- Email: [email protected]
Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism
The American Journal of Surgery xxx (2017) 1e6
Contents lists available at ScienceDirect
The American Journal of Surgery journal homepage: www.americanjournalofsurgery.com
Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism David C. Cron, Steven R. Kapeles, Elizabeth A. Andraska, Sebastian T. Kwon, Peter S. Kirk, Brendan L. McNeish, Christopher S. Lee, David T. Hughes* Department of Surgery, University of Michigan Medical School, 1500 E Medical Center Dr, Ann Arbor, MI 48109, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 2 August 2016 Accepted 8 January 2017
Introduction: Many adjuncts guide surgical decision making in parathyroidectomy, yet their independent associations with outcome are poorly characterized. We examined a broad range of perioperative factors and used multivariate techniques to identify independent predictors of operative failure (persistent disease) after parathyroidectomy. Methods: This was a retrospective review of 2239 patients with primary hyperparathyroidism who underwent parathyroidectomy at a single-center from 1999 to 2014. We used multivariate logistic regress to measure associations between multiple perioperative factors and an operative failure (persistent hypercalcemia). Results: Operative failure was identified in 67 patients (3.0%). The following variables were independently associated with operative failure on multivariate analysis: IOPTH criteria met (protective, OR ¼ 0.22, P < 0.001), preoperative calcium (risk factor, OR ¼ 2.27 per unit increase, P < 0.001), weight of excised gland(s) (protective, OR ¼ 0.70 per two-fold increase, P ¼ 0.003), and preoperative PTH (protective, OR ¼ 0.55 per two-fold increase, P ¼ 0.008). Conclusion: In addition to the well-established IOPTH criteria, we suggest that consideration of the above independent perioperative risk factors may further inform surgical decision-making in parathyroidectomy. Short summary: We studied the relationship between perioperative factors and operative failure (persistent hypercalcemia) after parathyroidectomy. In addition to the standard intraoperative parathyroid hormone criteria, the following were independent risk factors for operative failure on multivariate analysis: higher preoperative calcium, lower preoperative parathyroid hormone, and lower weight of excised glands. These perioperative factors may inform surgical decision making for parathyroidectomy. © 2017 Elsevier Inc. All rights reserved.
Keywords: Primary hyperparathyroidism Persistent hypercalcemia Parathyroidectomy Operative failure
1. Introduction Primary hyperparathyroidism (PHPT) has an incidence of 1e2% in the adult population, and parathyroidectomy is the only cure.1 Surgical decision-making in parathyroidectomy is guided largely
* Corresponding author. Department of Surgery, University of Michigan Health Systems, 1500 E. Medical Center Drive, 2920 Taubman Center, SPC 5331, Ann Arbor, MI 48109-5331, USA. E-mail addresses: [email protected] (D.C. Cron), [email protected] (S.R. Kapeles), [email protected] (E.A. Andraska), [email protected] (S.T. Kwon), [email protected] (P.S. Kirk), [email protected] (B.L. McNeish), [email protected] (C.S. Lee), [email protected] (D.T. Hughes).
by preoperative imaging and intraoperative parathyroid hormone (IOPTH) monitoring. Preoperative imaging can include sestamibi ± single-photon emission computed tomography (SPECT), cervical ultrasound, and 4D-computed tomography scan.2,3 IOPTH monitoring is frequently used to ensure appropriate resection and biochemical cure.4 In experienced centers, success rates for parathyroidectomy for PHPT range from 92% to 99%.3,5e11 Nevertheless, failure does occur after this common surgery, yet failure is infrequently studied. Little is known about patient-level predictors of operative failure and persistent PHPT. Previous studies have attributed operative failure to inadequate preoperative imaging localization.3,5,10 Achievement of IOPTH criteria is a known predictor of operative success, though the final target IOPTH level is not agreed upon.
http://dx.doi.org/10.1016/j.amjsurg.2017.01.012 0002-9610/© 2017 Elsevier Inc. All rights reserved.
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
2
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
Some researchers contend that final IOPTH levels should fall into the normal range,4,9 while others recommend lower levels.6 The independent contributions of preoperative localization, IOPTH biochemical cure, and preoperative biochemical severity to operative success are unclear. Better understanding of the relationship of these factors to operative failure is necessary to further improve outcomes after this surgery and optimize intraoperative decisionmaking. We investigated predictors of operative failure (persistent disease) in patients undergoing parathyroidectomy for PHPT. Using multivariate logistic regression, we evaluated the independent effects of the following factors: demographics, preoperative biochemical disease severity, preoperative imaging, IOPTH criteria, and histopathologic findings. We aimed to identify a subset of independent risk factors for operative failure.
criteria was not achieved, localization was incorrect, or multiglandular disease was otherwise suspected. The variable of interest in analysis was achievement of IOPTH criteria (biochemical cure). In rare cases, IOPTH criteria was not achieved prior to completion of an operation. For example, IOPTH monitoring was sometimes ceased upon conversion to four-gland operation in lieu of morphologic analysis of enlarged hyperplastic parathyroid glands. Additionally, patients with low baseline PTH (<100 pg/mL) may have final IOPTH levels that fall into normal range yet do not meet criteria of decreasing 50% from baseline. The number of parathyroid glands excised was dichotomized to single gland or multiple glands for the analysis. We did not distinguish between hyperplasia and adenoma based on pathological analysis, as this is generally considered a difficult distinction. Total weight of the excised gland(s) was recorded from the pathology report.
