Comparison of Parotidectomy With Observation for Treatment of Pleomorphic Adenoma in Adults

Key Points

Question  How does parotidectomy compare with observation for the management of pleomorphic adenoma in adults 50 years or older?

Findings  In this decision analytical model of patients 50 years or older, observation was more beneficial than surgery at 92 years of age for patients with superficial lobe tumors and at 85 years of age for patients with deep lobe tumors. There was no significant difference in outcomes between parotidectomy and observation among patients aged 70 to 80 years.

Meaning  This study suggests that the outcomes associated with parotidectomy and observation are similar at 70 years or older among patients with pleomorphic adenoma and that observation may be the favorable treatment in that age group.

Importance  There is no consensus regarding optimal management of pleomorphic adenoma in adults.

Objectives  To compare parotidectomy with observation for the management of pleomorphic adenoma in patients 50 years or older by age.

Design and Setting  This decision analytical model was performed from November 21, 2019, to June 15, 2020, using a Markov model. Model variables and ranges were selected based on a literature review. A 1-way sensitivity analysis was performed to evaluate the age threshold at which each algorithm, either upfront elective parotidectomy or observation, would be favored. A Monte Carlo probabilistic sensitivity analysis using variable ranges was then performed 5 times with patients in the model assigned a starting age of 50, 60, 70, 80, and 90 years to assess how age at diagnosis would be associated with the model results.

Main Outcomes and Measures  Model outcomes were measured with quality-adjusted life-years (QALYs).

Results  In the study models, the age thresholds at which observation became more beneficial than parotidectomy were 88.5 years for patients with superficial lobe tumors (5.37 QALYs in favor of parotidectomy below this age, and 5.37 QALYs in favor of observation above this age) and 83.4 years for patients with deep lobe tumors (7.51 QALYs in favor of surgery below this age, and 7.51 QALYs in favor of observation above this age). There was no significant difference in outcomes between parotidectomy and observation among patients aged 70 to 80 years.

Conclusions and Relevance  This study suggests that the outcomes associated with parotidectomy and observation are similar at 70 years or older among patients with pleomorphic adenoma and that observation may be the favorable treatment in that age group.

Pleomorphic adenoma, a slow-growing, benign salivary gland tumor with incidence rates between 2.4 and 4.9 per 100 000 persons per year, is the most common salivary gland tumor.^1,2 The current mainstay of treatment for pleomorphic adenoma is surgical excision.¹ However, parotidectomy is associated with significant risk of complications. The rate of permanent facial nerve injury after parotidectomy, for instance, is up to 26% according to a meta-analysis.³ Other less serious but still common adverse effects of parotidectomy include risks of Frey syndrome, first bite syndrome, and skin hypo-anesthesia.⁴ It is also important to consider the perioperative risk of general anesthesia. Consequently, the risk of complications should be considered when recommending surgery. Some of the reasons for recommending parotidectomy include correction of cosmetic deformity caused by a tumor, diagnosis and treatment of potential missed malignant neoplasms, and prevention of malignant transformation.

Pleomorphic adenoma is associated with an estimated 1.5% risk of malignant transformation to carcinoma ex pleomorphic adenoma within the first 5 years of diagnosis if left untreated.⁵ This risk may increase to 10% if left untreated for more than 15 years.⁵ Although the risk of malignant transformation is low, it is substantial for younger patients owing to their longer life expectancy and justifies surgery. However, many patients who receive a diagnosis of pleomorphic adenoma late in life may never develop malignant transformation, and surgery may be associated with unnecessary risk.

To our knowledge, there is no consensus regarding optimal management of pleomorphic adenoma in adults. Furthermore, it is unclear at what age parotidectomy ceases to be beneficial for elderly patients with pleomorphic adenoma. This decision analytical model compared parotidectomy with observation for the management of pleomorphic adenoma in adults 50 years or older. The study was intended to support clinical decision-making regarding the management of parotid pleomorphic adenoma in elderly patients.

