Methods

Top Methods Results Discussion Author & Article Info References

Population-based research is feasible in Rochester because medical care is almost self-contained within the community and the area has relatively few providers. Most endocrinologic care, for example, is provided by the Mayo Clinic, which has maintained a common medical record with its two large affiliated hospitals (St. Marys Hospital and Rochester Methodist Hospital) for 90 years. Recorded diagnoses and surgical procedures are indexed; this includes diagnoses made for outpatients seen in office or clinic consultations, persons seen in the emergency department, and nursing home residents, as well as diagnoses recorded for hospital inpatients, at autopsy examination, or on death certificates [9]. Medical records of the other providers who serve the local population, most notably the Olmsted Medical Group and its affiliate, Olmsted Community Hospital, are also indexed and retrievable. Thus, details of the medical care provided to the residents of Rochester are available for study through this unique medical records linkage system (the Rochester Epidemiology Project). Each year, more than 80% of the local population receives care from the Mayo Clinic or the Olmsted Medical Center and more than half of the local population receives care from the Mayo Clinic. In any 3-year period, more than 90% of Olmsted County residents are seen at the Mayo Clinic at least once. These figures have remained constant for decades [10]. Because of the complete population coverage and the redundancies in this data system, we believe that ascertainment of local residents with recognized primary hyperparathyroidism is complete.

Study criteria described elsewhere [6] yielded an incidence cohort of 90 cases of primary hyperparathyroidism that were diagnosed among Rochester residents between 1965 and 1976. We expanded the study to include all parts of Olmsted County and identified 1335 potential cases of hyperparathyroidism diagnosed from 1976 to 1993 by searching the following diagnostic rubrics: hyperparathyroidism, parathyroid adenoma, osteitis fibrosa cystica, familial benign hypercalcemia, malignant hypercalcemia, and hypercalcemia not otherwise specified. Because mild hypercalcemia is often ignored by practitioners, another 1784 residents in whom serum calcium levels exceeded 2.52 mmol/L at least twice between 1984 and 1993 were selected directly from the Mayo Clinic's Laboratory Information System. This system has preserved the results of all laboratory tests done at the Mayo Clinic from 1984 to the present. To allow patients who were in the process of being examined to be included on the diagnostic index, we assessed the incidence of hyperparathyroidism only through 1992.

For each potential case, the complete [inpatient and outpatient] medical record in the community was reviewed by one of the investigators. Patients were accepted as having a definite case of primary hyperparathyroidism if they met one or more of the following criteria: 1) histopathologic proof of parathyroid adenoma or hyperplasia, 2) hypercalcemia (calcium level >2.52 mmol/L) with an inappropriately elevated serum immunoreactive parathyroid hormone level (>2.1 pmol/L by two-site immunochemiluminometric assay [11] or >20 µLeq/mL by C-terminal radioimmunoassay [12, 13], or 3) hypercalcemia that had lasted longer than 1 year and for which no other cause (such as thiazide diuretics, cancer, creatinine level > 176.8 µmol/L, or lithium therapy) was identified after careful evaluation. The methods used at the Mayo Clinic to measure serum calcium levels changed over time; however, the normal range (2.22 to 2.52 mmol/L) remained the same because the instrumentation was calibrated not against the manufacturer's standard but against atomic absorption spectrophotometry (according to certified references from the National Bureau of Standards). Cut-off points for parathyroid hormone values were based on the clinical performance of the assays at our institution [11]. We also identified two groups of patients with possible hyperparathyroidism: 1) patients with at least two elevated serum calcium levels from at least three determinations who were followed for less than 1 year and 2) patients who had elevated serum calcium levels in at least 2 different years that were followed by three or more normal calcium values. Patients with possible hyperparathyroidism had not been included in the earlier study by Heath and colleagues [6]. Patients with familial benign hypercalcemia who had previously been identified in an extensive study at our institution were excluded [14, 15], as were patients with an incidental autopsy diagnosis of parathyroid adenoma or hyperplasia in whom no elevated serum calcium levels were recorded before death. Patients must also have had established residency in Rochester for at least 1 year before the initial elevated serum calcium level; use of this criterion minimized any effect caused by ill patients with unrecognized hyperparathyroidism moving into the community for care at the Mayo Clinic.

