|Home | About | Journals | Submit | Contact Us | Français|
Hypercalcemic crisis (HC) is a potentially life-threatening manifestation of primary hyperparathyroidism (PHPT). This study aims to identify patients with PHPT at greatest risk for developing HC.
This retrospective cohort study included patients with a pre-operative calcium of at least 12mg/dL undergoing initial parathyroidectomy for PHPT from 11/2000–03/2016. We compared those with HC, defined as needing hospitalization for hypercalcemia, to those without HC.
The study cohort included 29(15.8%) with HC and 154(84.2%) without. Demographics and comorbidities were similar between groups. Patients with HC were more likely to have a history of kidney stones (31.0% vs. 14.3%, P=0.039), higher pre-operative calcium (median 13.8 vs. 12.4mg/dL, P<0.001), higher parathyroid hormone (PTH) (median 318 vs. 160pg/mL, P=0.001), and lower vitamin D (median 16 vs. 26ng/mL, P<0.001) than patients without HC. Cure rates with parathyroidectomy were similar, but nearly double the proportion of patients with HC had multi-gland disease (24.1 vs. 12.3%, P=0.12). In multivariable analysis, higher pre-operative calcium (OR 1.7, 95% CI 1.1–2.5), higher PTH (OR 1.0, 95% CI 1.0–1.0), and kidney stones (OR 3.0, 95% CI 1.1–8.2) were independently associated with HC. A Classification and Regression Tree revealed that 91% of patients with a calcium ≥13.25 mg/dL and a Charlson Comorbidity Index ≥4 developed HC.
These data indicate that calcium, PTH, and kidney stones are important in predicting who is at greatest risk of HC. The Classification and Regression Tree can further help stratify risk for developing HC and allow surgeons to expedite parathyroidectomy accordingly.
Hypercalcemic crisis (HC) is an uncommon, potentially fatal complication of primary hyperparathyroidism (PHPT) that impacts numerous organ systems and requires acute hospitalization for management. While some elevations in serum calcium can be well tolerated, symptoms of crisis are severe. Presentations of HC vary but may include nausea, vomiting, pancreatitis, profound weakness, cognitive disturbances, somnolence, acute kidney injury, and cardiac arrhythmias (1–3). If treatment is delayed, HC can result in death (2).
Primary hyperparathyroidism (PHPT) is the most common cause of hypercalcemia with or without crisis (2, 4). Treatment for HC begins with aggressive hydration and determination of the underlying cause. Parathyroidectomy provides the only cure for HC caused by PHPT. Removing the overactive gland or glands lowers the levels of parathyroid hormone (PTH) released into the blood thereby decreasing serum calcium (5, 6). In contrast to the majority of parathyroid surgery, which is largely elective, parathyroidectomy for HC is urgent and recommended during the same admission.
It remains unclear, however, which patients with PHPT are at greatest risk of developing HC. This knowledge would allow providers to expedite parathyroidectomy in patients with increased risk thereby avoiding hospitalization and complications associated with HC. Thus, the primary purpose of this study is to identify risk factors of HC in patients with PHPT. The secondary aim of this study is to characterize the presentation, operative outcomes, and pathology of patients with HC.
We performed a single-institution retrospective cohort study of patients who underwent parathyroidectomy from November 2000 to March 2016. Data were collected prospectively and the University of Wisconsin Institutional Review Board approved the study. Some data were retrospectively gathered from the medical record including details on hospital admission, less common symptoms of hypercalcemia, and detailed medication lists. Patients included in the overall cohort had PHPT with a pre-operative calcium level greater than or equal to 12 mg/dL. We used a minimum calcium level to limit our analysis to patients with a reasonably high risk of developing HC and to reduce the total number of cases available in the database to a manageable size (2). We excluded patients if they had prior parathyroid surgery, a family history of related genetic syndromes, or age less than 18 years. Patients taking lithium were also excluded because lithium has been shown to increase serum calcium and PTH levels and likely reflects a different disease process (7). Each patient was then classified into one of two groups, those with HC and those without HC.
We defined HC as requiring hospitalization for hypercalcemia prior to undergoing a parathyroidectomy. We gathered information on medications prior to admission (or surgery for those not admitted in HC) with a particular emphasis on those that impact serum calcium level. These medications included bisphosphonates, loop diuretics, thiazide diuretics, calcimimetics, vitamin D, multivitamins, calcium supplements, and hormonal therapies, such as estrogen (8). Another variable we analyzed was the patients’ Charlson Comorbidity Index (CCI), which provided an appraisal of patient comorbidities (9).