This decision analytical model was performed from November 21, 2019, to June 15, 2020. A Markov model was constructed using TreeAgePro, version 2019 (TreeAge Software Inc) to compare 2 potential treatment algorithms for pleomorphic adenoma of the parotid gland. Specifically, the model compared upfront elective parotidectomy with observation in patients 50 years or older. To more accurately reflect true clinical decision-making, the initial diagnosis of pleomorphic adenoma was assumed to have been made based on fine-needle aspiration (FNA) cytologic findings. Given the imperfect diagnostic accuracy of FNA, the model accounted for the false-negative rates of FNA as they relate to pleomorphic adenoma.⁶ Therefore, a set proportion of patients in the model were assumed to have undetected malignant disease at diagnosis. False-positive cases (FNA suggestive of malignant neoplasm but final pathologic findings consistent with pleomorphic adenoma) were not included in the model because treatment for presumed parotid malignant neoplasm requires an altogether different clinical decision-making algorithm.

In the upfront elective parotidectomy algorithm, patients were assumed to have either true-positive results (pleomorphic adenoma) or false-negative results (malignant neoplasm). Patients with true-positive results were modeled to receive parotidectomy within 1 year of diagnosis and then require no additional treatment. Patients with false-negative results were assumed to have either high-grade or low-grade disease based on previously established proportions by Xiao et al.⁷ Those with high-grade carcinoma were modeled to receive total parotidectomy followed by adjuvant radiotherapy, and those with low-grade carcinoma were modeled to receive total parotidectomy alone. The decision analysis was then performed under 2 distinct sets of circumstances to represent either superficial parotidectomy (for pleomorphic adenoma confined to the superficial lobe of the parotid) or total parotidectomy (for pleomorphic adenoma extending into the deep lobe). The probability of facial nerve injury after surgery was different depending on which of these 2 procedures was modeled.

Patients in the observation group were assumed to require no additional treatment unless the pleomorphic adenoma transformed to carcinoma ex pleomorphic adenoma. If a transformation occurred, patients were modeled to receive total parotidectomy followed by adjuvant radiotherapy in cases of high-grade disease or total parotidectomy alone in cases of low-grade disease. Patients with a false-negative diagnosis were assumed to immediately undergo transformation and then receive the same treatment as above.

Each cycle in the model was assumed to represent 1 year, and the model was allowed to run for 50 cycles or until all patients had died. Patients were assumed to be otherwise healthy. Perioperative mortality was estimated to increase with age on the basis of preexisting values estimated by the Centers for Disease Control and Prevention.⁸ Perioperative morbidity was estimated using the risk of medical complications previously determined by Bovenzi et al⁹ and was also set to increase with age. The full model is available in the eFigure in the Supplement.

A total of 9 health states were incorporated into the model, including (1) life with untreated pleomorphic adenoma, (2) normal facial nerve function after elective parotidectomy, (3) permanent facial nerve injury after elective parotidectomy, (4) normal facial nerve function after total parotidectomy and adjuvant radiotherapy for high-grade carcinoma ex pleomorphic adenoma, (5) permanent facial nerve injury after total parotidectomy and adjuvant radiotherapy for high-grade carcinoma ex pleomorphic adenoma, (6) active adjuvant radiotherapy for high-grade carcinoma ex pleomorphic adenoma, (7) normal facial nerve function after total parotidectomy for low-grade carcinoma ex pleomorphic adenoma, (8) permanent facial nerve injury after total parotidectomy for low-grade carcinoma ex pleomorphic adenoma, and (9) death. Potential transitions between states were modeled to occur every year and are shown in Figure 1 and Figure 2.

Utility scores and their associated ranges were assigned for each state based on previously validated scores from the literature or best estimates if specific utility values were unavailable (Table 1).^3,6,7,10-16 Life with untreated pleomorphic adenoma was assumed to have a utility of 1, and death was modeled with a utility of 0. Normal facial nerve function after parotidectomy for pleomorphic adenoma was also modeled with a utility of 1 because of the excellent quality of life experienced by most patients after parotidectomy.¹⁷ Permanent facial nerve injury after elective parotidectomy for pleomorphic adenoma was modeled to have a utility of 0.79 (range, 0.79-1.00) based on previously determined values by Sinno et al.¹⁵ The associated range was assumed to constitute only the higher range of the 95% CI found by Sinno et al¹⁵ to account for most patients with facial nerve injury having partial and not complete paralysis.¹¹ Normal facial nerve function after total parotidectomy and adjuvant radiotherapy for high-grade carcinoma ex pleomorphic adenoma was assigned a utility of 0.87 (range, 0.70-0.87) based on predetermined utility values by Noel et al¹⁶ for patients with head and neck cancer who have undergone surgery and adjuvant radiotherapy. Normal facial nerve function after total parotidectomy for low-grade carcinoma ex pleomorphic adenoma was assigned a utility of 0.95 (range, 0.77-0.96) based on predetermined utility values by Noel et al¹⁶ for patients with head and neck cancer who did not require adjuvant radiotherapy.

