Ghaheri, Bobak A. MD; Liebler, Sarah L. MS, PA-C; Andersen, Peter E. MD; Schuff, Kathryn G. MD; Samuels, Mary H. MD; Klein, Robert F. MD; Cohen, James I. MD, PhD

Objective: Perioperative hypocalcemia from temporary parathyroid gland dysfunction is common after thyroid surgery. No reliable cutoff values for parathyroid hormone (PTH) and the subsequent possibility of developing hypocalcemia exist. The purpose of this study is to determine a criterion for predicting hypocalcemia based on different PTH levels as cutoff values.

Study Design: Retrospective chart review.

Methods: A centralized database of intraoperative PTH levels was analyzed. PTH values approximately 10 minutes after excision of the thyroid gland and in the recovery room were obtained; serial ionized calcium levels were also analyzed. PTH values were then compared using chi-square analysis with significance defined as P < .05. A receiver operator characteristic (ROC) curve was also constructed to define sensitivities and specificities of different PTH levels as potential cutoff values.

Results: Eighty patients were identified meeting the study criteria between January 1999 and February 2005. Fourteen of the 80 (17.5%) patients became hypocalcemic during the hospital stay; none experienced permanent hypocalcemia. Patients who became hypocalcemic during their hospitalization were more likely to have a PTH level below 15 pg/mL (P < .01). Patients with a PTH level less than 15 pg/mL were more likely to develop hypocalcemia (P < .01). Finally, an ROC curve was constructed, allowing the surgeon to determine acceptable sensitivities and specificities and various PTH cutoff values.

Conclusion: Low perioperative PTH levels significantly correlate with the presence of postoperative hypocalcemia but cannot be used to predict it. Using the ROC curve allows different chosen cutoff values to predict hypocalcemia with varying sensitivity and specificity.

Postoperative hypocalcemia is one of the most common complications of thyroid surgeries that place overall parathyroid gland function at risk (total thyroidectomy, completion thyroidectomy, or bilateral paratracheal lymph node dissection [PTLND]). Published reports have put the risk of temporary hypocalcemia as high as 30%.1 To monitor for this potentially serious complication, the current standard of care for these patients is admission to the hospital postoperatively for measurement of serial calcium levels and subsequent calcium and vitamin D supplementation if hypocalcemia develops.

Within the last 10 years, parathyroid surgery has benefited from the development of a rapid parathyroid hormone (PTH) assay. This assay has traditionally been used in the setting of parathyroid adenoma resection, where an intraoperative 50% to 60% drop in PTH levels when comparing postexcision with preoperative levels defines a successful resection.2 It has allowed the traditional four gland, and even the unilateral exploration, to be supplanted by more minimally invasive procedures. However, the immediate perioperative measurement of PTH has been less widely used for assessing parathyroid function at the time of thyroid surgery even though published reports have shown statistical significance when correlating a single postoperative PTH value to the presence or absence of hypocalcemia.2,3

The drawback of these studies is that none have defined clinically useful criteria that can be translated into practice guidelines. Statistically correlating low postoperative PTH values with the presence of hypocalcemia does not give an absolute risk of developing this complication. These correlations only demonstrate that patients who are normocalcemic have statistically different PTH values than those who become hypocalcemic. The purpose of this study is to define the sensitivity and false-positive rates (FPR) of using PTH levels in predicting postoperative hypocalcemia and to develop criteria that help define the absolute risk of hypocalcemia at different PTH values.

This retrospective study was approved by the institution’s review board. A centralized database of patients who have had intraoperative PTH assays performed was analyzed. Data were tabulated for all patients undergoing thyroid-related surgery believed to place total parathyroid function at risk: total thyroidectomy, completion thyroidectomy, and postthyroidectomy PTLND. Patients who had only undergone unilateral surgery were excluded from the study.

Patients undergoing surgeries included in the study had an intraoperative PTH level drawn approximately 10 minutes after removal of the thyroid gland or paratracheal lymph nodes. Another PTH level was drawn in the recovery room. Patients who had only one PTH level drawn were also excluded from the analysis. These levels were then immediately analyzed by the General Clinical Research Center Core Laboratory, and values were entered in a central database. The Immulite turbo intact PTH assay (a chemiluminescent assay; Los Angeles, CA) was used to measure intraoperative PTH levels.