2. Methods 2.3. Outcome 2.1. Study population This was a retrospective analysis of a prospectively maintained, Institutional Review Board approved database containing patients who underwent parathyroidectomy for PHPT at the University of Michigan Health System between 1999 and 2014. We excluded patients with the following: secondary or tertiary hyperparathyroidism, reoperative parathyroidectomy, multiple endocrine neoplasia type 1, and history of lithium use. 2.2. Preoperative factors Preoperative laboratory measurements included peak serum calcium and serum PTH levels (biochemical severity of disease). Nearly all study patients underwent preoperative imaging for gland localization, including surgeon performed ultrasound and/or sestamibi scan ± SPECT. At our institution, ultrasound is the primary imaging modality and is performed by the endocrine surgeon during initial clinical evaluation. Sestamibi is ordered if ultrasound is non-localizing or equivocal. For logistic regression analysis, the imaging localization variable was coded as: neither imaging modalities localizing or no imaging performed (the reference group), ultrasound or sestamibi localizing, ultrasound and sestamibi localizing but discordant, or ultrasound and sestamibi localizing and concordant. Imaging was considered correct if the diseased glands were located intraoperatively on the same side as indicated by imaging (without distinguishing superior/inferior). If imaging suggested an adenoma but multiglandular disease was discovered intraoperatively, this was considered incorrect imaging. The imaging correctness variable was coded as: neither imaging modalities correct or no imaging performed (the reference group), ultrasound or sestamibi correct, or both ultrasound and sestamibi correct. Patients with positive localizing studies were candidates for minimally invasive parathyroidectomy (MIP). A bilateral exploration was elected in absence of localization or for high preoperative suspicion of multiglandular disease. For logistic regression analysis, the variable of interest was four-gland operation (either planned or converted). IOPTH monitoring was used in nearly every case. Levels were drawn from the jugular vein or peripherally at initiation of surgery and/or just prior to parathyroid gland excision and at 5, 10, and 15 min post-excision. Baseline IOPTH was defined as the highest value of two samples recorded following anesthesia induction and immediately prior to gland excision. Biochemical cure was defined as both a decrease in IOPTH 50% from baseline and a post-excision IOPTH within the normal range (12e75 pg/mL). MIP was converted to four-gland exploration intraoperatively if IOPTH
Postoperative serum calcium values were collected on all patients. The lowest and highest postoperative calcium levels were recorded. At the time of analysis, for patients with peak postoperative serum calcium levels 10.2 mg/dL, charts were retrospectively reviewed for additional follow-up to rule out transient postoperative hypercalcemia, and to verify outcome. The outcome of interest was operative failure with resultant persistent hyperparathyroidism. Persistent disease was defined by postoperative serum calcium measurements persistently 10.2 mg/dL or a single calcium 11.0 mg/dL when no additional data was available (all measurements within 6 months postoperatively). Postoperative PTH labs are not routinely ordered at our institution, and thus PTH was not a component of the definition of failure. Due to biochemical followup constraints, recurrent disease (incidence beyond 6 months postoperatively) was not assessed or analyzed. 2.4. Statistical analysis Pairwise comparisons were computed using a Student's t-test, Wilcoxon rank-sum test, or Fisher's exact test, as appropriate. The following variables were considered as potential predictors of operative failure: age, sex, preoperative peak PTH, preoperative peak calcium, imaging localization, imaging correctness, four-gland approach, achievement of IOPTH criteria, number of glands excised, and total weight of excised gland(s). Highly skewed variables were transformed using a logarithm of base two. Univariate and multivariate logistic regression were utilized to assess the associations between the perioperative factors and operative failure. For multivariate analysis, we randomly divided our sample into a training set (60% of sample) and an internal validation set (40% hold-out sample) to assess predictive ability of the models. Backward stepwise selection was used to identify independent predictors of operative failure (P 0.100 for inclusion in the model). Because our predictors of interest included both preoperative and intraoperative variables, we developed two main models: a preoperative model and a combined intraoperative model (the latter considered all variables). Hosmer-Lemeshow test was used to ensure model goodness of fit. Variance inflation factor was used to ensure no problems with multicollinearity. Area under the receiver operating characteristic curve (AUROC) was used to assess model predictive ability. We performed secondary analyses to evaluate consistency of our findings. First, we grouped preoperative calcium and PTH by tertile, and we examined rates of operative failure by tertile of calcium and PTH. As a secondary outcome, we assessed predictors of elevated postoperative calcium (calcium 10.2 mg/dL regardless
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
of persistence; binary outcome) using logistic regression. To determine if any statistical effects were mediated by the number of glands excised, we tested for interactions between the variable multiple glands excised and the following variables: preoperative peak calcium, preoperative peak PTH, and IOPTH criteria. Lastly, since IOPTH criteria is known to be strongly predictive of operative success, we performed a multivariate analysis excluding IOPTH criteria as a variable, to assess the effects of other predictors without accounting for IOPTH criteria. A significance level of P 0.05 was used for all significance tests. Statistical analyses were performed using Stata v 13.0 (College Station, TX). 3. Results 3.1. Study population 2239 patients met inclusion criteria. We identified 67 failures (3.0%). Of these failures, 14 (21%) had a reoperation at our institution at the time of the completion of this study's analysis. The median follow-up for postoperative calcium values (time between operation and last available calcium laboratory) was 18 days. Of the failures, the median follow-up for postoperative calcium values was 46 days. Case mix included 64.2% MIPs, 10.6% MIPs converted to four-gland explorations, and 25.3% planned four-gland explorations. 78% of our sample was female, and the mean age was 59 ± 13 years. Table 1 shows patient characteristics for all patients, successful cases, and operative failure cases. Preoperative PTH and weight of excised gland(s) were significantly skewed and were therefore log transformed for regression analyses. 3.2. Univariate logistic regression Fig. 1 shows the results of univariate logistic regression. Risk factors for operative failure included the following: multiple glands excised (Odds Ratio [OR] ¼ 2.96, 95% CI: 1.79e4.90, P < 0.001), fourgland exploration (OR ¼ 2.44, 95% CI: 1.48e4.02, P < 0.001), and higher preoperative peak calcium (OR ¼ 1.38 per unit increase, 95% CI: 1.03e1.83, P ¼ 0.028). Protectors against failure included IOPTH criteria met (OR ¼ 0.21, 95% CI 0.13e0.35, P < 0.001), larger weight of excised gland(s) (OR ¼ 0.70 per two-fold increase, 95% CI: 0.59e0.82, P < 0.001), ultrasound and sestamibi both correct (OR ¼ 0.23 compared to no correct imaging, 95% CI: 0.09e0.61, P ¼ 0.003), higher preoperative peak PTH (OR ¼ 0.66 per two-fold increase, 95% CI: 0.48e0.91, P ¼ 0.011), and ultrasound and sestamibi localizing and concordant (OR ¼ 0.31 compared to no
3
localizing imaging, 95% CI: 0.13e0.78, P ¼ 0.013). Factors not significantly associated with operative failure included age (P ¼ 0.118), ultrasound or sestamibi correct (P ¼ 0.179), ultrasound or sestamibi localizing (P ¼ 0.391), female sex (P ¼ 0.577), and ultrasound and sestamibi localizing but discordant (P ¼ 0.693). 3.3. Multivariate analysis: combined intraoperative model Fig. 2 shows the variables included in the combined intraoperative model. The multivariate model included the following independent predictors of operative failure: IOPTH criteria met (OR ¼ 0.22, 95% CI: 0.11e0.44, P < 0.001), preoperative peak calcium (OR ¼ 2.27, 95% CI: 1.46e3.53, P < 0.001), weight of excised gland(s) (OR ¼ 0.70 per two-fold increase in weight, 95% CI: 0.55e0.89, P ¼ 0.003), and preoperative peak PTH (OR ¼ 0.55 per two-fold increase, 95% CI: 0.36e0.86, P ¼ 0.008). The training and validation set AUROC values for this model were 0.79 (95% CI: 0.71e0.87) and 0.73 (95% CI: 0.62e0.84), respectively. 3.4. Multivariate analysis: preoperative model Two variables were included in the preoperative model: preoperative peak PTH (OR ¼ 0.47 per two-fold increase, 95% CI: 0.30e0.74, P ¼ 0.001) and preoperative peak calcium (OR ¼ 2.01 per unit increase, 95% CI: 1.31e3.09, P ¼ 0.001). The training and validation set AUROC values for this model were 0.71 (95% CI: 0.62e0.79) and 0.58 (95% CI: 0.46e0.70), respectively. 3.5. Secondary analyses To assess the consistency of these findings, we performed secondary analyses as described in the methods section. Fig. 3 shows rates of operative failure by tertile of preoperative calcium and PTH. Compared to the lowest calcium tertile, patients in the highest tertile had 3.4-fold higher rates of operative failure (P < 0.001). Compared to patients in the highest PTH tertile, patients in the lowest tertile had 2.2-fold higher rates of operative failure (P ¼ 0.005). We used multivariate logistic regression to identify predictors of elevated postoperative calcium (calcium 10.2 regardless of persistence). In multivariate logistic regression, the following factors were independent predictors of elevated postoperative calcium: preoperative peak calcium (OR ¼ 2.05, 95% CI: 1.66e2.54, P < 0.001), preoperative peak PTH (OR ¼ 0.71, 95% CI: 0.58e0.87, P ¼ 0.001), multiple glands excised (OR ¼ 1.77, 95% CI: 1.22e2.57, P ¼ 0.003), and IOPTH criteria met (OR ¼ 0.60, 95% CI: 0.40e0.89,
Table 1 Clinical Characteristics of the Study Cohort. The p-value compares success vs. failure. Continuous variables are represented as mean ± standard deviation or median (interquartile range) if skewed. Binary variables are represented as count (frequency). Calcium measurements are in mg/dL. PTH measurements are in pg/mL.