We classified the types of parathyroidectomy that the patients underwent as a minimally invasive procedure, minimally invasive procedure–converted to open, or open four-gland exploration. Pathology was categorized as adenoma or hyperplasia, atypical adenoma, or cancer by a pathologist. We extracted these data from the pathology report. Our pathologists use the term “atypical adenoma” to describe adenomas with features of malignancies, such as banding fibrosis or mitotic activity, but lacking unequivocal evidence of invasive growth, as seen in carcinomas (10).
We performed a univariable analysis to compare patients with HC to those without HC by analyzing categorical variables with chi-square tests, and continuous variables with Mann-Whitney U tests, due to lack of normality. We then performed forward stepwise multivariable logistic regression to identify predictors of hypercalcemic crisis. Patient age, gender, and all variables with P<0.10 in the univariable analysis served as candidate predictors for the multivariable analysis. We used a Classification and Regression Tree (CART) analysis to create a decision-making tool to classify patients into groups based on the risk of a patient developing HC. CART combines recursive partitioning and regression methods to classify patients according to the outcome of interest (11). The CART analysis algorithm goes through each value for the variable to determine the cutoff point that does the best job of differentiating patients with and without crisis. The cutoffs for each variable are computer generated, taking into account the relationships between the variables at the different levels (or branch points) of the classification tree. For this reason, some variables that are non-significant in logistic regression become significant at later branch points in the CART analysis (eg. Charlson comorbidity index). The decision tree model was ten-fold cross-validated. All analyses except for the CART analysis were performed with SPSS 23 (IBM, Chicago, IL). The CART analysis was performed with R 3.3.1 “rpart” (Vienna, Austria).
From 11/2000–03/2016 there were 2,344 initial parathyroidectomies performed for sporadic PHPT in adults. The final study cohort included 183 patients with a serum calcium greater than or equal to 12 mg/dL. There were 29 (15.8%) patients with HC and 154 (84.2%) patients without HC. Of HC patients with records available before admission, 72% (13/18) had heathcare encounters prior to experiencing HC where PHPT could potentially have been recognized and treated. Patients with and without HC were similar in age, gender, CCI, Body Mass Index, and smoking status (Table 1). There were significant differences in several of the pre-operative lab values between the two groups of patients. As expected, patients with HC had a higher median calcium level (13.8 mg/dL vs. 12.4 mg/dL, P<0.001), a higher median PTH level (318 pg/mL vs. 160 pg/mL, P=0.001), and a lower median total vitamin D level (16 ng/mL vs. 26 ng/ml, P<0.001) than patients without crisis (Table 1). In addition, a history of kidney stones was more frequently observed in patients with crisis than it was in patients without crisis (31.0% vs. 14.3%, P=0.039).
Patients in both groups were taking similar calcium-altering medications. Hormonal therapies were the only category of these medications that showed a significant difference between the two groups. Patients who developed crisis were less likely to be taking hormonal therapies than those patients who did not develop crisis (0.0% vs. 9.1%, P=0.025, Table 2). It is also of note that vitamin D supplements were not associated with HC (Crisis 20.7% vs. Non-Crisis 18.8%, P=0.817).
The most common symptoms in patients during admission for HC were related to gastrointestinal problems. These symptoms, which included vomiting, diarrhea, and abdominal pain were present in 14 of 29 (48%) patients, while 12 of 29 (41%) patients presented with altered mental status. Other common symptoms included fatigue (12/29 or 41%) weakness (12/29 or 41%), and syncope or near-syncope (4/29 or 14%).
Nearly twice the number of patients with HC had multiple parathyroid glands resected during their parathyroidectomy compared to the patients who did not have HC although this difference was not statistically significant (24.1% vs. 12.3%, P=0.12). Cure rates were high and similar between groups (Crisis 96.6% vs. Non-Crisis 95.5%, P=0.79). Interestingly, parathyroid cysts were more common in patients with HC than in patients without HC (6.9% vs. 0.7%, P=0.05). We also found no difference between these two groups in frequency of atypical adenoma (3.6% vs. 2.7%, P=0.81) or cancer (0.0% vs. 1.4%, P=0.40).
