the national laboratory test coding table (NABM) (25).

This study was authorized by the French Data Protection Agency (Commission nationale de l’informatique et des libertés—Cnil (26)) n°2019-082 and 2019-083.

In this study, we included all adult patients who underwent, between 2011 and 2015, total thyroidectomy or completion thyroidectomy for cancer (ICD-10 codes: C73 or D093 or D440 or E070 or D448), with or without central and/or lateral lymph node dissection (CCAM codes: KCFA005 or KCFA007 or KCFA002 or KCFA003 or KCFA006 or KCMA001). We excluded patients who were taking calcium and/or active vitamin D up to the last 30 days before the procedure (except if there was only one drug delivery in the 90 days preceding the procedure), as well as patients with a previous history of parathyroid pathology or resection (see codes in appendix).

In this study we used the ATC codes A12AA (Calcium), A12AA01 (Calcium phosphate), A12AA02 (Calcium glubionate), A12AA03 (Calicm gluconate), A12AA04 (Calcium carbonate), A12AA20 (different Calcium salts in association), A12AX (combinations of Calcium with vitamin D and/or other drugs) to search for what we defined as « Calcium treatments ». Our definition of « Vitamin D treatment » was selective: we searched either for Calcitriol (ATC code A11CC03) or for Alfacalcidiol (ATC code A11CC04)

We divided patients into two groups according to whether they could be considered as having had postoperative hypoparathyroidism or not. Group 1 included patients who started calcium and/or vitamin D supplementation within the first postoperative month and/or were hospitalized for severe hypocalcemia at any time in the first year (ICD-10 code E83.51: Code for hypocalcemia). Group 2 included the other patients.

Subgroups were also defined to better characterize potential hypoparathyroidism. Subgroup 1A consisted of patients treated during the first postoperative month and who continued the treatment continuously during the first postoperative year (patients treated for probable permanent hypoparathyroidism), subgroup 1B consisted of patients treated during the first postoperative month and who discontinued their treatment during the first postoperative year (patients treated for probable temporary hypoparathyroidism), and subgroup 1Z consisted of patients who were hospitalized with hypocalcemia <1.5mmol/l (ICD-10 code E83.51) without any treatment during the first postoperative month (patients with hypoparathyroidism probably untreated or undetected during the first postoperative month). Subgroup 2C consisted of patients never treated during the first postoperative year (patients without hypoparathyroidism), subgroup 2D consisted of patients untreated during the first postoperative month, but who started treatment in the second or third postoperative month (patients with indeterminate postoperative parathyroid status), and subgroup 2E consisted of patients untreated during the first postoperative month, who started treatment after the third postoperative month (patients treated for a reason probably unrelated to hypoparathyroidism).

The study population was described using the following criteria: age, gender, chronic comorbidities at any time in the last 365 days prior to thyroidectomy (list of 17 included in the Charlson comorbidity index) (27–30), French deprivation index (31), year of surgery, length of thyroidectomy hospital stay, geographical area of thyroidectomy hospital stay, type of thyroid resection (total thyroidectomy, completion thyroidectomy, complex total thyroidectomy), type of lymph node dissection performed (none, central compartment, lateral compartment), radioactive iodine (RAI) treatment, other complications (laryngeal, hemorrhagic, infectious and cutaneous complications), and rehospitalization for calcium disorders (at any time during the first postoperative year for the last three criteria).

Costs were estimated from the payer’s perspective (French National Health Insurance).

All healthcare consumption reimbursed by the payer during the first postoperative year were identified in the database, i.e., from the end of surgery hospitalization up to day 365 after the date of surgery, or until the date of death, whichever occurred first.

Inpatient (hospitalization) care costs based on the Diagnosis-Related Group tariffs and drugs or medical devices billable in addition to the DRGs in public and private hospitals, outpatient care costs (drugs, biological tests, imaging procedures, total medical consultations, transportation and other procedures, such as physiotherapy or speech therapy, etc.), and indirect costs (all-cause sick leave daily allowance and invalidity allowance) were included in the overall reimbursed healthcare expenditures.