Total N Age Female Preoperative peak PTH Preoperative peak calcium Ultrasound performed Localizing Correct Sestamibi performed Localizing Correct Four-gland approach IOPTH criteria achieved Weight excised gland(s) (grams)
Total
Success
Failure
P-value
2239 59 ± 13 1742 (78%) 122 (81) 11.0 ± 0.7 1798 (84%) 1521 (85%) 1247 (82%) 1253 (56%) 856 (68%) 746 (87%) 765 (36%) 1887 (84%) 0.6 (0.7)
2172 59 ± 13 1688 (78%) 123 (82) 11.0 ± 0.7 1743 (84%) 1474 (85%) 1213 (82%) 1217 (59%) 842 (69%) 737 (88%) 728 (35%) 1850 (85%) 0.6 (0.7)
67 57 ± 13 54 (81%) 97 (69) 11.2 ± 0.6 55 (85%) 47 (85%) 34 (72%) 36 (55%) 14 (39%) 9 (64%) 37 (57%) 37 (55%) 0.4 (0.5)
0.117 0.656 0.003 0.029 1 1 0.085 0.611 <0.001 0.025 0.001 <0.001 <0.001
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
4
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
Fig. 1. Univariate logistic regression analysis of predictors of operative failure in parathyroidectomy. This figure displays the unadjusted odds ratios and 95% confidence intervals for all variables.
P ¼ 0.012). The training and validation set AUROC values for this model were 0.68 (95% CI: 0.63e0.71) and 0.62 (95% CI: 0.57e0.67), respectively. There were no significant statistical interactions between number of glands excised and calcium, PTH, or IOPTH criteria. This indicates that the effects of calcium, PTH, and IOPTH criteria on operative failure were independent of number of glands excised. Lastly, when we performed a multivariate analysis without including IOPTH criteria as a covariate, the following factors were independently predictive of operative failure: preoperative peak calcium (OR ¼ 2.47, 95% CI: 1.59e3.84, P < 0.001), preoperative peak PTH (OR ¼ 0.62, 95% CI: 0.40e0.94, P ¼ 0.025), multiple glands excised (OR ¼ 3.19, 95% CI: 1.59e6.43, P ¼ 0.001), and total weight of excised gland(s) (OR ¼ 0.70, 95% CI: 0.55e0.88, P ¼ 0.002).
4. Discussion This study used multivariate techniques to identify independent predictors of operative failure after parathyroidectomy for PHPT. The prevalence of failure in this cohort was 3%. In multivariate analysis, factors independently and significantly associated with increased risk of failure included lower preoperative PTH, higher preoperative calcium, failure to achieve IOPTH criteria of biochemical cure, and lower weight of excised gland(s). We reported similar findings for predictors of elevated postoperative calcium (regardless of persistence). When analyzed by tertile of preoperative calcium level, operative failure rates were highest in the upper tertile. Conversely, for preoperative PTH, operative failure rates were highest in the lower tertile. These results provide
Fig. 2. Multivariate logistic regression analysis of predictors of operative failure in parathyroidectomy. This figure displays the odds ratios and 95% confidence intervals for the 4 variables that were in the final combined intraoperative multivariate model.
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
Fig. 3. Rates of operative failure by tertile of preoperative calcium and PTH. This figure displays observed rates of operative failure (persistent hyperparathyroidism) by tertile of preoperative calcium (in mg/dL) and preoperative PTH (in pg/mL).
insight into the independent relationships between biochemical disease severity, commonly used perioperative adjuncts, and operative failure. This combination of factors may help inform surgical decision-making beyond use of IOPTH criteria alone. Our failure rate of 3% was similar to rates reported by other studies.3,5e11 An unexpected finding was the opposite effects of preoperative PTH compared to preoperative calcium. We expected that lower values of both preoperative PTH and preoperative calcium would be associated with higher rates of failure. We expected this because of the relation of low baseline PTH values to increased incidence of multiglandular parathyroid disease and the possible increased failure rate associated with this.12,13 In our study, lower PTH and higher calcium were risk factors for failure. The positive association between calcium and persistent disease could be due in part to the fact that persistent disease is defined based on calcium levels. However, this does not completely explain the direction or strength of the relationship, as the strong association was observed across the range of calcium values. It is possible that patients with higher preoperative calcium have a higher physiologic set-point for calcium homeostasis and are thus more likely to remain hypercalcemic postoperatively. It is also possible that higher calcium levels inhibit IOPTH levels and impair detection of additional hyperfunctioning tissue intraoperatively. Future studies are needed to explore these hypotheses and elucidate the differential effects of PTH and calcium on persistent disease. Imaging variables were predictive of failure in univariate analysis but not multivariate analysis. This does not imply that preoperative imaging is not an important adjunct; rather, it implies that the effect of imaging can be explained by other factors. Localizing imaging is inherently highly correlated with size and location of the enlarged glands, and scans that localize to one side favor MIP for the operative approach. Because weight of excised gland(s) was a strong predictor and present in the final model, it is not surprising that imaging variables and operative approach were excluded from the final multivariate model. Preoperative imaging is important to inform operative approach and to facilitate localization of the diseased glands. However, for the purpose of predicting operative success intraoperatively, our results suggest that the weight of excised parathyroid tissue, IOPTH criteria, and preoperative PTH and calcium are more important factors. Smaller weight of excised parathyroid gland(s) was a risk factor for failure. This is intuitive, as large glands are easier to locate and generally associated with single gland disease. Lastly, failure to meet IOPTH criteria was a strong predictor
5
of operative failure; however, our relatively stringent IOPTH criteria still yielded a significant amount of false positives, as 55% of failures met IOPTH criteria for biochemical cure. IOPTH data should be considered in the context of other patient specific perioperative factors. The limitations to this study include its retrospective study design. Also, this is a cohort from a single large academic institution's endocrine surgery division; therefore, our results may not be generalizable across all centers. Since our center is a tertiary care center, many of our patients had limited long-term biochemical follow-up within our healthcare system. Patients who were eucalcemic at their two-week postoperative appointment generally did not have additional calcium measurements ordered by the surgeon. Our reported failure rate may be an underestimate due to lack of follow-up; however, most cases of persistent disease have elevated calcium levels at their postoperative visit. Postoperative PTH was not a component of this definition; however, postoperative PTH levels are generally non-informative, as PTH elevation in the setting of eucalcemia is common postoperatively.14 Our failure rate