To assess changes in the frequency of primary hyperparathyroidism over time, we calculated apparent incidence rates as of the date of the initial elevated serum calcium level. In contrast, Heath and colleagues [6] used the date on which each case was clinically diagnosed. We report these incidence rates with the understanding that the actual onset of primary hyperparathyroidism may predate the clinical recognition of the condition by a decade or more. This is also true of some other chronic diseases, such as diabetes mellitus. When estimating incidence rates, we considered the entire population to be at risk. To allow our results to be comparable with the findings of Heath and colleagues, long-term trends in incidence were determined only for Rochester residents. For the final decade of our study, rates were estimated not only for Rochester but also for the rest of Olmsted County, which is largely rural. Denominator age- and sex-specific person-years were estimated from decennial census data; linear interpolation was done between census years [16]. We assumed that, given a fixed number of person-years, the number of incidence cases follows a Poisson distribution. This allows for the estimation of SEs and the calculation of 95% CIs for the incidence rates. Rates were directly adjusted for age or for age and sex according to the population structure of white persons in the United States in 1990. Standard errors and CIs for the adjusted rates were based on the same assumptions. Survival was assessed by using standardized mortality ratios; the number of deaths observed in the different periods was compared with an expected number that was estimated by applying age- and sex-specific mortality rates for white persons living in Minnesota to the person-years of follow-up in each subgroup. Ninety-five percent CIs for the standardized mortality ratios were also based on the Poisson distribution.

Results

Top Methods Results Discussion Author & Article Info References

During the study period (1965 to 1992), we identified 475 residents of Rochester (359 women and 116 men) who had definite (92%) or possible (8%) primary hyperparathyroidism. As shown in the top panel of Figure 1, the age- and sex-adjusted incidence rate per 100 000 increased from 5.7 in 1965 to 112.1 in 1975. In 1974, the overall incidence rate was 76.7 per 100 000 person-years; this rate, however, comprised rates of 24.0 per 100 000 person-years in the first half of the year (before the automated serum chemistry panel was introduced) and 129.4 per 100 000 person-years in the final 6 months. Thus, the peak rate was achieved in the second half of 1974. The overall adjusted incidence rate then decreased to 40.7 per 100 000 person-years in 1977 before increasing to 91.0 per 100 000 person-years in 1979. It then began to steadily decrease to 4.0 per 100 000 in 1992. The rates in recent years were lower despite more thorough ascertainment of cases obtained directly from the Mayo Clinic's Laboratory Information System (Figure 1, top).

View larger version (24K): [in this window] [in a new window]   Figure 1. Incidence of definite and possible primary hyperparathyroidism among residents of Rochester, Minnesota, from 1965 to 1992. Top. Incidence curves are presented for total cases (solid line) and cases excluding those identified solely from the Mayo Clinic's Laboratory Information System (dashed line). Although the latter cases could be identified only from 1984 onward, cases recognized in this way could be dated back to the patient's initial elevated serum calcium level through review of existing medical records. Rates were adjusted for age and sex according to the population distribution of white persons in the United States in 1990. Bottom. Incidence rates are presented for women (solid line) and men (dashed line). Rates were adjusted for age alone according to the population distribution of white persons in the United States.