Some health expenditure items thought to be specifically related to hypoparathyroidism were also described, and included some specific medications, medical consultations, bioassays, imaging and rehospitalization for calcium disorders (other than severe hypocalcemia). Specific drug prescriptions included conventional treatments prescribed to limit the long-term harm of hypoparathyroidism: calcium, active vitamin D, magnesium, thiazide diuretics, and phosphate binders. The specific medical consultations concerned doctors likely to have been directly involved in the treatment (general practitioner, endocrinologist, general/visceral surgeon, ENT surgeon) or its complications (plastic surgeon, rheumatologist, nephrologist, urologist, psychiatrist, cardiologist). Specific bioassays concerned tests associated with the monitoring of hypoparathyroidism and its potential renal complications: calcium, phosphorus, parathormone, 25-(OH)-vitamin D (D2+D3), plasma magnesium, urea, creatinine, urine calcium level, urine creatinine level, creatine clearance (32). Specific imaging tests were those associated with monitoring potential complications of treated hypoparathyroidism: urinary tract ultrasound, brain CT and/or MRI, skull X-ray and bone densitometry (32). Specific rehospitalizations for calcium disorders also provide information on monitoring of potential parathyroid-related complications. ATC, NABM, CCAM and ICD-10 codes used to collect drugs, bioassays, imaging and hospitalization are shown in the Appendix.

Categorical variables were described with frequency and percentage, continuous variables with mean and standard deviation ( ± SD) or median.

Univariate cost analyses were performed with the non-parametric Wilcoxon test, multivariate analyses were carried out using generalized linear models (GLMs) with log link and gamma distribution. Incremental expenditures associated with hypocalcemia and covariates were estimated from respective regression coefficients (independent differentials) in GLM including all the following covariates: age (1^st quartile to 4^th quartile), gender, Charlson comorbidity index (0, 1 or 2, > 2), French deprivation index (from 1^st quintile: the least deprived quintile to the 5^th quintile: the most deprived quintile), year of surgery (2011 to 2015), geographical area of thyroidectomy hospital stay (13 administrative areas), type of thyroid resection (total thyroidectomy, completion thyroidectomy, complex total thyroidectomy), type of lymph node dissection (none, central compartment, lateral compartment), laryngeal complications, hemorrhagic complications, infectious complications, cutaneous complications (complications at any time during the first postoperative year).

Statistical tests were two-sided, a p-value < 0.05 was considered as significant.

All statistical analyses were performed using World Programming System 4.02.

Between January 1st, 2011, and December 31st, 2015, 34,398 patients underwent total thyroidectomy for thyroid cancer in France. After taking into account for exclusion factors (preoperative calcium or vitamin D-calcium therapy for 2,040 patients, associated parathyroid pathology for 1,080 patients, age <18 years for 288 patients, and uninterpretable data for 28 patients), 31,175 patients were analyzed, as detailed in the flow chart, Figure 1 .

In total, out of the 31,175 patients analyzed, 13,247 patients (42.49%) were deemed to have postoperative hypoparathyroidism on account of calcium or vitamin D+calcium replacement therapy started within the first postoperative month (n=13,224: 4,344 with calcium alone, and 8,880 with vitamin D+calcium) or hospitalization for severe hypocalcemia at any time in the first year (n=23). They were included in Group 1.

The remaining 17,928 patients (57.51%) were assigned to Group 2. The main characteristics of both groups were described ( Table 1 ).

Among the 13,224 patients who started calcium and/or vitamin D supplementation within the first postoperative month, 2,855 (22%) were still treated at one year (Subgroup 1A). Of the other 10,369 patients who discontinued their treatment before 1 year (Subgroup 1B), 9,530 (92%) were treated for less than 6 months.