could also be an overestimate if elevated calcium levels dropped after patients were lost to follow-up. Prior to analysis, we performed an additional extensive review of the electronic medical records of all patients with elevated postoperative calcium levels (10.2 mg/dL) in order to rule out hypercalcemia due to postoperative calcium supplementation. Lastly, our models showed moderate predictive ability, and we were limited to using an internal validation set. However, the predictive strength of the models was not a primary focus of this study. We were more interested in identifying independent predictors, and future work including more perioperative factors is needed to create robust predictive models. The strengths of this study include its large sample size, the use of multivariate techniques, and a focus on operative failure. In the context of operative failure, many studies focus on the diagnostic accuracy and predictive ability of individual adjuncts only. Multivariate techniques are important to identify the independent contributions of individual adjuncts. Our results emphasize this by showing that factors such as preoperative imaging can predict outcome in univariate analysis but can be explained by other variables (such as weight of enlarged glands). An important novel finding of this study is the correlation between preoperative biochemical severity of disease and outcome. Patients with high calcium and low PTH were at increased risk of failure independent of intraoperative factors like IOPTH. These risk factors should raise a surgeon's suspicion of multiglandular disease and failure risk, and this information may lower one's threshold for converting to a fourgland operation or continuing to monitor IOPTH beyond the standard criteria of biochemical cure. Predictive models can aid risk assessment and may provide objective metrics for intraoperative decision-making. The challenge is to make a prediction tool easy and efficient to use intraoperatively. Mazeh et al. recently developed a nomogram based on preoperative PTH and calcium to be used intraoperatively to predict likelihood of additional hyperfunctioning parathyroid tissue.15 Kebebew et al. developed a scoring system using multiple preoperative factors to distinguish single-gland vs. multiglandular cases.16 At this time, our model is too complex to be practically used intraoperatively, and these results serve as a proof of concept. Future work may extend these methods to an easy-to-use nomogram predicting failure risk based on preoperative biochemical severity, IOPTH levels, histopathologic findings, and more. Incorporating additional measurements such as renal function, 25-hydroxy vitamin D, and more granular IOPTH measurements may yield models with higher predictive ability than we show here. At the least, these results provide important insight into the independent contributions of patient-level
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
6
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
characteristics to outcome after parathyroidectomy. 5. Conclusion With this work, we show that multivariate techniques can be used to identify independent risk factors for operative failure after parathyroidectomy. We identified four independent predictors of persistent postoperative hypercalcemia: lower preoperative PTH, higher preoperative calcium, failure to meet IOPTH criteria, and smaller weight of excised glands. In addition to the wellestablished IOPTH criteria, we suggest that consideration of these independent perioperative risk factors may further inform surgical decision-making in parathyroidectomy. Future studies can build upon these models to create robust yet practical metrics that can be used in clinics and operating rooms to assess risk of persistent disease. Acknowledgements Presented in abstract form at the 2015 Academic Surgical Congress; Las Vegas, Nevada; February 2015. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. References 1. Coker LH, Rorie K, Cantley L, et al. Primary hyperparathyroidism, cognition, and health-related quality of life. Ann Surg. 2005;242:642e650. 2. Haber RS, Kim CK, Inabnet WB. Ultrasonography for preoperative localization of enlarged parathyroid glands in primary hyperparathyroidism: comparison with (99m)technetium sestamibi scintigraphy. Clin Endocrinol (Oxf). 2002;57: 241e249. 3. Gawande AA, Monchik JM, Abbruzzese TA, Iannuccilli JD, Ibrahim SI, Moore Jr FD. Reassessment of parathyroid hormone monitoring during parathyroidectomy for primary hyperparathyroidism after 2 preoperative
localization studies. Arch Surg. 2006;141:381e384. discussion 384. 4. Richards ML, Thompson GB, Farley DR, Grant CS. An optimal algorithm for intraoperative parathyroid hormone monitoring. Arch Surg. 2011;146: 280e285. 5. Yeh MW, Wiseman JE, Chu SD, et al. Population-level predictors of persistent hyperparathyroidism. Surgery. 2011;150:1113e1119. 6. Rajaei MH, Bentz AM, Schneider DF, Sippel RS, Chen H, Oltmann SC. Justified follow-up: a final intraoperative parathyroid hormone (ioPTH) over 40 pg/mL is associated with an increased risk of persistence and recurrence in primary hyperparathyroidism. Ann Surg Oncol. 2015;22:454e459. 7. Schneider DF, Mazeh H, Chen H, Sippel RS. Predictors of recurrence in primary hyperparathyroidism: an analysis of 1386 cases. Ann Surg. 2014;259:563e568. 8. Chiu B, Sturgeon C, Angelos P. Which intraoperative parathyroid hormone assay criterion best predicts operative success? A study of 352 consecutive patients. Arch Surg. 2006;141:483e487. discussion 487e488. 9. Hughes DT, Miller BS, Doherty GM, Gauger PG. Intraoperative parathyroid hormone monitoring in patients with recognized multiglandular primary hyperparathyroidism. World J Surg. 2011;35:336e341. 10. Bagul A, Patel HP, Chadwick D, Harrison BJ, Balasubramanian SP. Primary hyperparathyroidism: an analysis of failure of parathyroidectomy. World J Surg. 2014;38:534e541. 11. Lee S, Ryu H, Morris LF, et al. Operative failure in minimally invasive parathyroidectomy utilizing an intraoperative parathyroid hormone assay. Ann Surg Oncol. 2014;21:1878e1883. 12. Miller BS, England BG, Nehs M, Burney RE, Doherty GM, Gauger PG. Interpretation of intraoperative parathyroid hormone monitoring in patients with baseline parathyroid hormone levels of <100 pg/mL. Surgery. 2006;140: 883e889. discussion 889e890. 13. Wachtel H, Cerullo I, Bartlett EK, Kelz RR, Karakousis GC, Fraker DL. What can we learn from intraoperative parathyroid hormone levels that do not drop appropriately? Ann Surg Oncol. 2015;22:1781e1788. 14. Solorzano CC, Mendez W, Lew JI, et al. Long-term outcome of patients with elevated parathyroid hormone levels after successful parathyroidectomy for sporadic primary hyperparathyroidism. Arch Surg. 2008;143:659e663. discussion 663. 15. Mazeh H, Chen H, Leverson G, Sippel RS. Creation of a “Wisconsin index” nomogram to predict the likelihood of additional hyperfunctioning parathyroid glands during parathyroidectomy. Ann Surg. 2013;257:138e141. 16. Kebebew E, Hwang J, Reiff E, Duh QY, Clark OH. Predictors of single-gland vs multigland parathyroid disease in primary hyperparathyroidism: a simple and accurate scoring model. Arch Surg. 2006;141:777e782. discussion 782.