Although the apparent incidence of primary hyperparathyroidism in Rochester was no higher in 1992 than in 1965, the clinical spectrum of the disorder was quite different (Table 1). Overall, most patients had either a pathologic diagnosis (122 patients [26%]) or a parathyroid hormone level that was inappropriately elevated given the degree of hypercalcemia (209 patients [44%]). In only 22% of patients (n = 104) was the diagnosis made by exclusion. The 40 remaining patients (8%) had possible primary hyperparathyroidism. When the study period was divided into thirds, however, the proportion of histologically proven cases decreased from 37% in the prescreening era (1965 through June 1974) to 27% between July 1974 and the end of 1982 to only 17% between 1983 and 1992. Among patients with pathologically confirmed hyperparathyroidism, the diagnosis was parathyroid adenoma in 90% of cases and parathyroid hyperplasia in 10%. This finding was consistent over time. The proportion of patients whose condition was diagnosed on the basis of an inappropriately elevated serum parathyroid hormone level remained fairly constant, whereas the proportion whose condition was diagnosed by exclusion increased from about 21% between 1965 and 1982 to 25% in the final third of the study (Table 1). The rapid decrease in incidence rates during the latter period occurred despite an increased contribution from cases of possible hyperparathyroidism, which increased from 5% of patients or less before 1983 to 18% of all patients between 1983 and 1992.

The 359 women comprised approximately 75% of the cases; this proportion was consistent over time (Table 1). However, the women with primary hyperparathyroidism were older than the men with primary hyperparathyroidism (median ages of 59 and 49 years, respectively; P < 0.001). After adjustment for age, the ratio of incidence rates between women and men was 2.5:1 (incidence rates of 46.3 and 18.1 per 100 000 person-years, respectively). This ratio was also constant over time (Table 2) because the sex-specific trends in incidence were similar (Figure 1, bottom). The mode of diagnosis was similar for men and women and across age groups. Thus, the proportions of patients with pathologic confirmation before and after the mean age of 55 years were 27% and 24%, respectively. Incidence rates were greatest in patients who were 55 to 64 years of age; this finding was consistent across time among the women and varied somewhat among the men (Table 2). Within each age group, the rates increased and decreased with the overall trends in incidence.

During the entire study period, 39 patients (8%) had possible complications of hyperparathyroidism (urolithiasis in 25 patients; osteoporosis, fractures, or both in 5; hypercalcemic crisis in 5; peptic ulcer disease in 2; pseudogout in 1; and band keratopathy in 1). This proportion declined dramatically over time, from 22% in the prescreening era to 8% between 1974 and 1982 to 2% between 1983 and 1992. Most cases were found incidentally, almost always because of an elevated serum calcium level (Table 1). Only three cases were discovered because of some other biochemical abnormality (elevated alkaline phosphatase levels in 1 patient and hypertriglyceridemia leading to pancreatitis in 2 patients). No cases were found as a result of a radiologic abnormality. Two patients were recognized because of an incidental autopsy diagnosis of parathyroid adenoma or hyperplasia, but these patients had elevated serum calcium levels before death. The maximum serum calcium levels did not change: The mean highest levels were 2.72 mmol/L in the first two periods and 2.67 mmol/L in the final period (P = 0.15). Parathyroid surgery was ultimately done in 126 patients (27%). In the prescreening era, surgery was recommended as the initial form of management for 41% of the patients and was done within 6 months of diagnosis in 29%; the other patients refused surgery or were too ill to have surgery (Table 1). Surgery was recommended for 30% of the patients whose condition was diagnosed between July 1974 and 1982 but for only 15% of the patients who had an initial elevated serum calcium level between 1983 and 1992.

These 475 patients were subsequently followed for 6103 person-years (1206 person-years among the cases diagnosed from 1965 to 1974, 3974 person-years for cases diagnosed from 1974 to 1982, and 923 person-years for cases diagnosed from 1983 to 1992). Among the patients from the prescreening era (1965 to June 1974), 26 deaths occurred and 27.4 were expected (standardized mortality ratio, 0.9 [95% CI, 0.7 to 1.4]). In the second period, significantly fewer deaths occurred than were expected (standardized mortality ratio, 0.8 [CI, 0.3 to 0.96]). The reduction in mortality rate was even greater among the patients whose condition was identified from 1983 to 1992 (standardized mortality ratio, 0.5 [CI, 0.1 to 2.0]); however, because only 7 deaths occurred, this result was not statistically significant (P = 0.073).