Among the 17,928 patients without any supplementation within the first postoperative month (Group 2), 17,030 patients (55% of the patients of this study) took no treatment during the first year (Subgroup 2C), 336 patients started supplementation in the second or third postoperative month (subgroup 2D, indeterminate postoperative parathyroid status), with 106 (32%) of these still treated at 1 year, and 562 patients started supplementation after the third postoperative month (subgroup 2E, indeterminate postoperative parathyroid status but probably not related to thyroid surgery), with 269 (49%) of these still treated at 1 year. Ultimately, 375 of these 898 patients with undetermined parathyroid status were still treated at 1 year.

A representation of the duration of supplementation (≥1 calcium ± vitamin D delivery) by time intervals: within the first month, >M1 and ≤M6, >M6 and <M12, or at M12 accounts for late recoveries ( Figure 2 ). For instance, patients still treated 1 year after thyroidectomy, from 9.1% (2,855) to 10.4% (3,230) of patients, can be found in the peripheral circle of the figure.

Regarding health care resource utilization within the first postoperative year, almost all patients received outpatient care, 70% of patients were hospitalized at least once: 9,523 patients (72%) in Group 1 versus 12,336 patients (69%) in Group 2; 91% had at least one specific bioassay: 12,886 patients (97%) in Group 1 versus 15 440 (86%) in Group 2; 46% received at least one specific drug: 1,069 patients (6%) in Group 2 versus all patients in Group 1, except 15 of 23 included because of hospitalization for severe hypocalcemia; imaging was observed in 8% of patients in each group; at least one rehospitalization for calcium disorders was observed in 277 patients (2.1%) in Group 1 versus 56 patients (0.3%) in Group 2. In addition, 37% of patients received daily allowances: 5,263 (40%) in Group 1 versus 6,259 (35%) in Group 2.

The overall health expenditure of the 31,175 patients during the first postoperative year was €95,809,106 in Group 1 and €124,315,208 in Group 2 ( Table 2 ). The mean overall health expenditure per patient during the first year was €7,233 (±€9,256) in Group 1 and €6,934 (±€8,751) in Group 2, with an average difference of €298 per patient (p<0.0001). The average difference between the two groups in inpatient care costs was €23 (p<0.0001), the difference in outpatient expenses was €111 (p<0.0001), and the difference in indirect expenses was €164 (p<0.0001).

Specifically, hypoparathyroidism-related health care expenditures during the first year were €6,339,607, with an average of €478.6 (€ ± 708.8) per patient in Group 1 versus a total of €5,964,120 and an average of €332.7 (€ ± 461.0) per patient in Group 2 ( Table 3 ). This resulted in a mean difference between the groups of €145.9 per patient (p<0.0001), mainly due to drugs and rehospitalizations for calcium disorders.

To take into account potential confounders, a multivariate analysis was performed. After adjustment, the 1-year incremental cost in patients treated for postoperative hypoparathyroidism was estimated at €142 (p<0.004) ( Table 4 ).

The mean differences between patients in Subgroup 1A (permanent hypoparathyroidism) and Subgroup 1B (temporary hypoparathyroidism) were €2,126.9 per patient (p<0.0001) for overall expenditures ( Table 5 ), and €434.6 (p<0.01) per patient for specific expenditures ( Table 6 ).

Similarly, the mean difference between patients in Subgroup 1A (permanent hypoparathyroidism) and patients in Subgroup 2C (no hypoparathyroidism) was €2,129.3 per patient (p<0.0001) for overall expenditures ( Table 5 ), and €493.3 (p<0.0001) per patient for specific expenditures ( Table 6 ).

After adjustment for potential confounders, the incremental cost for patients treated for a probable permanent postoperative hypoparathyroidism (Subgroup 1A) compared to patients without hypoparathyroidism (Subgroup 2C) was estimated at €776 (p<0.0001) ( Table 4 ).