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
Contents lists available at ScienceDirect
The American Journal of Surgery journal homepage: www.americanjournalofsurgery.com
Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism David C. Cron, Steven R. Kapeles, Elizabeth A. Andraska, Sebastian T. Kwon, Peter S. Kirk, Brendan L. McNeish, Christopher S. Lee, David T. Hughes* Department of Surgery, University of Michigan Medical School, 1500 E Medical Center Dr, Ann Arbor, MI 48109, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 2 August 2016 Accepted 8 January 2017
Introduction: Many adjuncts guide surgical decision making in parathyroidectomy, yet their independent associations with outcome are poorly characterized. We examined a broad range of perioperative factors and used multivariate techniques to identify independent predictors of operative failure (persistent disease) after parathyroidectomy. Methods: This was a retrospective review of 2239 patients with primary hyperparathyroidism who underwent parathyroidectomy at a single-center from 1999 to 2014. We used multivariate logistic regress to measure associations between multiple perioperative factors and an operative failure (persistent hypercalcemia). Results: Operative failure was identified in 67 patients (3.0%). The following variables were independently associated with operative failure on multivariate analysis: IOPTH criteria met (protective, OR ¼ 0.22, P < 0.001), preoperative calcium (risk factor, OR ¼ 2.27 per unit increase, P < 0.001), weight of excised gland(s) (protective, OR ¼ 0.70 per two-fold increase, P ¼ 0.003), and preoperative PTH (protective, OR ¼ 0.55 per two-fold increase, P ¼ 0.008). Conclusion: In addition to the well-established IOPTH criteria, we suggest that consideration of the above independent perioperative risk factors may further inform surgical decision-making in parathyroidectomy. Short summary: We studied the relationship between perioperative factors and operative failure (persistent hypercalcemia) after parathyroidectomy. In addition to the standard intraoperative parathyroid hormone criteria, the following were independent risk factors for operative failure on multivariate analysis: higher preoperative calcium, lower preoperative parathyroid hormone, and lower weight of excised glands. These perioperative factors may inform surgical decision making for parathyroidectomy. © 2017 Elsevier Inc. All rights reserved.
Keywords: Primary hyperparathyroidism Persistent hypercalcemia Parathyroidectomy Operative failure
1. Introduction Primary hyperparathyroidism (PHPT) has an incidence of 1e2% in the adult population, and parathyroidectomy is the only cure.1 Surgical decision-making in parathyroidectomy is guided largely
* Corresponding author. Department of Surgery, University of Michigan Health Systems, 1500 E. Medical Center Drive, 2920 Taubman Center, SPC 5331, Ann Arbor, MI 48109-5331, USA. E-mail addresses: [email protected] (D.C. Cron), [email protected] (S.R. Kapeles), [email protected] (E.A. Andraska), [email protected] (S.T. Kwon), [email protected] (P.S. Kirk), [email protected] (B.L. McNeish), [email protected] (C.S. Lee), [email protected] (D.T. Hughes).
by preoperative imaging and intraoperative parathyroid hormone (IOPTH) monitoring. Preoperative imaging can include sestamibi ± single-photon emission computed tomography (SPECT), cervical ultrasound, and 4D-computed tomography scan.2,3 IOPTH monitoring is frequently used to ensure appropriate resection and biochemical cure.4 In experienced centers, success rates for parathyroidectomy for PHPT range from 92% to 99%.3,5e11 Nevertheless, failure does occur after this common surgery, yet failure is infrequently studied. Little is known about patient-level predictors of operative failure and persistent PHPT. Previous studies have attributed operative failure to inadequate preoperative imaging localization.3,5,10 Achievement of IOPTH criteria is a known predictor of operative success, though the final target IOPTH level is not agreed upon.
http://dx.doi.org/10.1016/j.amjsurg.2017.01.012 0002-9610/© 2017 Elsevier Inc. All rights reserved.
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
2
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
Some researchers contend that final IOPTH levels should fall into the normal range,4,9 while others recommend lower levels.6 The independent contributions of preoperative localization, IOPTH biochemical cure, and preoperative biochemical severity to operative success are unclear. Better understanding of the relationship of these factors to operative failure is necessary to further improve outcomes after this surgery and optimize intraoperative decisionmaking. We investigated predictors of operative failure (persistent disease) in patients undergoing parathyroidectomy for PHPT. Using multivariate logistic regression, we evaluated the independent effects of the following factors: demographics, preoperative biochemical disease severity, preoperative imaging, IOPTH criteria, and histopathologic findings. We aimed to identify a subset of independent risk factors for operative failure.
criteria was not achieved, localization was incorrect, or multiglandular disease was otherwise suspected. The variable of interest in analysis was achievement of IOPTH criteria (biochemical cure). In rare cases, IOPTH criteria was not achieved prior to completion of an operation. For example, IOPTH monitoring was sometimes ceased upon conversion to four-gland operation in lieu of morphologic analysis of enlarged hyperplastic parathyroid glands. Additionally, patients with low baseline PTH (<100 pg/mL) may have final IOPTH levels that fall into normal range yet do not meet criteria of decreasing 50% from baseline. The number of parathyroid glands excised was dichotomized to single gland or multiple glands for the analysis. We did not distinguish between hyperplasia and adenoma based on pathological analysis, as this is generally considered a difficult distinction. Total weight of the excised gland(s) was recorded from the pathology report.
