Positive Effect of Parathyroidectomy Compared to Observation on BMD in a Randomized Controlled Trial of Mild Primary Hyperparathyroidism

Mild or asymptomatic disease is now the dominating presentation of primary hyperparathyroidism (PHPT). However, bone involvement with decreased bone mineral density (BMD) and an increased risk of fractures has been demonstrated. Indications for parathyroidectomy (PTX) in mild PHPT have been debated for years. There is a need of long‐term randomized studies comparing PTX with observation without intervention (OBS). Here, we present bone health data from the Scandinavian Investigation of Primary Hyperparathyroidism (SIPH), a randomized controlled trial, comparing PTX to OBS. This study included 191 patients (96 OBS/95 PTX), and 129 patients (64 OBS/65 PTX) were followed for 10 years to the end of study (EOS). BMD was measured with dual‐energy X‐ray absorptiometry (DXA), peripheral fractures were noted, and spine radiographs were obtained for vertebral fracture assessment. There was a significant treatment effect of PTX on BMD compared with OBS for all analyzed compartments, most explicit for the lumbar spine (LS) and femoral neck (FN) (p < 0.001). The mean changes in T‐score from baseline to 10 years were from 0.41 for radius 33% (Rad33) to 0.58 for LS greater in the PTX group than in the OBS group. There was a significant decrease in BMD for all compartments in the OBS group, most pronounced for FN, Rad33, and ultradistal radius (UDR) (p < 0.001). Even though there was a significant treatment effect of PTX compared with OBS, there was only a significant increase in BMD over time for LS (p < 0.001). We found no difference between groups in fracture frequency in the 10‐year cohort, neither with modified intention‐to‐treat (mITT) analysis nor per protocol analysis. Because BMD is only a surrogate endpoint of bone health and PTX did not reduce fracture risk, observation could be considered a safe option for many patients with mild PHPT regarding bone health in a 10‐year perspective. © 2023 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR).

Introduction P rimary hyperparathyroidism (PHPT) is potentially a serious disease with multi-organ involvement. However, in developed countries, PHPT is now most common in patients with only mild or no obvious symptoms, due to more frequent detection by routine clinical testing of calcium levels as part of multichannel blood testing. The only curative treatment is parathyroidectomy (PTX). Although there is consensus that patients with symptomatic PHPT should be offered PTX, (1) there are uncertainties regarding the optimal management of patients with mild PHPT. (2) This is reflected in four subsequent international workshops addressing criteria for surgery in mild PHPT, with extended indications over time. (3)(4)(5)(6) PHPT is known to affect bone health. A continuous excess of parathyroid hormone (PTH) increases bone turnover, potentially leading to a decreased bone mineral density (BMD). (7,8) Conversely, several studies have shown increased BMD after PTX. (8)(9)(10)(11) A long-term observational study found the BMD increase after PTX to persist for up to 15 years. The same study also described stable lumbar spine BMD in patients without PTX but a significant decrease in femoral neck and radial BMD starting just before 10 years of follow-up. (12) Two randomized studies comparing PTX with observation (OBS), with a followup time of 1 to 2 years, have demonstrated positive effects of PTX on lumbar spine and femoral neck BMD but no treatment effect on radial BMD compared with OBS. (13,14) There is, however, a lack of long-term randomized studies, which are needed for deeper knowledge on how PTX affects patients with mild PHPT over time.
We have previously presented 5-year BMD data from our prospective, randomized, controlled study on patients with mild PHPT, Scandinavian Investigation of Primary Hyperparathyroidism (SIPH), comparing PTX with OBS. (15) A treatment effect of PTX, compared with OBS, on bone mineral density was found in four of five DXA compartments, (15) but there was no significant difference in the frequency of vertebral and peripheral fractures between the groups. (16) The ultimate aim of the randomized SIPH study was to compare the long-term (10-year) effects of PTX compared with OBS. Recently, fracture data were presented, showing no difference in fracture frequency between the randomization groups. (17) Here, we present the 10-year BMD data.