Postoperatively, ionized calcium levels were obtained every 6 hours for monitoring. Hypocalcemia was defined as an ionized calcium level (corrected for blood temperature and pH) below 1.00 mmol/L. Normocalcemic levels were defined as ionized calcium levels of 1.00 mmol/L or greater. If patients became hypocalcemic, intravenous or oral calcium supplementation was started.

Statistical analysis was performed using [chi]² analysis on Microsoft Excel (Redmond, WA), and a statistical significance was assigned to a P value equal to or less than 0.05. A receiver operating characteristic (ROC) curve was constructed with the tabulated data using SPSS (Chicago, IL). Sensitivity was plotted against the FPR at each PTH value to construct the curve.

Ninety-one patients undergoing total thyroidectomy, completion thyroidectomy, or PTLND were identified between January 1999 and February 2005. Eleven patients were excluded because only one PTH value was drawn, leaving 80 patients for inclusion in the study. Fourteen (17.5%) patients experienced temporary hypocalcemia; all 14 patients have subsequently regained normal parathyroid function. None of the 80 patients had preexisting hypoparathyroidism or hypocalcemia.

There were 15 male patients and 65 female patients. The indications for surgery are displayed in Table I. Twenty-two (27.5%) patients were undergoing completion thyroid surgery or postthyroidectomy PTLND. Data for time point 1 (approximately 10 min postexcision) and time point 2 (in the recovery room) were analyzed separately. For time point 1, the mean PTH value for the 66 normocalcemic patients was 70.8 pg/mL. This was compared with the same group in time point 2, where the mean PTH level was 48.0 pg/mL. For the 14 hypocalcemic patients, mean PTH values for time points 1 and 2 were 31.3 pg/mL and 16.0 pg/mL, respectively.

An ROC curve was constructed using the raw data from the patient population. The curve was constructed in a nonparametric fashion, plotting the FPR (x-axis) against sensitivity (y-axis) to test the ability of intraoperative PTH levels in predicting postoperative hypocalcemia. This ROC curve allows the user to choose any point on the curve of PTH values and determine the corresponding sensitivity and FPR (1 - specificity).4 The curve was constructed for both time points, and the area under the curve (AUC) was determined. With use of the ROC curve, an AUC of 1.0 indicates a perfect test, with 100% sensitivity and a 0% FPR. The AUC for time point 1 was 0.76, whereas the AUC for time point 2 was 0.83, indicating that the PTH level drawn in the recovery room is the more accurate test. The two ROC curves are shown in Figure 1. With use of the ROC curve, representative PTH values were selected, and the associated sensitivity and FPR were calculated to demonstrate the variation in the statistical measures with different PTH level cutoffs (Table II).

With use of the ROC curve, it becomes evident that a PTH cutoff value of 58 pg/mL correlates with 100% sensitivity but a FPR of 65% (specificity of 35%). Similarly, a PTH value of 39 pg/mL has a 93% sensitivity and a 53% specificity. The PTH values and their corresponding sensitivity and specificity shown in Table II are derived from the data generated by the ROC curve. Each decrease in sensitivity comes from the PTH level encountered when each of the 14 hypocalcemic patients developed hypocalcemia. Table III demonstrates the differences between the intraoperative and recovery room PTH values among the 14 hypocalcemic patients.

In our own preliminary report, we found 15 pg/mL to be the statistically significant cutoff value for predicting hypocalcemia and that an increase in PTH levels between the postresection and recovery room value predicted normocalcemia.3 In the current study, patients with a recovery room PTH level of less than 15 pg/mL were again significantly more likely to experience hypocalcemia (P < .05). Conversely, patients who experienced hypocalcemia were significantly more likely to have a PTH level less than 15 pg/mL (P < .05). However, a postoperative PTH level less than 15 pg/mL did not have high predictive value for hypocalcemia in the current analysis. In addition, in this study, the trend between the intraoperative PTH level and the PTH level drawn in the recovery room did not correlate with predicting postoperative hypocalcemia.

