Primary Hyperparathyroidism and the Risk of Fracture: A Population-Based Study

Authors

  • Sundeep Khosla,

    Corresponding author
    1. Division of Endocrinology and Internal Medicine, Mayo Clinic and Mayo Foundation, Rochester, Minnesota, U.S.A.
    • Address reprint requests to: Sundeep Khosla, M.D. Mayo Clinic 200 First Street SW 5–164 West Joseph Rochester, MN 55905 U.S.A.
    Search for more papers by this author
  • L. Joseph Melton III,

    1. Department of Health Sciences Research, Mayo Clinic and Mayo Foundation, Rochester, Minnesota, U.S.A.
    Search for more papers by this author
  • Robert A. Wermers,

    1. Division of Endocrinology and Internal Medicine, Mayo Clinic and Mayo Foundation, Rochester, Minnesota, U.S.A.
    Search for more papers by this author
  • Cynthia S. Crowson,

    1. Department of Health Sciences Research, Mayo Clinic and Mayo Foundation, Rochester, Minnesota, U.S.A.
    Search for more papers by this author
  • W. Michael O'Fallon,

    1. Department of Health Sciences Research, Mayo Clinic and Mayo Foundation, Rochester, Minnesota, U.S.A.
    Search for more papers by this author
  • B. Lawrence Riggs

    1. Division of Endocrinology and Internal Medicine, Mayo Clinic and Mayo Foundation, Rochester, Minnesota, U.S.A.
    Search for more papers by this author

Abstract

While severe primary hyperparathyroidism (HPT) is clearly associated with osteitis fibrosa cystica, it remains uncertain whether mild, asymptomatic primary HPT adversely affects the skeleton. Thus, we assessed the incidence of age-related fractures in a large, population-based inception cohort of 407 cases of primary HPT (93 men and 314 women) recognized during the 28-year period, 1965–1992. Fracture risk was assessed by comparing new fractures at each site to the number expected from gender- and age-specific fracture incidence rates for the general population (standardized incidence ratios, SIRs). These community patients with primary HPT mostly had mild disease (mean ± SD serum calcium, 10.9 ± 0.6 mg/dl). Altogether, 471 fractures occurred during 5766 person-years of follow-up. Overall fracture risk was significantly increased in these patients (SIR 1.3, 95% confidence interval [CI] 1.1–1.5). Primary HPT was associated with an increased risk of vertebral (SIR 3.2, 95% CI 2.5–4.0), distal forearm (SIR 2.2, 95% CI 1.6–2.9), rib (SIR 2.7, 95% CI 2.1–3.5), and pelvic fractures (SIR 2.1, 95% CI 1.1–3.5). The risk of proximal femur fractures was only marginally increased (SIR 1.4, 95% CI 1.0–2.0). By univariate analysis, increasing age and female gender were significant predictors of fracture risk, although higher serum calcium levels were also associated with increased fracture risk, and parathyroid surgery may have had a protective effect. By multivariate analysis, however, only age (relative hazard [RH] per 10-year increase, 1.6, 95% CI 1.4–1.9) and female gender (RH 2.3, 95% CI 1.2–4.1) remained significant independent predictors of fracture risk. Thus, primary HPT among unselected patients in the community is associated with a significant increase in the risk of vertebral, Colles', rib, and pelvic fractures. These data have important implications for the current trend to recommend nonsurgical management for patients with mild primary HPT.

INTRODUCTION

The introduction of multichannel biochemical testing, which included automated serum calcium determinations, led to a marked increase in the detection of asymptomatic patients with mild primary hyperparathyroidism (HPT).(1,2) While the only recent population-based data suggest that the true incidence of primary HPT may be declining,(3) it still remains a relatively common disorder in clinical practice. Moreover, most patients continue to be diagnosed on the basis of asymptomatic elevations in serum calcium levels rather than signs or symptoms specifically related to HPT.(3) Since it is often unclear which patients should undergo parathyroidectomy, a National Institutes of Health (NIH) Consensus Development Conference agreed upon a list of guidelines for surgery in primary HPT.(4) These included: serum calcium above 12 mg/dl, marked hypercalciuria (>400 mg/day), any overt manifestation of primary HPT (nephrolithiasis, osteitis fibrosa cystica, or classic neuromuscular disease), markedly reduced cortical bone mineral density (BMD), reduced creatinine clearance in the absence of other cause, and age <50 years. On the basis of these recommendations, a substantial proportion of primary HPT patients are currently managed conservatively in the community.(3,5)