Materials and Methods
Participants Between 1998 and2005, 191 patients meeting the NIH 1991 criteria for mild PHPT (3) were included in the SIPH study in Sweden, Norway, and Denmark (ClinicalTrials.gov: NCT00522028). They were randomized (in blocks of 2 to 10 patients at each center) either to PTX (95 patients; 13 men) or OBS (96 patients; 13 men). Inclusion criteria were untreated mild PHPT, no thiazide medication within 4 weeks of calcium and PTH measurements for diagnosis, age 50 to 80 years, albumin-corrected calcium between 2.60 and 2.80 mmol/L, and informed consent. Exclusion criteria were previous neck surgery, osteitis fibrosa cystica, serum-creatinine >130 μmol/L, kidney stones, complicated medical conditions, psychiatric disorders, and familial forms of hyperparathyroidism. The study has previously been described in detail. (18) According to the protocol, the end-of-study (EOS) visit was planned to occur 10 years after inclusion. Of the 191 patients included, 145 patients remained in the study at the 5-year visit and 129 patients concluded the 10-year visit. In the 10-year cohort, 65 patients (9 men) had been randomized to PTX and 64 (7 men) to OBS. No long-term surgical complications (chronic hypocalcemia or recurrent laryngeal nerve palsy) of PTX were revealed. Reasons for early termination and withdrawals, as well as the number of patients included in the EOS analyses, are shown in Fig. 1.
Seventeen patients (17.7%) in the OBS group underwent PTX within the 10-year study period, 4 due to medical predefined protocol criteria and the rest due to patient's decision or nonprotocol specified medical conditions. Thirteen of these concluded the EOS visit (20.3% of the 129 patients concluding the EOS visit, ie, of the 10-year cohort). According to a modified intention-to-treat (mITT; including only the 129 patients in the 10-year cohort) principle, these patients remained in the OBS group for statistical analyses.
The study was approved by the local ethics committees in the respective countries and was conducted according to the Declaration of Helsinki II. All participants gave informed consent.

Biochemistry
Calcium, albumin, and PTH blood levels were measured locally at each center at baseline and at the yearly visits. Data management and albumin correction of calcium levels were performed at the coordinating study center (Oslo University Hospital), and the reported biochemical measurements from baseline visits, 5-year visits, and 10-year visits in the case report forms (CRFs) have subsequently been validated to find and correct for reporting errors.
Dual-energy X-ray absorptiometry (DXA) BMD was measured with DXA at baseline and after 5 and 10 years. Patients were scanned at several different centers over a time span of almost 19 years. Standardized calibrating procedures were applied by all centers. Printouts of all scans were sent to the coordinating center. The dominating manufacturer was Lunar (GE Healthcare, Madison, WI, USA) (270 scans, 106 patients), with six different hardware used (Expert, DPX-L, DPX-IQ, Prodigy, Prodigy Advance, and iDXA). Scanners from Norland (Norland Corp., Fort Atkinson, WI, USA) (27 scans, 9 patients) and Hologic (Hologic Inc., Waltham, MA, USA) (31 scans, 12 patients) were used at one and two centers, respectively.
The BMD-T-score plot by Lundstam and colleagues, a simple tool for detecting random (eg, typos) and systematic BMD data irregularities, based on analysis of the BMD-T-score relationship, (19) was used for DXA BMD data validation from lumbar spine (LS), femoral neck (FN), radius 33% (Rad33), and ultradistal radius (UDR). A description of this process is found in Supplemental File S2.

ITT and per protocol analysis of any fracture in the 10-year cohort
Recently, new (incidental) vertebral and peripheral fractures, occurring after study start, were reported together with the primary endpoint of the SIPH study, mortality. (17) Here, we additionally present an overall analysis of patients with any new fracture in the 10-year cohort with both mITT analysis and per protocol analysis.

Statistical analysis
Depending on data distribution, mean AE standard deviation (SD) or median (interquartile range [IQR]) were used. Differences within and between groups regarding change in calcium and PTH levels (baseline to EOS), were analyzed with dependent and independent samples t test, respectively, as analyzed data were normally distributed. A linear mixed model was fitted to the DXA measurements from the different time points (baseline, 5 years, and 10 years). Based on the model fit, mean values for the randomization groups at the different time points, mean values for within-group change between baseline and 10 years, and mean values for treatment effect of PTX compared with OBS, from baseline to 10 years, were estimated with 95% confidence interval (95% CI). Chi-square test was used for comparing randomization groups with regard to development of any new fracture or not. Stata/SE 16.1 (StataCorp LLC, College Station, TX, USA) was used for the statistical analysis. A p value <0.05 was considered statistically significant.