Preexisting utility values for permanent facial nerve injury after total parotidectomy for either high-grade or low-grade carcinoma ex pleomorphic adenoma do not exist in the literature, to our knowledge. Therefore, these values were estimated based on predetermined values by Sinno et al¹⁵ and Noel et al¹⁶ and assigned a value of 0.69 (range, 0.51-0.87) for high-grade disease and 0.75 (range, 0.55-0.95) for low-grade disease. The high-grade value was evaluated by reducing the utility determined by Sinno et al¹⁵ for facial nerve paralysis by 13% to account for the decreased quality of life owing to the history of radiotherapy and cancer as determined by Noel at al.¹⁶ The low-grade value was evaluated by reducing the utility determined by Sinno et al¹⁵ for facial nerve paralysis by 5% to account for the decreased quality of life owing to cancer history alone, as determined by Noel at al.¹⁶ The associated ranges were assumed to contain the full 95% CI found by Sinno et al¹⁵; however, distribution was assumed to be triangular and focused around the upper quartile of values to reflect that not all patients with facial nerve injury from carcinoma ex pleomorphic adenoma have complete paralysis.¹⁰ Last, active adjuvant radiotherapy for carcinoma ex pleomorphic adenoma was modeled to have a utility value of 0.66 (range, 0.47-0.85) based on previously validated utility scores found by Liao et al¹⁴ for patients undergoing radiotherapy for head and neck cancer.

Patients modeled to develop a postoperative medical complication experienced a disutility of 0.12 (range, 0.06-0.24) for a single cycle. This disutility was selected based on previously validated utilities corresponding to 6 months of deep venous thrombosis. Although not all patients experience deep venous thrombosis as a medical complication, it was selected as a proxy and as a compromise between more severe complications, such as stroke, and less severe complications, such as acute kidney injury, all of which could not be modeled separately.

Values from the literature were used to inform the model variables for both fixed variables (Table 1)^3,6,7,10-16 and time-dependent variables (Table 2).^5,8-10,18 Fixed variables were assumed to have the same value regardless of when they occur in the model, such as the probability of facial nerve injury after superficial parotidectomy. The probabilities of facial nerve injury that were used in the model were based on data from multiple studies that had mixed use of intraoperative facial nerve monitoring. Time-dependent variables were assigned values that change depending on when they occurred in the model. This variation was based on how long a patient was in a particular state, such as the probability of death from carcinoma ex pleomorphic adenoma, or it reflected variation based on the patient’s age, such as the probability of perioperative death and perioperative medical complications. A full list of values for each time-dependent variable is presented in the eTable in the Supplement.

The model assumed that patients in the elective surgery group with false-negative results were more likely to be treated earlier and thus have improved survival.^10,19 Also, to provide an element of consistency in the model, all patients with false-negative results were assumed to have carcinoma ex pleomorphic adenoma. However, it is possible that patients with false-negative results of FNA could have various parotid malignant neoplasms, but this was not included in the model.

The expected values in the model and the associated ranges were selected from the published literature. When multiple published values existed, the values used in the model were selected by an iterative process between 2 of us (M.P.K. and N.A.) and were intended to represent the most likely values for each variable. The senior author (U.C.M.) independently verified the values used in the model. Distributions were selected based on available data and certainty of the mean or distribution around the base case. Given the relative uncertainty about the true mean for the utility values selected for the model, a uniform distribution was selected for most utilities. However, if there was a high degree of certainty about the true mean for other variables, a triangular distribution was selected. Time-dependent variables were estimated with only base-case values and not assigned ranges given the inherent limitations with modeling time-dependent variables.

Effectiveness outcomes were modeled as quality-adjusted life years (QALYs) based on utility scores. All patients in the model were assumed to die by age 100 years based on Centers for Disease Control and Prevention life tables.¹⁸

A 1-way sensitivity analysis was performed to evaluate the age threshold at which each algorithm, either upfront elective parotidectomy or observation, would be favored. The sensitivity analysis was performed using the estimated values based on variables ranges. The 1-way sensitivity analysis was performed to assess results both for elective superficial parotidectomy and for elective total parotidectomy.