For the last period of the study (1983 to 1992), we could determine the incidence of primary hyperparathyroidism not only for residents of Rochester but also for residents of rural Olmsted County (1990 population density of 57 persons per square mile, including the urban fringe of Rochester and one small town). These data indicate that apparent incidence rates were almost identical for urban and rural residents (Table 3). The overall age- and sex-adjusted incidence rate for Olmsted County was 21.1 per 100 000 person-years (CI, 18.0 to 24.1 per 100 000 person-years). When only the definite cases were considered, the overall age- and sex-adjusted incidence rate was estimated to be 17.5 per 100 000 person-years (CI, 14.7 to 20.2 per 100 000 person-years). However, the incidence of primary hyperparathyroidism declined in Olmsted County just as it did in Rochester; thus, the age- and sex-adjusted rate decreased from 41.8 per 100 000 in 1983 to 8.0 per 100 000 in 1992.

Discussion

Top Methods Results Discussion Author & Article Info References

After automated measurement of serum calcium levels was introduced, numerous studies reported increased incidence and prevalence of hypercalcemia and hyperparathyroidism. Most observers attributed the increase to more thorough case ascertainment [1, 2, 4, 6, 7, 17-19], which was consistent with recognition that the excess incidence almost entirely reflected patients who were free of the complications of hyperparathyroidism and had only modestly elevated serum calcium levels [20]. Indeed, 92% of Rochester residents whose primary hyperparathyroidism had been recognized since July 1974 came to clinical attention because of an incidental finding of an abnormal serum calcium level. Other investigators have reported similar results [1, 4, 8, 21, 22]. In addition to this improved ascertainment of a class of patients who had hyperparathyroidism not previously identifiable on clinical grounds, other cases were no doubt diagnosed earlier than they would have been before the introduction of automated measurement of serum calcium levels. This accelerated rate of diagnosis would be expected to cause a compensatory decrease in the incidence of primary hyperparathyroidism in subsequent years. Indeed, Heath and colleagues [6] documented a 46% reduction in the incidence of primary hyperparathyroidism, from 51.1 per 100 000 person-years between July 1974 and June 1975 to 27.7 per 100 000 person-years between July 1975 and December 1976 [6]. The latter Figure was still higher than the rate seen before the introduction of automated measurement of serum calcium levels. Heath and colleagues believed that this rate represented the true incidence rate in an era of routine measurement of calcium levels. Other contemporary estimates were similar [7, 8], but no studies had determined trends in the incidence of primary hyperparathyroidism in the United States since 1976.

One might have expected the incidence of primary hyperparathyroidism to stabilize, but rates in Rochester decreased steadily after 1979. This was unexpected, especially because our inclusion of cases identified solely from the Mayo Clinic's Laboratory Information System since 1984 should have introduced a positive bias by further improving ascertainment through the inclusion of cases that had been clinically overlooked [23-26]. Thus, the overall rate for Olmsted County in 1983 was 41.8 per 100 000; it had decreased to 8.0 per 100 000 by 1992. Other studies have reported low incidence rates, but these rates were primarily based on cases of primary hyperparathyroidism that were treated surgically and therefore reflect only a fraction of the total [19, 27-29]. The decrease in incidence rates was not an artifact of less frequent measurement of serum calcium levels because about 20% of Olmsted County residents were tested each year. In addition, between 1984 and 1993, this Figure varied only from 18.8% to 21.1%. Although part of this rate represented the performance of several tests on the same person, an estimated 66% of the population aged 45 years and older (70% of women and 61% of men) had at least one serum calcium determination in the 3-year period (1991 to 1993).