Previous studies of this type have tried to define a specific PTH level that would predict postoperative hypocalcemia. Lombardi and colleagues 5 analyzed 53 total thyroidectomy patients who had serial PTH levels drawn in addition to postoperative calcium levels. They found that a PTH level of less than 10 pg/mL at 4 or 6 hours after surgery identified all but one patient who suffered hypocalcemia. Quiros et al.6 also used 10 pg/mL as a cutoff value; in their study, patients who were below this level postoperatively were significantly more likely to be on vitamin D supplementation at 1 month postthyroidectomy. Payne and colleagues 7 used a combination of PTH values (<28 pg/mL) and corrected calcium levels (<8.56 mg/dL) to define a critical level at 6 hours postthyroidectomy. A further study by Lam and Kerr 8 where hypocalcemia was defined as less than 0.9 mmol/L demonstrated a critical PTH level of 8 pg/mL to accurately separate hypocalcemic from normocalcemic patients. Finally, Richards et al.9 also used a PTH level of 10 pg/mL to demonstrate which patients actually became symptomatic of hypocalcemia. Of 10 patients in that study who experienced symptoms of hypocalcemia, 5 experienced tetany. The current study also demonstrated statistical significance when comparing PTH levels and the development of hypocalcemia. At 15 pg/mL or lower, more patients became hypocalcemic when compared with patients whose postoperative PTH levels were greater than 15 pg/mL.

Despite the determination of statistical significance for certain values, these studies do not establish the clinical usefulness of the rapid PTH value as an intraoperative tool. Although a low PTH value may statistically correlate with a risk of hypocalcemia, that risk is not absolute. Similarly, normocalcemia is not guaranteed by a value that is higher than these published cutoff levels. Although the presence of postoperative hypocalcemia can significantly correlate with lower PTH values, the presence of a low calcium level postoperatively is not always associated with a low PTH value. To better use the intraoperative PTH value, there must first exist criteria that define the sensitivity and specificity of the test at different PTH cutoff values.

The ROC curve has typically been used to measure the ability of a test, such as a computed tomography scan, to accurately predict the presence or absence of disease, such as a malignancy.4 The current study has used the ROC curve to test the ability of an intraoperative PTH value (the test) to accurately predict the presence of disease (hypoparathyroidism and subsequent hypocalcemia). The ROC curve allows users to identify a PTH value and determine what the corresponding sensitivity and FPR are. Examples in Table II help to illustrate this point. For instance, with use of the data from this study, to insure a patient was not at risk for hypocalcemia, the PTH value would need to be greater than 58 pg/mL. At this cutoff value, the sensitivity is 100%, but the FPR (calculated as 1 - specificity) at this level is quite high at 65%, meaning 65% of patients under 58 pg/mL also remain normocalcemic. One of the primary goals of using the intraoperative PTH level is to be able to safely and reliably send thyroidectomy patients home the day of surgery without the need for hospital admission and serial calcium levels because this would allow for a better allocation of resources in the setting of an increasingly costly health care environment. The most important statistical tool for making this decision is the sensitivity of the PTH cutoff value, which measures the ability of the perioperative PTH value to show the presence of hypocalcemia. Using the data from this study, a conservative surgeon would use a PTH cutoff value of 58 pg/mL. However, the data also allow the surgeon to understand that 65% of the patient admissions using these criteria will be unnecessary because those patients will remain normocalcemic.

The ROC curve constructed from the two time points in this study also demonstrated that the timing of the intraoperative PTH level is of significance. The PTH value drawn 10 minutes postexcision was less accurate than the level drawn in the recovery room. Perhaps manipulation of the surgical bed may artificially elevate serum PTH levels and mask a true underlying hypoparathyroidism. The data set obtained from the patients does include patients who have undergone completion thyroidectomy; two of these patients were among the 14 who ultimately experienced temporary postoperative hypocalcemia. Inclusion of these patients is clinically relevant because the functional status of the parathyroid glands from the previous operation is unknown and reflects the reality of referral and reoperation.