With these guidelines, a key component in the decision regarding surgery in patients with primary HPT is their skeletal status.(4) While osteitis fibrosa cystica is rare in current practice, even mild primary HPT may have significant effects on the skeleton, including possible reductions in BMD and increases in bone turnover.(6) However, there are sparse population-based data at present on fracture risk in the patient with mild primary HPT, who represents the type most commonly encountered in today's general medical practice. Thus, a majority of primary HPT patients in the community are currently being managed nonoperatively,(3,5) and it is generally assumed that they are not at significantly increased risk of fracture. To test the validity of this assumption, we assessed the incidence of age-related fractures in a large, population-based cohort of 407 Rochester, Minnesota residents with primary HPT recognized during the 28-year period, 1965–1992, and compared this with the incidence in the general population.

MATERIALS AND METHODS

Population-based epidemiologic research can be conducted in Rochester, Minnesota because medical care is virtually self-contained within the community and there are relatively few providers. Most endocrinologic and orthopedic care, for example, is provided by the Mayo Clinic, which has maintained a common medical record system with its two large affiliated hospitals in the community (Saint Marys and Rochester Methodist) for over 90 years. The diagnoses and surgical procedures recorded in these records are indexed, as are the medical records of the other providers who serve the local population, most notably the Olmsted Medical Group and its affiliated Olmsted Community Hospital.(7) Following approval by Mayo's Institutional Review Board, we used this unique database (the Rochester Epidemiology Project) to identify 435 Rochester residents with primary HPT during the period, 1965–1992.(3,5) Seven of these patients refused subsequent authorization for chart review,(8) and 21 had no follow-up after age 35 years, resulting in a final cohort of 407 subjects. Details regarding the specific criteria used to identify these patients have been published previously.(3,5)

These subjects were then followed forward in time through their linked medical records in the community (retrospective cohort study) until death or the most recent clinical contact, and backward in time to the first medical record entry in the community. For consistency, the index date for the diagnosis of primary HPT was the date when hypercalcemia was first evident, and not when the clinician recorded the diagnosis of primary HPT. For each subject, all inpatient and outpatient medical records at any local provider of medical care were searched for the occurrence of specific fractures. Mayo Clinic records, for example, contain the details of every inpatient hospitalization at its two hospitals, every outpatient office or clinic visit, emergency room and nursing home care, as well as radiographic reports and pathology reports, including autopsies.(9) Emphasis was on fractures at the skeletal sites usually associated with osteoporosis, and these were recorded regardless of whether they occurred before or after the recognition of primary HPT. The records contained the clinical history and the radiologist's report of each fracture, but the original roentgenograms were not available for review. Thus, the diagnosis of vertebral fracture was accepted on the basis of a radiologist's report of compression or collapse of one or more thoracic or lumbar vertebrae. All fractures were classified according to the circumstances of the injury. By convention, falls from standing height or less were considered moderate trauma, while motor vehicle accidents and falls from heights were deemed severe trauma. Ascertainment of the fractures of interest is believed to be complete except for vertebral fractures, some of which are never diagnosed.(10)

The influence of primary HPT on fracture incidence was evaluated using three basic methods of analysis. In the primary analysis, we calculated standardized incidence ratios (SIRs), comparing the number of fractures that were observed at each skeletal site (based on the first fracture of a given type per person) to the number expected in this cohort during their follow-up in the community. Expected numbers were derived by applying age- and gender-specific incidence rates from the general population for these fractures to the age- and gender-specific person-years of follow-up in the cohort. Incidence rates from the general population of Rochester (age ≥35 years) were available for fractures of the distal forearm (Colles'),(11) proximal femur,(12) proximal humerus,(13) pelvis,(14) and vertebrae,(10) as well as other specific fracture sites.(15) Fractures that occurred prior to the diagnosis of primary HPT were also assessed and compared with the number expected. Ninety-five percent confidence intervals (95% CI) for the SIRs were calculated assuming that the expected rates are fixed and the observed fractures follow a Poisson distribution.(16)

In the second method of analysis, the cumulative incidence of new fractures (1 minus survival-free-of-fracture) was projected for up to 20 years following the initial diagnosis of primary HPT using product-limit life table methods.(17) Cumulative incidence curves were compared with the log-rank statistic.(18)

Finally, Cox proportional hazards models,(19) which do not incorporate the population expected rates, were used to assess the impact of various covariates (age at HPT diagnosis, year of diagnosis, gender, highest serum calcium levels, and parathyroid surgery [yes/no]) on the risk of fracture. Parathyroid surgery was treated as a time-dependent covariate. The stepwise process was used for model selection. Interactions among the significant main effects were assessed.