Using a forward stepwise approach to logistic regression, we constructed a model of important factors associated with HC from age, gender, and all variables with a univariable p value of .1 or less. The final model indicated that preoperative calcium (OR 1.6 per 1 mg/dL increase, 95% CI 1.1 – 2.5, P=0.01), PTH (OR 1.003 per 1 pg/mL increase, 95% CI 1.001 – 1.006, P=0.01), and a history of kidney stones (OR 3.0, 95% CI 1.1 – 8.2, P=0.04) were independently associated with HC while controlling for effects of other predictors in the model.
The CART analysis identified the three most important factors in patient classification: calcium, CCI, and PTH. Based on patient characteristics associated with these three factors, the CART analysis classified patients into the following four groups with different risk of developing HC: (1) having calcium level between 12 and 13.25 mg/dL (with a 6.4% chance of developing HC); (2) having calcium level no less than 13.25 mg/dL and a CCI score of 4 or greater (90.9% risk); (3) having calcium level no less than 13.25 mg/dL, a CCI score less than 4, and PTH level less than 394 pg/mL (19.0% risk), and (4) having calcium level no less than 13.25 mg/dL, a CCI score less than 4, and PTH level no less than 394 pg/mL (60.0% risk). The ten-fold cross-validation yielded a prediction accuracy of 0.87 for the decision tree model (Figure 1).
In this study, we identify factors that are associated with a higher risk of developing HC in patients with PHPT. Specifically, both marked elevation in pre-operative calcium and PTH levels and a personal history of kidney stones appear to make HC more likely. We also provide a decision-making tool that can aid clinicians in classifying a patient into one of the four groups with different likelihoods of developing HC.
The current study is unique in its definition of HC. Most previous studies use an arbitrary plasma calcium level to define HC (2, 6, 12–15). Our study took a different approach, allowing us to evaluate the association of plasma calcium levels and the risk of HC. We found that as a patient’s plasma calcium level increases, the patient’s likelihood of requiring inpatient management for HC increases as well. While other investigations have also reported that higher calcium levels correlate with HC, the significance of their finding is unclear since HC was defined by the serum calcium levels (2). As a result, not all patients defined as having HC in these studies actually experienced clinically significant hypercalcemia. Additionally, Bargren and colleagues have previously shown that calcium levels do not always correlate with symptoms in patients with hyperparathyroidism (16). Therefore, defining HC with clinical parameters rather than serum calcium levels provides a more accurate analysis of factors associated with HC.
The only medications we found that were associated with a lower risk of HC were hormonal therapies, including estrogen/progesterone formulations and selective estrogen receptor modulators. Given the other effects these medications have, it is impractical to think they would have any utility in bridging patients with markedly elevated calcium to parathyroidectomy. Furthermore, Schneider et al. reported that these and other calcium lowering medications may place patients at higher risk for post-operative hypocalcemia (8).
Interestingly, other calcium-altering medications displayed little association with the risk of HC. A prior study from our group has shown that HC was more common in patients with a vitamin D deficiency and recommended avoiding vitamin D supplements in hypercalcemic patients for fear of exacerbating hypercalcemia (17). Other investigators also recommend against vitamin D supplementation in patients with PHPT to avoid exacerbating the hyperalecmia (18). While the current study was unable to show an association between vitamin D supplementation and the development of HC, we may be underpowered to do so. Additonally, in patients with markedly elevated calcium and clear evidence of PHPT, expedited parathyroidectomy is more appropriate than taking time to replete vitamin D so that HC can be avoided even if vitamin D supplementation is safe in these patients. Although patients with preoperative vitamin D deficiency are at theoretically increased risk for post-operative hypocalcemia, prior large, retrospective reviews have shown no association between vitamin D deficiency and post-operative hypocalcemia (17, 19). While Phityakorn and colleagues have reported on the use of bisphosphonates to bridge patients from HC to parathyroidectomy (14), we did not find an association between bisphosphonate use and the risk of HC, suggesting that bisphosphonates do not play a role in preventing crisis. Again, we may be underpowered to detect an association between their use and a lower risk of HC, we generally avoid bisphosphonates in crisis due to concerns for post-operative hypocalcemia.