Baseline characteristics and biochemistry
Data presented here are from the 10-year cohort, the 129 patients who completed the EOS visit. This is in concordance with our previously published results from the patients concluding the 5-year visit. (15,16) Baseline characteristics for the 10-year cohort are shown in Table 1.
There were treatment effects on both albumin-corrected calcium and PTH levels in the 10-year perspective (p < 0.001 and p < 0.001, respectively). Even so, albumin-corrected calcium decreased over time in both groups (OBS: mean difference [diff] À0.09 AE 0.23 mmol/L, p = 0.007; PTX: mean diff À0.30 AE 0.20 mmol/L, p < 0.001). However, if the patients in the OBS group who underwent PTX were excluded, there was no significant change in calcium levels in the OBS group (Fig. 2). There was a decrease in PTH levels over 10 years in the PTX group (mean diff À4.64 AE 5.04 pmol/L, p < 0.001) but no significant change in the OBS group.

DXA
There were treatment effects of PTX compared with OBS for all analyzed compartments with mixed model statistics of BMD and T-scores from baseline to EOS (FN and LS: p < 0.001; forearm: p = 0.006 to p = 0.012), as shown in Table 2. At EOS, the between-group differences in interval change in T-scores (PTX group compared with the OBS group) ranged from 0.41 (Rad33) to 0.58 (LS). T-score differences over time for the two groups are illustrated in Fig. 3.   A decrease in BMD and T-score was noted for all compartments in the OBS group (Table 2, Fig. 3). The mean T-score change ranged from À0.87 (Rad33: p < 0.001) to À0.22 (LS: p = 0.045) ( Table 2). BMD and T-scores also decreased in the forearm compartments in the PTX group (Rad33 BMD: p < 0.001; Rad33 T-score: p < 0.001; UDR BMD: p = 0.043; UDR T-score: p = 0.048). The only compartment with a T-score increase in the PTX group was LS (0.36; p < 0.001; Table 2). New (incidental) fractures in the 10-year cohort, as per ITT and per protocol After exclusion of one patient with an indeterminate vertebral fracture, (17) there were, in the 10-year cohort, 21 of 64 patients (32.8%) in the OBS group and 18 of 64 patients (28.1%) in the PTX group with any (peripheral or vertebral) new fracture, with no difference between groups (p = 0.565; mITT). If the patients with PTX in the OBS group were excluded, 17 of 51 patients (33.3%) had any new fracture, with no difference between groups (p = 0.546; per protocol). Mean T-scores with 95% confidence intervals (95% CI) calculated from raw data (not mixed model derived data) for both randomization groups at baseline, 5 years, and 10 years for (A) lumbar spine, (B) femoral neck, (C) radius 33%, and (D) ultradistal radius. P denotes p value for longitudinal changes from baseline to 10 years within each group. P(groups) denotes p value for longitudinal changes between groups from baseline to 10 years. OBS = observation; PTX = parathyroidectomy. ᵃNumber of validated scans for each compartment in the OBS group. b Number of validated scans for each compartment in the PTX group.