RESULTS

During the 28-year study period, 1965–1992, 407 cases of primary HPT diagnosed among the residents of Rochester, Minnesota constituted the inception cohort. The majority were women (314, 77%) and most were 45 years of age or older at the diagnosis of HPT (335, 82%). The mean age at diagnosis was 57.8 years (median, 58.5 years; range, 18.5–89.4 years). The average maximum serum calcium level (mean ± SD) was 10.9 ± 0.6 mg/dl (normal range, 8.9–10.1 mg/dl). Median serum PTH (reported as such since it was highly skewed), measured by a C-terminal assay, was 47 μl · eq/ml (25–75% interval, 33–71 μl · eq/ml; normal, ≤50 μl · eq/ml). The majority of the patients were asymptomatic, and the relative proportion of patients presenting with a symptom or complication of primary HPT (urolithiasis, fracture as a presenting symptom, hypercalcemic crisis, peptic ulcer disease, pseudogout, or band keratopathy) declined as biochemical screening was introduced in this population: 21% in the prescreening era (1965–June 1974), 9% following the introduction of automated serum calcium determinations (July 1974–1982), and 2% in the postscreening era (1983–1992). Most of these patients were managed conservatively, with parathyroid surgery ultimately performed on only 93 patients (23%).

The cohort of 407 patients with primary HPT was followed for 5766 person-years, and 295 patients (73%) were still alive at last follow-up. A total of 471 fractures occurred in 202 patients after the index date. The types of fractures and their causes are summarized in Table 1. The majority of fractures of the vertebrae (92%), distal forearm (91%), pelvis (75%), and proximal femur (90%) were due to mild or moderate trauma. Fracture cause was coded as “uncertain” if the etiology of the fracture was not clearly stated in the medical records or there was more than one possible cause. This explains the relatively high number of uncertain causes, especially among rib fractures, most of which probably occurred after mild or moderate trauma.

Table Table 1. Fractures by Skeletal Site and Cause Among 407 Rochester, Minnesota, Residents Following the Initial Diagnosis of Primary HPT in 1965–1992
original image

Figure 1 depicts the SIRs for fractures at selected skeletal sites as well as the overall SIR for any fracture in this cohort. More detailed data are provided in Table 2. As is evident, there was greater than a 3-fold increase in the risk of vertebral fractures and a 2- to 2.5-fold increase in the risk of distal forearm, rib, and pelvic fractures following a diagnosis of primary HPT. The risk of a proximal femur fracture was increased by ∼40%, although the 95% CI for this estimate included 1.0. The overall risk of any fracture was increased significantly by about 30%, due largely to the increase in fracture risk at the sites shown in Fig. 1. Fracture risk at the skeletal sites other than those shown in Fig. 1 was not significantly increased in the primary HPT patients, with the exception of the combined fracture risk at the sternum/clavicle/scapula (Table 2). When each of the latter sites was considered individually, however, the 95% CI for the SIR estimates included 1.0 (data not shown). It should be noted that, while the average of the SIRs in Table 2 is 1.7, the overall SIR is somewhat lower (1.3) since the analysis for each type of fracture separately and all fractures combined was based on the first fracture of a given type per person (see Materials and Methods). Thus, for example, a person who sustained first a vertebral fracture and then a rib fracture would be counted under both (vertebral and rib) categories, but only once (on the basis of the vertebral fracture) in the “any fracture” category.

Table Table 2. Observed (Obs) Fractures in Comparison with Expected Numbers (Exp) and Standardized Incidence Ratios (SIRs) at All Skeletal Sites Among 407 Rochester, Minnesota, Residents with Primary HPT First Diagnosed in 1965–1992
original image
Figure FIG. 1.