It is unclear why patients with a history of kidney stones seem more likely to develop HC. It is possible that hydration plays a role and that patients who are more prone to dehydration are more likely to have concentrated urine and acute rises in serum calcium. In general, adequate hydration is thought to prevent both nephrolithiasis and HC. Lotan and colleagues were able to show that the incidence of nephrolithiasis is greater in a population at risk of dehydration (20). Vomiting and diarrhea were common symptoms in our cohort of patients with HC. While our study was not designed to determine if these symptoms led to or were caused by HC, it is possible that gastrointestinal losses resulted in dehydration which further exacerbated hypercalcemia.
Although higher calcium levels are generally associated with single adenomas (21), and the majority of patients with HC have single adenomas (14), we found that almost a quarter of patients with HC actually had multi-gland disease. Although not statistically significant, the patients with high calcium levels who did not experience HC were about half as likely to require the resection of more than one gland. Thus, the cohort of patients with HC was similar to the general population of patients with PHPT, though conventional wisdom would suggest that HC is likely to be caused by single adenoma. This finding is in contrast to the series reported by Beck and colleagues where there were equivalent rates of multi-gland disease in crisis and non-crisis patients (12). However, for an accurate comparison, it is important to note that while our study only included patients with calcium levels of 12 mg/dL or higher, their study included all patients being treated for hyperparathyroidism. Also, their definition of crisis included anyone with a calcium level of at least 14 mg/dL. These differences in study design make direct comparisons challenging. Similar to Beck and colleagues, cure rates for those with and without HC were comparable indicating that despite variation in the proportion of multi-gland disease, surgery remains an effective treatment for these patients (12). Nevertheless, in the absence of the expected decrease in intraoperative PTH during a minimally invasive approach, we encourage a low threshold for four-gland exploration in patients with HC, given the proven severity of the symptoms and the importance of providing a cure with the first surgery.
Although the treatment for HC, namely hydration followed by surgery, is effective, there is a paucity of literature aimed at preventing HC in patients with PHPT. Providers can use the decision-making tool developed in this study to determine the likelihood of a patient with PHPT developing crisis based on that patient’s pre-operative calcium, CCI score, and PTH levels. In the case of most patients with PHPT who are unlikely to develop HC, they can wait months to have a parathyroidectomy. However, a provider can use this tool to determine if a patient has a high likelihood of developing HC, and expedite parathyroidectomy, scheduling it within a shorter period of time.
While pre-operative calcium above 13.25 mg/dL was determined to be the most important branch point in this study. If the calcium is greater than this value, a CCI score greater than or equal to four can be used as a second branch point to classify the patient. One possible explanation for the CCI being the next branch point is that if a patient has many comorbidities, it may be more difficult for that patient to compensate for a markedly elevated calcium level. Meanwhile, compensation is less of an issue if the patient has a lower calcium.
The current study does have some limitations. While a relatively large cohort of patients with HC were included, the rare nature of this manifestation of PHPT makes a small sample size unavoidable and renders our analyses at risk of a type II error. For instance, the rate of multi-gland disease was twice as high in patients with HC, but this difference was not statistically significant. While most of the data was collected prospectively and in a consistent manner, some symptoms of HC were collected retrospectively from hospital documentation and may be subject to the quality and completeness of inpatient documentation. Additionally, if a patient’s presentation for treatment of HC was the initial documentation of PHPT, there may not have been laboratory values that could be analyzed from prior to the admission. Similarly, because much of this data was collected retrospectively and many patients were refered from other healthcare systems, we do not know how long they had markedly elevated calcium and were at risk of developing HC. Lastly, while our definition of HC focuses more on symptoms of hypercalcemia that require inpatient treatment and likely represents a more clinically relevant measure of HC, it is different from other publications looking at HC and may make our results difficult to compare to other literature.
In conclusion, patients with PHPT and calcium levels greater than or equal to 13.25 mg/dL and a history of kidney stones are at risk for developing HC. Once calcium levels reach 12 or greater, the providers can utilize the decision-making tool developed in this study to determine which patients are at greatest risk for HC and expedite parathyroidectomy appropriately.
Research reported in this publication was supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health under award number T35DK062709. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Presented in February 2017 at the 12th Annual Academic Surgical Congress in Las Vegas, Nevada
Author Contributions: Conception/ design- Lowell, Schneider, Randle. Acquisition of data- all authors. Analysis/ data interpretation- Lowell, Schneider, Wang, Ma, Randle. Drafting- Lowell, Randle. Revision- all authors. Approval/ accountability- all authors.
Disclosures: No authors have anything to disclose