One possible explanation for the decreasing incidence could be increasing use of estrogen replacement therapy. The appearance of hyperparathyroidism in women is often related to menopause [30, 31], and the excess of hyperparathyroidism among postmenopausal women is well known [20]. In our study, incidence rates were the same for women and men before age 45 years but were greater among women at all ages thereafter. National trends in the use of estrogen-containing prescriptions, however, are not consistent with the trends we report for primary hyperparathyroidism [32-34]. In addition, only 10% to 20% of postmenopausal women in the United States receive long-term estrogen replacement therapy [34-36]. Moreover, identical trends in the incidence of hyperparathyroidism were seen among male residents of Rochester, for whom estrogen replacement therapy is not a relevant explanation. The fact that the same trends were seen in both middle-aged and elderly persons also discounts the possibility that the decrease was due to an increase in exclusionary conditions.

Another possibility is a systematic dietary change in the population that has modified the risk for the disease. Numerous reports indicate that parathyroid hormone levels increase with age [37-42]; studies from our group have shown changes in parathyroid hormone secretory dynamics in elderly women that are consistent with parathyroid hyperplasia [40] and that can be reversed by long-term increases in dietary calcium intake [43]. These studies suggest that the calcium requirement probably increases with age, perhaps because of age-related decreases in intestinal calcium absorption [44-46] or renal calcium conservation [40, 47, 48]. Thus, failure of the elderly to meet this increased requirement by adequately increasing calcium intake could provide a chronic stimulus to the parathyroid glands and lead to secondary hyperparathyroidism [49]. Chronic stimulation may also increase the likelihood of adenomatous transformation of the parathyroid glands [50]. Further studies assessing changes in dietary calcium intake over time are needed to address this hypothesis.

A final, intriguing possibility is the potential adverse effect of exposure to ionizing radiation [6]. A fourfold increase in the risk for hyperparathyroidism was found among atomic bomb survivors from Hiroshima [51], and the association was seen at lower radiation exposures (mean dose, 0.4 Gy) than those previously reported [52]. More recently, researchers documented an 11-fold increase in the risk for hyperparathyroidism per 1.0-Gy dose of radiation to the head and neck [53]. Therefore, exposure to relatively low levels of radiation might promote parathyroid hyperplasia or adenoma and lead to hyperparathyroidism decades later. Many studies have reported such an association. In one study done in Rochester, for example, the risk for histologically proven primary hyperparathyroidism associated with previous therapeutic radiation to the head and neck was increased 2.3-fold [54]. The excess risk was confined to women and was related to radiation exposure that occurred an average of 29 years earlier, mostly for treatment of such benign conditions as acne. However, only one quarter of the patients, whose primary hyperparathyroidism was diagnosed between 1975 and 1983, had a history of radiation treatment. Other studies have found that 1% to 31% of patients with hyperparathyroidism have a history of radiation exposure [18, 55-62]. Since the association between radiation to the head and neck and the development of thyroid cancer was recognized [63, 64], the use of radiation has decreased significantly [65]. Therefore, the changing incidence of primary hyperparathyroidism could reflect changes in the number of patients previously exposed to radiation to the head and neck.

In summary, our study is the first to systematically examine long-term changes in the incidence of primary hyperparathyroidism. Despite more thorough ascertainment of cases in recent years, we found a progressive decline in the incidence of primary hyperparathyroidism. Although several factors may explain this phenomenon, the precise cause of the decline remains unclear. A better understanding of the factors responsible might lead to public health measures (for example, increased dietary intake of calcium) to further reduce the medical and socioeconomic effect of this disease in the population. However, the unexpected decline in the incidence of primary hyperparathyroidism has immediate practical implications with respect to the prior probability of disease and, thus, to the clinical usefulness of measuring serum calcium levels as a screening technique. Moreover, reductions in the use of serum calcium testing in response to efforts to control costs promise a return to an era in which most clinically recognized cases of primary hyperparathyroidism were symptomatic. Because only 2% of patients recognized in the past decade had complications of the disease, however, optimal management even of patients with clinically recognized primary hyperparathyroidism will probably remain controversial.

Drs. Khosla, Atkinson, Hodgson, O'Fallon, and Melton: Mayo Clinic, 200 First Street Southwest, Rochester, MN 55905.