A probabilistic sensitivity analysis (PSA) was then performed 5 times with patients in the model assigned a starting age of 50, 60, 70, 80, and 90 years to assess how age at diagnosis would impact the model results. A Monte Carlo analysis consisting of 1000 individual simulations was performed. All fixed probability values and utility values with a root definition other than 1 or 0 were assessed as part of the Monte Carlo PSA. The PSA was performed to assess results for both elective superficial parotidectomy and elective total parotidectomy.

The 1-way sensitivity analysis for superficial parotidectomy demonstrated that below age 88.5 years, elective parotidectomy would be the more effective algorithm, whereas above this age, observation would be more effective (5.37 QALYs in favor of surgery below this age and 5.37 QALYs in favor of observation above this age). When considering total parotidectomy, 1-way sensitivity analysis demonstrated that below age 83.4 years, elective surgery would be the more effective algorithm, whereas above this age, observation would be more effective (7.51 QALYs in favor of surgery below this age and 7.51 QALYs in favor of observation above this age).

Table 3 gives the results of the PSA at ages 50, 60, 70, 80, and 90 years for both total parotidectomy and superficial parotidectomy. At ages 50, 60, 70, and 80 years, elective surgery would be the optimal strategy, whereas at age 90 years, observation would be equivalent or optimal.

The incremental advantage of parotidectomy compared with observation was greatest at the youngest ages, and the gap closed as age increased. At age 50 years, for instance, the incremental effect of superficial parotidectomy compared with observation was 1.05 QALYs and of total parotidectomy compared with observation was 0.95 QALYs. At age 70 years, the incremental effect of superficial parotidectomy compared with observation was 0.20 QALYs and of total parotidectomy compared with observation was 0.15 QALYs. By age 90 years, the incremental effect of superficial parotidectomy compared with observation was –0.01 QALYs and of total parotidectomy compared with observation was –0.02 QALYs.

To our knowledge, this study is the first to compare surgery with observation for treatment of pleomorphic adenoma of the parotid gland in adults 50 years or older using a Markov chain decision analysis model. The findings suggest that when balancing the risk of malignant transformation with the risk of complications from parotidectomy, observation may be reasonable for patients 70 years or older. Furthermore, the present study suggests that the age thresholds at which observation becomes more beneficial than surgery are 88.5 years for patients with superficial lobe tumors and 83.4 years for patients with deep lobe tumors. However, although there was a significant difference in outcomes between parotidectomy and observation among patients aged 70 to 80 years the difference is so small that it is likely not clinically significant, suggesting that observation may be reasonable for this age group.

There remains a paucity of studies in the literature evaluating observation alone as a treatment modality for pleomorphic adenoma in elderly patients. Malik et al²⁰ suggest that the negative cognitive impacts of general anesthesia in elderly patients (not accounted for in this model) may make observation more favorable compared with upfront surgery; however, few other studies have addressed this option. Some authors suggest that parotidectomy performed under local anesthesia may be a reasonable option for elderly patients.²¹ Regarding recurrent pleomorphic adenoma, some studies indicate that observation may be preferable for elderly patients, but a detailed explanation with age recommendations is lacking.^22,23

In the sensitivity analysis, 88.5 years was the threshold after which observation maximized QALYs compared with superficial parotidectomy in patients with superficial lobe tumors. In patients with deep lobe tumors, 83.4 years was the age threshold after which observation maximized QALYs compared with total parotidectomy. These thresholds should be interpreted with caution because they are based on the age at which one strategy provided greater QALYs compared with the other and these differences may not be clinically significant. A more nuanced interpretation of the Monte Carlo PSA demonstrated a near convergence in outcomes at a younger age, which is shown in Table 3. In fact, the difference between both strategies was small at most ages analyzed in the model. At the youngest age assessed in this model, 50 years, the maximal difference between superficial parotidectomy and observation was 1.05 QALYs (or the equivalent of approximately 1 year of good health). By age 70 years, the difference between superficial parotidectomy and observation was reduced to 0.20 QALYs (or the equivalent of slightly more than 2 months of good health). By age 80 years, the difference between superficial parotidectomy and observation was 0.05 QALYs (or the equivalent of approximately 2 weeks of good health). This convergence suggests that the clinically significant age threshold after which observation and superficial parotidectomy are roughly equivalent is closer to 70 years of age and likely not more than 80 years. The convergence in values was similar for total parotidectomy. A significant degree of clinician discretion, therefore, may be indicated based on individual patient preferences and patient comorbidities. For example, it may be reasonable to offer surgery for an otherwise healthy 79-year-old patient who wants to avoid the risk of malignant transformation, while also offering observation to a 70-year-old patient with substantial comorbidities who is averse to surgery.