2. Methods 2.3. Outcome 2.1. Study population This was a retrospective analysis of a prospectively maintained, Institutional Review Board approved database containing patients who underwent parathyroidectomy for PHPT at the University of Michigan Health System between 1999 and 2014. We excluded patients with the following: secondary or tertiary hyperparathyroidism, reoperative parathyroidectomy, multiple endocrine neoplasia type 1, and history of lithium use. 2.2. Preoperative factors Preoperative laboratory measurements included peak serum calcium and serum PTH levels (biochemical severity of disease). Nearly all study patients underwent preoperative imaging for gland localization, including surgeon performed ultrasound and/or sestamibi scan ± SPECT. At our institution, ultrasound is the primary imaging modality and is performed by the endocrine surgeon during initial clinical evaluation. Sestamibi is ordered if ultrasound is non-localizing or equivocal. For logistic regression analysis, the imaging localization variable was coded as: neither imaging modalities localizing or no imaging performed (the reference group), ultrasound or sestamibi localizing, ultrasound and sestamibi localizing but discordant, or ultrasound and sestamibi localizing and concordant. Imaging was considered correct if the diseased glands were located intraoperatively on the same side as indicated by imaging (without distinguishing superior/inferior). If imaging suggested an adenoma but multiglandular disease was discovered intraoperatively, this was considered incorrect imaging. The imaging correctness variable was coded as: neither imaging modalities correct or no imaging performed (the reference group), ultrasound or sestamibi correct, or both ultrasound and sestamibi correct. Patients with positive localizing studies were candidates for minimally invasive parathyroidectomy (MIP). A bilateral exploration was elected in absence of localization or for high preoperative suspicion of multiglandular disease. For logistic regression analysis, the variable of interest was four-gland operation (either planned or converted). IOPTH monitoring was used in nearly every case. Levels were drawn from the jugular vein or peripherally at initiation of surgery and/or just prior to parathyroid gland excision and at 5, 10, and 15 min post-excision. Baseline IOPTH was defined as the highest value of two samples recorded following anesthesia induction and immediately prior to gland excision. Biochemical cure was defined as both a decrease in IOPTH 50% from baseline and a post-excision IOPTH within the normal range (12e75 pg/mL). MIP was converted to four-gland exploration intraoperatively if IOPTH
Postoperative serum calcium values were collected on all patients. The lowest and highest postoperative calcium levels were recorded. At the time of analysis, for patients with peak postoperative serum calcium levels 10.2 mg/dL, charts were retrospectively reviewed for additional follow-up to rule out transient postoperative hypercalcemia, and to verify outcome. The outcome of interest was operative failure with resultant persistent hyperparathyroidism. Persistent disease was defined by postoperative serum calcium measurements persistently 10.2 mg/dL or a single calcium 11.0 mg/dL when no additional data was available (all measurements within 6 months postoperatively). Postoperative PTH labs are not routinely ordered at our institution, and thus PTH was not a component of the definition of failure. Due to biochemical followup constraints, recurrent disease (incidence beyond 6 months postoperatively) was not assessed or analyzed. 2.4. Statistical analysis Pairwise comparisons were computed using a Student's t-test, Wilcoxon rank-sum test, or Fisher's exact test, as appropriate. The following variables were considered as potential predictors of operative failure: age, sex, preoperative peak PTH, preoperative peak calcium, imaging localization, imaging correctness, four-gland approach, achievement of IOPTH criteria, number of glands excised, and total weight of excised gland(s). Highly skewed variables were transformed using a logarithm of base two. Univariate and multivariate logistic regression were utilized to assess the associations between the perioperative factors and operative failure. For multivariate analysis, we randomly divided our sample into a training set (60% of sample) and an internal validation set (40% hold-out sample) to assess predictive ability of the models. Backward stepwise selection was used to identify independent predictors of operative failure (P 0.100 for inclusion in the model). Because our predictors of interest included both preoperative and intraoperative variables, we developed two main models: a preoperative model and a combined intraoperative model (the latter considered all variables). Hosmer-Lemeshow test was used to ensure model goodness of fit. Variance inflation factor was used to ensure no problems with multicollinearity. Area under the receiver operating characteristic curve (AUROC) was used to assess model predictive ability. We performed secondary analyses to evaluate consistency of our findings. First, we grouped preoperative calcium and PTH by tertile, and we examined rates of operative failure by tertile of calcium and PTH. As a secondary outcome, we assessed predictors of elevated postoperative calcium (calcium 10.2 mg/dL regardless
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
of persistence; binary outcome) using logistic regression. To determine if any statistical effects were mediated by the number of glands excised, we tested for interactions between the variable multiple glands excised and the following variables: preoperative peak calcium, preoperative peak PTH, and IOPTH criteria. Lastly, since IOPTH criteria is known to be strongly predictive of operative success, we performed a multivariate analysis excluding IOPTH criteria as a variable, to assess the effects of other predictors without accounting for IOPTH criteria. A significance level of P 0.05 was used for all significance tests. Statistical analyses were performed using Stata v 13.0 (College Station, TX). 3. Results 3.1. Study population 2239 patients met inclusion criteria. We identified 67 failures (3.0%). Of these failures, 14 (21%) had a reoperation at our institution at the time of the completion of this study's analysis. The median follow-up for postoperative calcium values (time between operation and last available calcium laboratory) was 18 days. Of the failures, the median follow-up for postoperative calcium values was 46 days. Case mix included 64.2% MIPs, 10.6% MIPs converted to four-gland explorations, and 25.3% planned four-gland explorations. 78% of our sample was female, and the mean age was 59 ± 13 years. Table 1 shows patient characteristics for all patients, successful cases, and operative failure cases. Preoperative PTH and weight of excised gland(s) were significantly skewed and were therefore log transformed for regression analyses. 3.2. Univariate logistic regression Fig. 1 shows the results of univariate logistic regression. Risk factors for operative failure included the following: multiple glands excised (Odds Ratio [OR] ¼ 2.96, 95% CI: 1.79e4.90, P < 0.001), fourgland exploration (OR ¼ 2.44, 95% CI: 1.48e4.02, P < 0.001), and higher preoperative peak calcium (OR ¼ 1.38 per unit increase, 95% CI: 1.03e1.83, P ¼ 0.028). Protectors against failure included IOPTH criteria met (OR ¼ 0.21, 95% CI 0.13e0.35, P < 0.001), larger weight of excised gland(s) (OR ¼ 0.70 per two-fold increase, 95% CI: 0.59e0.82, P < 0.001), ultrasound and sestamibi both correct (OR ¼ 0.23 compared to no correct imaging, 95% CI: 0.09e0.61, P ¼ 0.003), higher preoperative peak PTH (OR ¼ 0.66 per two-fold increase, 95% CI: 0.48e0.91, P ¼ 0.011), and ultrasound and sestamibi localizing and concordant (OR ¼ 0.31 compared to no
3
localizing imaging, 95% CI: 0.13e0.78, P ¼ 0.013). Factors not significantly associated with operative failure included age (P ¼ 0.118), ultrasound or sestamibi correct (P ¼ 0.179), ultrasound or sestamibi localizing (P ¼ 0.391), female sex (P ¼ 0.577), and ultrasound and sestamibi localizing but discordant (P ¼ 0.693). 3.3. Multivariate analysis: combined intraoperative model Fig. 2 shows the variables included in the combined intraoperative model. The multivariate model included the following independent predictors of operative failure: IOPTH criteria met (OR ¼ 0.22, 95% CI: 0.11e0.44, P < 0.001), preoperative peak calcium (OR ¼ 2.27, 95% CI: 1.46e3.53, P < 0.001), weight of excised gland(s) (OR ¼ 0.70 per two-fold increase in weight, 95% CI: 0.55e0.89, P ¼ 0.003), and preoperative peak PTH (OR ¼ 0.55 per two-fold increase, 95% CI: 0.36e0.86, P ¼ 0.008). The training and validation set AUROC values for this model were 0.79 (95% CI: 0.71e0.87) and 0.73 (95% CI: 0.62e0.84), respectively. 3.4. Multivariate analysis: preoperative model Two variables were included in the preoperative model: preoperative peak PTH (OR ¼ 0.47 per two-fold increase, 95% CI: 0.30e0.74, P ¼ 0.001) and preoperative peak calcium (OR ¼ 2.01 per unit increase, 95% CI: 1.31e3.09, P ¼ 0.001). The training and validation set AUROC values for this model were 0.71 (95% CI: 0.62e0.79) and 0.58 (95% CI: 0.46e0.70), respectively. 3.5. Secondary analyses To assess the consistency of these findings, we performed secondary analyses as described in the methods section. Fig. 3 shows rates of operative failure by tertile of preoperative calcium and PTH. Compared to the lowest calcium tertile, patients in the highest tertile had 3.4-fold higher rates of operative failure (P < 0.001). Compared to patients in the highest PTH tertile, patients in the lowest tertile had 2.2-fold higher rates of operative failure (P ¼ 0.005). We used multivariate logistic regression to identify predictors of elevated postoperative calcium (calcium 10.2 regardless of persistence). In multivariate logistic regression, the following factors were independent predictors of elevated postoperative calcium: preoperative peak calcium (OR ¼ 2.05, 95% CI: 1.66e2.54, P < 0.001), preoperative peak PTH (OR ¼ 0.71, 95% CI: 0.58e0.87, P ¼ 0.001), multiple glands excised (OR ¼ 1.77, 95% CI: 1.22e2.57, P ¼ 0.003), and IOPTH criteria met (OR ¼ 0.60, 95% CI: 0.40e0.89,
Table 1 Clinical Characteristics of the Study Cohort. The p-value compares success vs. failure. Continuous variables are represented as mean ± standard deviation or median (interquartile range) if skewed. Binary variables are represented as count (frequency). Calcium measurements are in mg/dL. PTH measurements are in pg/mL.