SIRs and 95% CIs of selected fractures among 407 Rochester, Minnesota residents following the first diagnosis of primary HPT in 1965–1992 (ratio of 1.0 indicates no increased risk).

Figure 2 shows the observed cumulative incidence curves for vertebral, distal forearm, rib, and all fractures in the primary HPT patients compared with the expected rates in this population. Consistent with the SIRs shown in Fig. 1, there was a significantly greater cumulative incidence of fractures at these specific skeletal sites as well as overall fractures in the primary HPT patients.

Figure FIG. 2.

Observed and expected cumulative incidence (1 minus survival free-of-fracture) of any vertebral fracture (A), distal forearm (Colles') fracture (B), rib fracture (C), and all fractures (D) among 407 Rochester, Minnesota residents following the initial diagnosis of primary HPT in 1965–1992. p values < 0.001 for all of the observed versus expected curves.

To better define the effect of duration of primary HPT on fracture risk at the above sites, a time-dependent analysis was performed, which revealed that the SIRs at these selected skeletal sites were already elevated within the first 5 years after diagnosis (Table 3). The SIRs remained relatively constant over time for vertebral and rib fractures. There did appear to be a decrease in the SIR for distal forearm fractures after 15 years and for any fracture after 10 years, although the 95% CIs for these SIRs were relatively wide.

Table Table 3. Standardized Incidence Ratios (SIRs) as a Function of Time Since Diagnosis at Selected Skeletal Sites Among 407 Rochester, Minnesota, Residents with Primary HPT First Diagnosed in 1965–1992
original image

Since automated screening for serum calcium levels was introduced into this population in July, 1974, we also assessed whether there was a significant difference in fracture risk between the individuals diagnosed before versus after this date (Table 4). The patients in the prescreening era were slightly, but not significantly, younger than the patients in the postscreening era (55 ± 15 years vs. 58 ± 15 years, respectively), but the age distribution of the whole population was younger then, too. When differences in age were taken into account by the calculation of SIRs, fracture risk was not higher in the prescreening versus the postscreening cohort; if anything, the SIRs tended to be lower in the prescreening cohort, perhaps related to limited power to detect increases in fracture risk due to the much smaller number of subjects in this group.

Table Table 4. Standardized Incidence Ratios (SIRs) at Selected Skeletal Sites Among 58 Rochester, Minnesota, Residents with Primary HPT First Diagnosed in January 1965–June 1974 Versus 349 Rochester Residents with Primary HPT First Diagnosed in July 1974–December 1992
original image

While fracture risk at these sites was clearly increased following the diagnosis of primary HPT, the risk of vertebral, distal forearm, pelvic, proximal femur, or any fracture was no different from expected in these patients prior to the development of primary HPT (data not shown).

Table 5 shows the results of the univariate and multivariate analyses for predictors of vertebral, distal forearm, or proximal femur fractures due to mild/moderate trauma in the primary HPT patients. These are traditionally considered osteoporotic fractures, and they constitute the majority of osteoporotic fractures in the population. By univariate analysis, increasing age and female gender were clearly predictive of fractures at these sites. Highest serum calcium was positively associated with fracture risk, although the 95% CI for this estimate included 1.0. Parathyroid surgery appeared to be protective, although again the 95% CI was relatively large due principally to the small number of primary HPT patients (23%) that underwent parathyroid surgery in this population. Serum PTH was allowed to enter the model, but was not a significant predictor of fracture risk. By multivariate analysis, only increased age and female gender remained as independent predictors of an elevated fracture risk.

Table Table 5. Predictors of the Risk of Mild/Moderate Trauma Vertebral, Distal Forearm (Colles'), or Proximal Femur Fractures Among 407 Rochester, Minnesota, Residents with Primary HPT First Diagnosed in 1965–1992
original image

DISCUSSION

The potential adverse skeletal consequences of mild, asymptomatic primary HPT have been the subject of ongoing controversy, particularly since many such patients are managed conservatively in the community.(3,5) Our findings indicate that these patients have a significantly greater than expected risk of vertebral, distal forearm, rib, and pelvis fractures, resulting in a greater overall risk of fracture. The risk of hip fractures appears to be marginally, if at all, increased. By univariate analysis, increasing age and female gender were significant predictors of fracture risk, although higher serum calcium levels were also associated with increased fracture risk, and parathyroid surgery may have had a protective effect. Only age and female gender remained significant independent predictors by multivariate analysis. These findings, thus, have potentially important implications for the current trend, based largely on the recommendations of the NIH Consensus Conference,(4) for nonsurgical management of patients with primary HPT.