Scurry et al.10 recently published a study with methodology similar to this one. They identified two separate tools to help predict postoperative hypocalcemia. Their data showed a cutoff value of 7 pg/mL as a useful measure to predict hypocalcemic patients because it maximized the AUC for their particular ROC curve. Unfortunately, maximization of the AUC is determined by the specificity in addition to the sensitivity for that particular PTH level. We maintain that the specificity of the PTH cutoff value is a clinically irrelevant statistic and that only the sensitivity should be analyzed to determine who can safely be discharged. Similarly, they find that a 75% drop in PTH level when comparing preoperative levels with those drawn 10 minutes after thyroidectomy also maximizes the AUC for that particular test. Again, the study overemphasizes the contribution of the specificity in forming the ROC curve. The best criterion for using PTH levels for predictive purposes should focus on the sensitivity alone. If the test can detect all patients who will ultimately become hypocalcemic (sensitivity), then none will be inadvertently discharged home. This will obviously lead to more unnecessary hospital admissions (higher FPR), but the nature of this disease demands conservatism.

Perioperative PTH levels, especially those drawn in the recovery room, can be used in a clinically relevant way to predict the relative risk of postoperative hypocalcemia after thyroid-related surgeries that place total parathyroid function at risk.

The authors thank Kristi Buxton, Dennis Trune, and Susan Griest for their assistance with data compilation and statistical analysis.

1. Marohn MR, La Civita KA. Evaluation of total/near-total thyroidectomyin a short stay hospitalization: safe and cost effective. Surgery 1995;118:943–947. ExternalResolverBasic Bibliographic Links [Context Link]

2. Warren FM, Andersen PE, Wax MK, et al. Intraoperative parathyroid hormone levels in thyroid and parathyroid surgery. Laryngoscope 2002;112:1866–1870. Ovid Full Text ExternalResolverBasic Bibliographic Links [Context Link]

3. Warren FM, Andersen PE, Wax MK, et al. Perioperative parathyroid hormone levels in thyroid surgery: preliminary report. Laryngoscope 2004;114:689–693. Ovid Full Text ExternalResolverBasic Bibliographic Links [Context Link]

4. Park SH, Goo JM, Jo CH. Receiver operating characteristic (ROC) curve: practical review for radiologists. Korean J Radiol 2004;5:11–18. ExternalResolverBasic Bibliographic Links [Context Link]

5. Lombardi CP, Raffaelli M, Princi P, et al. Early prediction of postthyroidectomy hypocalcemia by one single iPTH measurement. Surgery 2004;136:1236–1240. ExternalResolverBasic Full Text Bibliographic Links [Context Link]

6. Quiros RM, Pesce CE, Wilhelm SM, et al. Intraoperative parathyroid hormone levels in thyroid surgery are predictive of postoperative hypoparathyroidism and need for vitamin D supplementation. Am J Surg 2005;189:306–309. ExternalResolverBasic Full Text Bibliographic Links [Context Link]

7. Payne RJ, Hier MP, Tamilia M, et al. Same-day discharge after total thyroidectomy: the value of 6-hour serum parathyroid hormone and calcium levels. Head Neck 2005;27:1–7. ExternalResolverBasic Full Text Bibliographic Links [Context Link]

8. Lam A, Kerr PD. Parathyroid hormone: an early predictor of postthyroidectomy hypocalcemia. Laryngoscope 2003;113:2196–2200. Ovid Full Text ExternalResolverBasic Bibliographic Links [Context Link]

9. Richards ML, Bingener-Casey J, Pierce D, et al. Intraoperative parathyroid hormone assay: an accurate predictor of symptomatic hypocalcemia following thyroidectomy. Arch Surg 2003;138:632–636. Ovid Full Text ExternalResolverBasic Full Text Bibliographic Links [Context Link]

10. Scurry WC, Beus KS, Hollenbeak CS, et al. Peroperative parathyroid hormone assay for diagnosis and management of postthyroidectomy hypocalcemia. Laryngoscope 2005;115:1362–1366. Ovid Full Text ExternalResolverBasic Bibliographic Links [Context Link]

Key Words: Hypocalcemia; parathyroid hormone; perioperative; thyroid surgery