Total N Age Female Preoperative peak PTH Preoperative peak calcium Ultrasound performed Localizing Correct Sestamibi performed Localizing Correct Four-gland approach IOPTH criteria achieved Weight excised gland(s) (grams)
Total
Success
Failure
P-value
2239 59 ± 13 1742 (78%) 122 (81) 11.0 ± 0.7 1798 (84%) 1521 (85%) 1247 (82%) 1253 (56%) 856 (68%) 746 (87%) 765 (36%) 1887 (84%) 0.6 (0.7)
2172 59 ± 13 1688 (78%) 123 (82) 11.0 ± 0.7 1743 (84%) 1474 (85%) 1213 (82%) 1217 (59%) 842 (69%) 737 (88%) 728 (35%) 1850 (85%) 0.6 (0.7)
67 57 ± 13 54 (81%) 97 (69) 11.2 ± 0.6 55 (85%) 47 (85%) 34 (72%) 36 (55%) 14 (39%) 9 (64%) 37 (57%) 37 (55%) 0.4 (0.5)
0.117 0.656 0.003 0.029 1 1 0.085 0.611 <0.001 0.025 0.001 <0.001 <0.001
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
4
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
Fig. 1. Univariate logistic regression analysis of predictors of operative failure in parathyroidectomy. This figure displays the unadjusted odds ratios and 95% confidence intervals for all variables.
P ¼ 0.012). The training and validation set AUROC values for this model were 0.68 (95% CI: 0.63e0.71) and 0.62 (95% CI: 0.57e0.67), respectively. There were no significant statistical interactions between number of glands excised and calcium, PTH, or IOPTH criteria. This indicates that the effects of calcium, PTH, and IOPTH criteria on operative failure were independent of number of glands excised. Lastly, when we performed a multivariate analysis without including IOPTH criteria as a covariate, the following factors were independently predictive of operative failure: preoperative peak calcium (OR ¼ 2.47, 95% CI: 1.59e3.84, P < 0.001), preoperative peak PTH (OR ¼ 0.62, 95% CI: 0.40e0.94, P ¼ 0.025), multiple glands excised (OR ¼ 3.19, 95% CI: 1.59e6.43, P ¼ 0.001), and total weight of excised gland(s) (OR ¼ 0.70, 95% CI: 0.55e0.88, P ¼ 0.002).
4. Discussion This study used multivariate techniques to identify independent predictors of operative failure after parathyroidectomy for PHPT. The prevalence of failure in this cohort was 3%. In multivariate analysis, factors independently and significantly associated with increased risk of failure included lower preoperative PTH, higher preoperative calcium, failure to achieve IOPTH criteria of biochemical cure, and lower weight of excised gland(s). We reported similar findings for predictors of elevated postoperative calcium (regardless of persistence). When analyzed by tertile of preoperative calcium level, operative failure rates were highest in the upper tertile. Conversely, for preoperative PTH, operative failure rates were highest in the lower tertile. These results provide
Fig. 2. Multivariate logistic regression analysis of predictors of operative failure in parathyroidectomy. This figure displays the odds ratios and 95% confidence intervals for the 4 variables that were in the final combined intraoperative multivariate model.
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
Fig. 3. Rates of operative failure by tertile of preoperative calcium and PTH. This figure displays observed rates of operative failure (persistent hyperparathyroidism) by tertile of preoperative calcium (in mg/dL) and preoperative PTH (in pg/mL).
insight into the independent relationships between biochemical disease severity, commonly used perioperative adjuncts, and operative failure. This combination of factors may help inform surgical decision-making beyond use of IOPTH criteria alone. Our failure rate of 3% was similar to rates reported by other studies.3,5e11 An unexpected finding was the opposite effects of preoperative PTH compared to preoperative calcium. We expected that lower values of both preoperative PTH and preoperative calcium would be associated with higher rates of failure. We expected this because of the relation of low baseline PTH values to increased incidence of multiglandular parathyroid disease and the possible increased failure rate associated with this.12,13 In our study, lower PTH and higher calcium were risk factors for failure. The positive association between calcium and persistent disease could be due in part to the fact that persistent disease is defined based on calcium levels. However, this does not completely explain the direction or strength of the relationship, as the strong association was observed across the range of calcium values. It is possible that patients with higher preoperative calcium have a higher physiologic set-point for calcium homeostasis and are thus more likely to remain hypercalcemic postoperatively. It is also possible that higher calcium levels inhibit IOPTH levels and impair detection of additional hyperfunctioning tissue intraoperatively. Future studies are needed to explore these hypotheses and elucidate the differential effects of PTH and calcium on persistent disease. Imaging variables were predictive of failure in univariate analysis but not multivariate analysis. This does not imply that preoperative imaging is not an important adjunct; rather, it implies that the effect of imaging can be explained by other factors. Localizing imaging is inherently highly correlated with size and location of the enlarged glands, and scans that localize to one side favor MIP for the operative approach. Because weight of excised gland(s) was a strong predictor and present in the final model, it is not surprising that imaging variables and operative approach were excluded from the final multivariate model. Preoperative imaging is important to inform operative approach and to facilitate localization of the diseased glands. However, for the purpose of predicting operative success intraoperatively, our results suggest that the weight of excised parathyroid tissue, IOPTH criteria, and preoperative PTH and calcium are more important factors. Smaller weight of excised parathyroid gland(s) was a risk factor for failure. This is intuitive, as large glands are easier to locate and generally associated with single gland disease. Lastly, failure to meet IOPTH criteria was a strong predictor
5
of operative failure; however, our relatively stringent IOPTH criteria still yielded a significant amount of false positives, as 55% of failures met IOPTH criteria for biochemical cure. IOPTH data should be considered in the context of other patient specific perioperative factors. The limitations to this study include its retrospective study design. Also, this is a cohort from a single large academic institution's endocrine surgery division; therefore, our results may not be generalizable across all centers. Since our center is a tertiary care center, many of our patients had limited long-term biochemical follow-up within our healthcare system. Patients who were eucalcemic at their two-week postoperative appointment generally did not have additional calcium measurements ordered by the surgeon. Our reported failure rate may be an underestimate due to lack of follow-up; however, most cases of persistent disease have elevated calcium levels at their postoperative visit. Postoperative PTH was not a component of this definition; however, postoperative PTH levels are generally non-informative, as PTH elevation in the setting of eucalcemia is common postoperatively.14 Our failure rate could also be an overestimate if elevated calcium levels dropped after patients were lost to follow-up. Prior to analysis, we