In a previous study of a cohort of 90 primary HPT patients,(20) we reported that overall fracture risk was increased in these patients prior to the diagnosis of HPT, but not afterward. In contrast, in the present, larger cohort of 407 patients, we demonstrate that the increase in fracture risk primarily follows the diagnosis of primary HPT. There are important differences between the two studies that may explain the somewhat different results. First, because of the much larger number of subjects, the present study had considerably greater power to detect increases in fracture risk. Second, since the majority of the patients in the previous study (which spanned the time period of 1965–1976) had their disease diagnosed prior to the introduction of routine screening of the population, they likely had had primary HPT for a much longer period prior to diagnosis than the subjects in the present study, and thus had a greater “opportunity” to fracture prior to diagnosis. A third possible difference is that, in the previous study, the date of diagnosis was based on the clinical diagnosis in the medical record. In the present study, however, we assigned the index date as the first recorded elevation in serum calcium levels, rather than the later date when the clinicians noted the diagnosis. The latter is unlikely to explain the different findings of the two studies, however, since the average difference in time from discovery of hypercalcemia to diagnosis of primary HPT was only about 1 year in the previous cohort.

The major strength of our study is that it is based on all identifiable patients with primary HPT in the community and our results are therefore applicable to the average patient with mild primary HPT that is commonly seen in current practice. In addition, the clinical characteristics and diagnosis of primary HPT were recorded prior to any knowledge of resultant fractures, and fracture ascertainment should be nearly complete since the medical records contained detailed documentation of the fractures that occurred during each subject's residency in the community, the vast majority of which come to medical attention.(15) Finally, fracture risk was assessed at all skeletal sites. In contrast, all previous studies of this issue have been done in selected patients, with fractures assessed only at a limited number of sites. Thus, Larsson et al.(21) studied hospitalized patients and examined only the risk of hip fractures. Consistent with our data, they found that HPT was not associated with a significant increase in hip fracture risk. Similarly, an earlier study from this group compared the incidence of distal forearm fractures in 39 women with primary HPT versus age-matched controls and, again consistent with our data, found a significant increase in distal forearm fractures in these patients.(22)

Several other studies have assessed fracture risk in selected patients with primary HPT, with somewhat conflicting results. In a case-control study from our institution, Dauphine et al.(23) found an increased prevalence of vertebral fractures in referral patients with surgically proven primary HPT. In contrast, Wilson et al.(24) found no increase in the prevalence of vertebral fractures in patients with primary HPT attending a bone and mineral metabolism clinic. Kochersberger et al.(25) compared lateral chest roentgenograms of 206 patients from a registry of parathyroidectomies with similar X-rays from a control group of patients undergoing cholecystectomy and found a significantly greater prevalence of vertebral fractures in the primary HPT patients. Finally, Kenny et al.(26) studied 46 women following parathyroidectomy and found that a greater percentage of the primary HPT patients had a history of fractures compared with the controls. Thus, with the exception of the study of Wilson et al.,(24) these previous studies are consistent with our finding of a greater risk of fracture in primary HPT patients. However, since they involved selected patients, and primarily those treated surgically, these studies clearly need to be interpreted with caution.