Discussion
In this randomized long-term study on mild PHPT comparing PTX with OBS, we found a significant PTX treatment effect on BMD for all analyzed DXA compartments. However, this treatment effect did not lead to a lower incidence of vertebral or peripheral fractures in patients randomized to PTX, as fracture events were similar between randomization groups. (17) Because fracture outcome is a hard endpoint with relevance for the patient, observation of patients with mild PHPT can be considered safe in a 10-year perspective, with no proven practical bone health benefits of PTX. However, decreasing BMD may be a concern in a longer perspective.
In accordance with our previously published 5-year BMD data, (15) a significant PTX treatment effect compared with OBS was found for LS, FN, and UDR. Cortical bone is known for poor recovery after PTX, (20,21) and our results, showing a decrease in Rad33 BMD in the PTX group, support this. However, the BMD decrease curve appears flatter for the PTX group than for the OBS group after 10 years, resulting in a significant difference between the groups, in contrast to our 5-year data where no difference was found. (15) It seems that the positive effect of PTX compared with OBS on BMD increases with time. Other randomized studies with shorter follow-up times found a positive BMD treatment effect of PTX only on LS and the femur compartments. (13,14) Our previously published 5-year data from the SIPH study found a positive treatment effect of PTX compared with OBS on LS, FN, UDR, and TB BMD. (15) The 10-year data reported in this article could add Rad33 to the list of compartments with a PTX treatment effect compared with OBS. The graphical illustration of T-scores over time (Fig. 3) supports this conclusion. We found a significant decrease in BMD in the OBS group for all compartments in a 10-year perspective. The decrease was most pronounced for the forearm compartments with a mean decrease of almost one T-score unit. There are only a few observational studies with follow-up for 10 years, all with very limited numbers of patients. (12,20,22) Thus, the idea of BMD being stable in observed patients with PHPT are based on either a short follow-up time or very few followed patients, and our results contradict this idea. There are difficulties when comparing studies, as patients with PHPT do not constitute a homogenous group. For example, included patients may be symptomatic and/or asymptomatic, an obvious cause of inhomogeneity. In addition, PHPT presentation differs between postmenopausal women compared with premenopausal women and men, (23) and there are sex differences in patients diagnosed after the age of 50 years, with more osteoporosis in females. (24) The present study was based on the NIH Consensus meeting from 1991, (3) but since then, criteria for surgery have been extended. Based on the T-score ≤ À2.5 at LS, FN, or Rad33 or presence of a vertebral fracture at baseline, 38 (20 OBS/18 PTX) of our patients in the 10-year cohort (29.5%) would have met the 2014 criteria for surgery. (6) The majority met the updated criteria due to low BMD. However, no difference in fracture outcome was found among those meeting the updated surgical criteria; eight new fractures in the OBS group and seven in the PTX group, p = 0.942 (one patient with indeterminate vertebral fracture excluded). Interestingly, the positive BMD effect of PTX did not translate into reduced fracture rates compared with OBS, which could have been expected. Per protocol analysis of fracture frequency, with exclusion of patients in the OBS group who underwent PTX, did not change that. Possible causes include type 2 error or BMD representing bone density and not bone strength. Previous studies have found BMD to be less predictive of fractures in PHPT patients, (25,26) supporting the latter. A recent large-scale, retrospective observational study found PTX to be associated with reduced fracture risk, supporting the former. (27) However, as rightly pointed out by the authors, the retrospective design by itself, the lack of biochemical and bone mineral density data, and uncertainty about the severity of PHPT in the included patients make their conclusions uncertain. As the BMD effect of PTX compared with OBS seems to increase with time, a future difference in fracture outcome might appear. There was only one hip fracture in our study population, (17) a classical fragility fracture. (28,29) In a recent Gothenburg registerbased study, hip fractures were the second most common fracture, with a mean patient age at time of fracture of 81.1 years. (29) The low number of hip fractures in our study is likely explained by our study population being younger. Therefore, one can assume that more patients in our study population will suffer from hip fractures in the future. The overall fracture incidence increases with age, (29) and with the BMD difference between PTX and OBS seemingly increasing with time, a future difference in fracture events cannot be ruled out.
This study's main strength is its randomized, controlled design with a follow-up time of 10 years of 129 patients with defined mild PHPT. The study also has limitations. Because DXA scans were obtained over a period of 19 years from several different machines, the optimal retrieval of raw data and analysis of all scans in one run with the same settings (19) was not possible. However, the BMD-T-score plot facilitated the identification of random and systematic differences and made it possible to correct for most of these. Missing DXA data is also a concern, but mixed model analysis of DXA data allowed all available data to be used statistically, even if the patient did not have data from all three time points. Drugs against osteoporosis were allowed in the study. During the study, 18 of the 129 patients used bisphosphonates at some time point (13 OBS/5 PTX), with a significant difference between groups (p = 0.039). Despite the fact that more patients in the OBS group received bisphosphonates in the 10-year cohort, a treatment effect of PTX compared with OBS on BMD was found. Likewise, estrogens were allowed but without difference in use between the groups. The 1991 NIH conference (3) recommended surgery for patients with creatinine clearance reduced by 30%, which in the pragmatic study protocol was defined for the age group to be reflected by a creatinine level >130 μmol/L. The initially relatively high drop-out rate is also a limitation. Surgically treated patients in the OBS group can be considered a limitation but also reflects a real-world situation with some patients meeting criteria for or wanting surgery at some stage after inclusion in the OBS group. This is also inherent to the modified intention-to-treat principle used for analyses in our study. Nevertheless, the per protocol analysis did not change the conclusions of the study. No power calculations were made for fractures, one of the secondary endpoints of the study. Thus, we cannot rule out a type two error due to the study potentially being underpowered as a cause of no detectable difference in fracture rates between groups.
In conclusion, PTX showed positive treatment effects on BMD for all analyzed compartments compared with OBS. However, BMD can be considered a surrogate endpoint, and we could not demonstrate a treatment effect of PTX on fractures in the 10-year cohort, neither with mITT analysis nor with per protocol analysis. Thus, from a bone health perspective, observation can be considered a safe option in a 10-year perspective for the majority of patients with mild PHPT.

Conflicts of Interest
All authors state that they have no conflicts of interest.