This study has limitations. In particular, although the tested treatment algorithms are representative of real-life situations, there are also, by necessity, simplifications. For instance, the clinical course for pleomorphic adenoma is more nuanced than accounted for in the model and may include recurrences of benign or malignant disease,²⁴ transformation with late-stage presentations of metastatic carcinoma,^19,25 and complications beyond facial nerve injury, deep venous thrombosis, and perioperative death. Modeling this degree of detail, however, would not have been practical or feasible. The model results are specific to pleomorphic adenoma of the parotid gland and should not be extrapolated to pleomorphic adenoma of other sites.

Tumor size was not specifically accounted for in the model. Although some articles directly implicate tumor size as a factor associated with facial nerve outcomes,^26,27 others reported that size is not significantly associated with permanent facial nerve outcomes.^11,28-30 Without a meta-analysis to synthesize these data, tumor size was not included. According to the study by Jin et al,¹¹ which is the largest single series included in this analysis, tumor location (superficial vs deep lobe) was associated with permanent postoperative facial nerve outcomes but not tumor size. Jin et al¹¹ suggest that many articles that find tumor size to be a factor associated with facial nerve outcomes do not account for tumor location. Larger tumors may have different rates of transformation compared with smaller tumors; however, definitive data on this remain unclear. Given this limitation in the model, large tumors may not be accurately reflected in this decision analysis.

Other key assumptions and limitations of the model include the assumption that untreated pleomorphic adenoma has a utility value of 1. Some patients may experience psychological distress and/or aesthetic changes owing to tumor growth that affect their quality of life. Another key limitation was the inability to assign value ranges for the time-dependent variables. This limitation is most important for the variable describing the probability of transformation from pleomorphic adenoma to carcinoma ex pleomorphic adenoma because there remains considerable uncertainty about the true value for this variable owing to a paucity of natural history studies. Another limitation was the need to estimate the utility value for the health state of permanent facial nerve injury after total parotidectomy and adjuvant radiotherapy for carcinoma ex pleomorphic adenoma. However, wherever possible, the model was constructed to account for these limitations. By performing a PSA with a uniform distribution about the mean for all utility values, for instance, the model accounted for considerable uncertainty. In addition, by using a Markov model with time-dependent variables, such as death from unrelated causes and probability of transformation, the model was able to precisely evaluate the association of both age and time since diagnosis. In addition, the model accounted for the false-negative FNA results given the imperfect diagnostic accuracy of FNA. Consequently, the model realistically depicts how treatment decisions are made with imperfect information.

The present study showed that, when balancing the risk of malignant transformation with the risk of complications from parotidectomy, observation may be reasonable in elderly patients. Furthermore, by age 70 years, observation and surgery may be associated with similar outcomes. Selection of the appropriate treatment option (parotidectomy vs observation) should involve shared decision-making between the patient and the surgeon. Development of specific observation protocols may also help streamline and optimize patient outcomes. Future studies evaluating the cost-effectiveness of each strategy may help inform health systems and public health policy decisions regarding the management of pleomorphic adenomas in elderly patients.

Accepted for Publication: July 21, 2020.

Corresponding Author: Uchechukwu C. Megwalu, MD, MPH, Department of Otolaryngology–Head and Neck Surgery, Stanford University School of Medicine, 801 Welch Rd, Stanford, CA 94305 (umegwalu@stanford.edu).

Published Online: September 24, 2020. doi:10.1001/jamaoto.2020.2944

Author Contributions: Dr Kligerman had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Kligerman, Ayoub, Megwalu.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Kligerman, Ayoub.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Kligerman, Jin.

Administrative, technical, or material support: Ayoub, Megwalu.

Supervision: Megwalu.

Conflict of Interest Disclosures: None reported.