performed an additional extensive review of the electronic medical records of all patients with elevated postoperative calcium levels (10.2 mg/dL) in order to rule out hypercalcemia due to postoperative calcium supplementation. Lastly, our models showed moderate predictive ability, and we were limited to using an internal validation set. However, the predictive strength of the models was not a primary focus of this study. We were more interested in identifying independent predictors, and future work including more perioperative factors is needed to create robust predictive models. The strengths of this study include its large sample size, the use of multivariate techniques, and a focus on operative failure. In the context of operative failure, many studies focus on the diagnostic accuracy and predictive ability of individual adjuncts only. Multivariate techniques are important to identify the independent contributions of individual adjuncts. Our results emphasize this by showing that factors such as preoperative imaging can predict outcome in univariate analysis but can be explained by other variables (such as weight of enlarged glands). An important novel finding of this study is the correlation between preoperative biochemical severity of disease and outcome. Patients with high calcium and low PTH were at increased risk of failure independent of intraoperative factors like IOPTH. These risk factors should raise a surgeon's suspicion of multiglandular disease and failure risk, and this information may lower one's threshold for converting to a fourgland operation or continuing to monitor IOPTH beyond the standard criteria of biochemical cure. Predictive models can aid risk assessment and may provide objective metrics for intraoperative decision-making. The challenge is to make a prediction tool easy and efficient to use intraoperatively. Mazeh et al. recently developed a nomogram based on preoperative PTH and calcium to be used intraoperatively to predict likelihood of additional hyperfunctioning parathyroid tissue.15 Kebebew et al. developed a scoring system using multiple preoperative factors to distinguish single-gland vs. multiglandular cases.16 At this time, our model is too complex to be practically used intraoperatively, and these results serve as a proof of concept. Future work may extend these methods to an easy-to-use nomogram predicting failure risk based on preoperative biochemical severity, IOPTH levels, histopathologic findings, and more. Incorporating additional measurements such as renal function, 25-hydroxy vitamin D, and more granular IOPTH measurements may yield models with higher predictive ability than we show here. At the least, these results provide important insight into the independent contributions of patient-level
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012
6
D.C. Cron et al. / The American Journal of Surgery xxx (2017) 1e6
characteristics to outcome after parathyroidectomy. 5. Conclusion With this work, we show that multivariate techniques can be used to identify independent risk factors for operative failure after parathyroidectomy. We identified four independent predictors of persistent postoperative hypercalcemia: lower preoperative PTH, higher preoperative calcium, failure to meet IOPTH criteria, and smaller weight of excised glands. In addition to the wellestablished IOPTH criteria, we suggest that consideration of these independent perioperative risk factors may further inform surgical decision-making in parathyroidectomy. Future studies can build upon these models to create robust yet practical metrics that can be used in clinics and operating rooms to assess risk of persistent disease. Acknowledgements Presented in abstract form at the 2015 Academic Surgical Congress; Las Vegas, Nevada; February 2015. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. References 1. Coker LH, Rorie K, Cantley L, et al. Primary hyperparathyroidism, cognition, and health-related quality of life. Ann Surg. 2005;242:642e650. 2. Haber RS, Kim CK, Inabnet WB. Ultrasonography for preoperative localization of enlarged parathyroid glands in primary hyperparathyroidism: comparison with (99m)technetium sestamibi scintigraphy. Clin Endocrinol (Oxf). 2002;57: 241e249. 3. Gawande AA, Monchik JM, Abbruzzese TA, Iannuccilli JD, Ibrahim SI, Moore Jr FD. Reassessment of parathyroid hormone monitoring during parathyroidectomy for primary hyperparathyroidism after 2 preoperative
localization studies. Arch Surg. 2006;141:381e384. discussion 384. 4. Richards ML, Thompson GB, Farley DR, Grant CS. An optimal algorithm for intraoperative parathyroid hormone monitoring. Arch Surg. 2011;146: 280e285. 5. Yeh MW, Wiseman JE, Chu SD, et al. Population-level predictors of persistent hyperparathyroidism. Surgery. 2011;150:1113e1119. 6. Rajaei MH, Bentz AM, Schneider DF, Sippel RS, Chen H, Oltmann SC. Justified follow-up: a final intraoperative parathyroid hormone (ioPTH) over 40 pg/mL is associated with an increased risk of persistence and recurrence in primary hyperparathyroidism. Ann Surg Oncol. 2015;22:454e459. 7. Schneider DF, Mazeh H, Chen H, Sippel RS. Predictors of recurrence in primary hyperparathyroidism: an analysis of 1386 cases. Ann Surg. 2014;259:563e568. 8. Chiu B, Sturgeon C, Angelos P. Which intraoperative parathyroid hormone assay criterion best predicts operative success? A study of 352 consecutive patients. Arch Surg. 2006;141:483e487. discussion 487e488. 9. Hughes DT, Miller BS, Doherty GM, Gauger PG. Intraoperative parathyroid hormone monitoring in patients with recognized multiglandular primary hyperparathyroidism. World J Surg. 2011;35:336e341. 10. Bagul A, Patel HP, Chadwick D, Harrison BJ, Balasubramanian SP. Primary hyperparathyroidism: an analysis of failure of parathyroidectomy. World J Surg. 2014;38:534e541. 11. Lee S, Ryu H, Morris LF, et al. Operative failure in minimally invasive parathyroidectomy utilizing an intraoperative parathyroid hormone assay. Ann Surg Oncol. 2014;21:1878e1883. 12. Miller BS, England BG, Nehs M, Burney RE, Doherty GM, Gauger PG. Interpretation of intraoperative parathyroid hormone monitoring in patients with baseline parathyroid hormone levels of <100 pg/mL. Surgery. 2006;140: 883e889. discussion 889e890. 13. Wachtel H, Cerullo I, Bartlett EK, Kelz RR, Karakousis GC, Fraker DL. What can we learn from intraoperative parathyroid hormone levels that do not drop appropriately? Ann Surg Oncol. 2015;22:1781e1788. 14. Solorzano CC, Mendez W, Lew JI, et al. Long-term outcome of patients with elevated parathyroid hormone levels after successful parathyroidectomy for sporadic primary hyperparathyroidism. Arch Surg. 2008;143:659e663. discussion 663. 15. Mazeh H, Chen H, Leverson G, Sippel RS. Creation of a “Wisconsin index” nomogram to predict the likelihood of additional hyperfunctioning parathyroid glands during parathyroidectomy. Ann Surg. 2013;257:138e141. 16. Kebebew E, Hwang J, Reiff E, Duh QY, Clark OH. Predictors of single-gland vs multigland parathyroid disease in primary hyperparathyroidism: a simple and accurate scoring model. Arch Surg. 2006;141:777e782. discussion 782.
Please cite this article in press as: Cron DC, et al., Predictors of operative failure in parathyroidectomy for primary hyperparathyroidism, The American Journal of Surgery (2017), http://dx.doi.org/10.1016/j.amjsurg.2017.01.012