While our data provide estimates of the increased fracture risk at specific skeletal sites in primary HPT patients, we cannot address the mechanism of this effect. Several cross-sectional studies assessing BMD in primary HPT patients have been published, with somewhat varying results. Cortical bone appears to be most affected in primary HPT patients,(22,27,28) although some studies have noted cancellous osteopenia,(29–32) at least in a subset of patients.(33) In contrast to the BMD data suggesting that cortical bone is preferentially affected in primary HPT, our data indicate that fracture risk is increased primarily at sites relatively enriched in cancellous bone. The vertebrae, for example, contain ∼70% cancellous bone, and the ribs and pelvis also contain significant amounts of cancellous bone.(34) The vast majority of distal forearm fractures in this study were Colles' fractures, which also occur at sites containing ∼70% cancellous bone.(35) The reasons for this discrepancy are at present unclear. It is possible, for example, that decreases in cortical bone mass, even at sites relatively enriched in cancellous bone, contribute to the increased fracture risk at these sites. Alternatively, increased fracture risk at the cancellous bone sites may, in part, be related to the fact that primary HPT is also associated with increased bone turnover.(36–38) Recent data indicate that increased bone turnover may represent an independent risk factor for fracture,(39,40) perhaps due to perforative resorption of trabeculae and loss of structural elements of bone(41) or a reduction in bone strength due to an enlarged remodeling space,(39,42) and these effects may be particularly detrimental at cancellous bone sites. However, cancellous bone structure has been reported to be preserved in primary HPT, at least in patients without vertebral fractures, and trabeculae tend not to perforate in primary HPT as they do in the case of estrogen deficiency.(43–45) Nonetheless, parathyroidectomy in primary HPT patients is associated with significantly greater increases in BMD at cancellous as opposed to cortical bone sites,(33,46) consistent with significant effects of excess PTH on cancellous bone. Clearly, more studies are needed to address these issues.

Given the nature of this retrospective cohort study, we do not have interpretable BMD data in this cohort. The patients were seen over the wide time span of 1965–1992 and, since they were seen in the community rather than specifically in the Metabolic Bone Clinic, BMD was not always obtained in these subjects. When it was obtained, it was on different clinical instruments over the years, without any consistent cross-calibration across instruments, and with variable reporting formats (i.e., absolute BMD values, percentiles, and more recently, T scores). Similarly, we do not have consistent information on other factors such as urinary calcium values, creatinine clearance, or use of hormone replacement therapy. Thus, we cannot address issues such as the percentage of patients that actually met NIH guidelines for surgery(4) but were, in fact, followed medically.

We also recognize that there is the possibility of selection bias in our study that may lead to an overestimation of the relative risk of certain fractures. Our data indicate, for example, that the risk of vertebral, rib, and distal forearm fractures was increased by approximately the same amount in patients with primary HPT. However, since these patients (and their physicians) are “aware” of their disease, they may be more likely to have spine X-rays for back pain, and possibly rib X-rays for chest pain, leading to increased detection of vertebral and rib fractures relative to the control population. However, this would clearly not be the case for distal forearm fractures. Thus, it is possible that the relative risk of distal forearm fractures is, in fact, higher than the relative risk of vertebral or rib fractures in patients with primary HPT, but due to an overestimate of the risk of vertebral and rib fractures, our data may not reflect this. As part of our analysis, we also searched for any consistent differences in the clinical characteristics of patients who sustained vertebral vs. distal forearm fractures, but failed to identify any unique features of the vertebral fracture group (data not shown).

There has been considerable interest recently in the use of intermittent PTH therapy as a treatment for osteoporosis,(47,48) Our findings of an increased risk of fracture in primary HPT patients does not, however, militate against such an approach. It is clear that continuous exposure of bone to PTH, as would be expected in primary HPT, can decrease bone mass.(49,50) In contrast, intermittent PTH exposure, as in daily injections of PTH, leads to a stimulation of bone formation and a net increase in bone mass.(49–52) Thus, intermittent PTH treatment remains a promising anabolic therapy for osteoporosis, and our data on fracture risk in primary HPT patients should not be extrapolated to fracture risk in patients treated with daily injections of PTH.

Definitive prospective studies using fracture as an endpoint will be required to better define the precise fracture risk associated with primary HPT or with different treatments for HPT. This represents a significant undertaking, however. Based on our data, in order to detect a 2-fold increase in vertebral or distal forearm fractures in primary HPT patients with 90% power, ∼900 primary HPT patients and an equal number of age-matched controls would have to be followed for at least 5 years. In the absence of such a study, our present population-based data provide the best available estimates for fracture risk in primary HPT patients. These data indicate that even patients with mild primary HPT in the community are at increased risk of vertebral, distal forearm, rib, and pelvic fractures. Our data also have important implications for conservative management of these patients in the community. They reinforce the need for careful evaluation of the skeletal status in these patients, with early surgical intervention in those patients with significant decreases in bone mass and/or ongoing bone loss.

Acknowledgements

We thank Ms. Celia Wright for help with data collection and Ms. Beth Atkinson for assistance with the statistical analyses. This project was supported in part by grants AG-04875 and AR-30682 from the National Institutes of Health, United States Public Health Service.

Ancillary