Based on comprehensive real-world data, our study makes it possible to describe observed health expenditure without the need for extrapolation. Furthermore, by providing a comparison with a control group of untreated patients, i.e., without hypoparathyroidism, our study makes it possible to estimate postoperative hypoparathyroidism-related health expenditure. Finally, the nature of the data available makes it possible to categorize health expenditure, and to estimate not only specific costs directly associated with hypoparathyroidism, but also indirect health costs, thus making it possible to detect unforeseen expenditure. These three items offer new concepts compared to the literature. Wang et al. (10) and Nicholson et al. in Markov cohort model studies (12), and Mercante et al. in a randomized trial (13), compared the cost of several drug supplementation regimens for hypoparathyroidism, while Fanget et al., in a retrospective series, estimated an approximate additional management cost of hypocalcemia patients, including the initial surgical stay, and without specifying the perspective or the source of cost data (11). These studies varied in the items chosen to estimate the cost, whether it was the price of drugs (10–12), biological assays (10, 13), caregiver time (12, 13), excess days of initial hospitalization (11), or rehospitalizations (10), but none of these studies included a control group of non-hypoparathyroidism patients. Other studies did not include cost quantification, such as Chen et al., who reported, in a retrospective multicenter study, health resource utilization: pills and medical visits (14) and Siggelkow et al. who conducted a survey assessing, among other things, medication use, caregiver burden and impact on work in patients with hypoparathyroidism (15). These latter two studies did not specifically target postoperative hypoparathyroidism, and none of the studies mentioned above have specifically focused on cancer patients. Finally, Mathonnet et al., in a large study describing the complete care pathway of total thyroidectomy patients in France, also quantified the rate of hypoparathyroidism after total thyroidectomy for cancer, but did not provide any cost estimation (4).

The inability to access patient records, which were anonymized, made it difficult to confirm the diagnosis of postoperative hypoparathyroidism. Rather than searching the SNDS database for patients with the ICD10 coding ‘hypoparathyroidism’, we preferred to use a more reliable definition of hypoparathyroidism based on automatically recorded drug reimbursement. With ICD10 codes, cases without hospitalization would not have been identified, only stays with hypoparathyroidism would have been located, but these codes do not provide sufficient clinical information, and they do not indicate the duration of the condition. Within the first postoperative month, these codes were found for only 21 patients: 19 in Group 1 and 2 in Group 2 (data not shown), therefore no bias was introduced. Even though laboratory tests are entered into the database (blood calcium, etc.), their results are not available, which implies the use of a proxy. We assumed that the initiation of vitamin D and/or calcium treatment immediately after total thyroidectomy, in previously untreated patients, left little doubt about the causal link between the treatment and the disease. However, we were interested to find that a group of patients (Subgroup 2D with n=336 patients), started their treatment after the first postoperative month. A further analysis of this small subgroup showed that one third of these patients (representing only 106 patients) started treatment after the first month, but continued it for the whole first year. These patients were most probably patients treated for permanent hypoparathyroidism, whose initial treatment was not been recorded, for example. Some patients started treatment after the third postoperative month, a timeframe that is inconsistent with primary hypoparathyroidism (Subgroup 2E with 562 patients, of whom 269 were still being treated at 1 year). Therefore, these were patients who were most likely treated for a reason other than postoperative hypoparathyroidism (renal osteodystrophy for instance), although it is theoretically possible that hypocalcemia requiring treatment could remain untreated for 3 months. However, given the low proportion of the potentially misclassified patient population, we consider that the use of the proxy we chose did not affect the findings of our study.

Finally, our study only considered the first postoperative year, because it corresponded to the first phase of the disease. This phase makes it possible to discriminate between temporary hypoparathyroidism and permanent hypoparathyroidism. After the first year, the occurrence of serious complications due to permanent hypoparathyroidism should allow the identification of the few cases potentially misclassified. These long-term complications can be expected to cause potentially significant additional costs. Therefore, a 5-year update of the current data has been planned.

In conclusion, our study found that 42.5% of patients who underwent total thyroidectomy for cancer in France were treated for postoperative hypoparathyroidism, and 9.1% remained treated one year post-surgery. During the first postoperative year, health expenditures were significantly higher for patients treated for postoperative hypoparathyroidism, even if the difference was small. A 5-year follow-up is planned to assess the more burdensome long-term complications and their costs.