1 INTRODUCTION

Graves’ disease (GD) is the most common cause of persistent hyperthyroidism in adults worldwide. It is estimated that 0.5% of men and 3% of women will develop the disease during their lifetime.1 GD is an autoimmune disorder caused by complex interactions between both environmental and genetic factors that cause an inappropriately high synthesis and secretion of thyroid hormone by the thyroid.2, 3 Known clinical presenting symptoms may include heart palpitations, heat intolerance, weight loss, tremulousness, anxiety and ophthalmopathy (Graves' thyroid disease), which can be found in up to 50% of patients.4 GD can be managed by antithyroid drugs (ATD), radioactive iodine (RAI) or thyroidectomy. A surgical option has the advantage of immediate and definite treatment of hyperthyroidism without the toxicity associated with life-long antithyroid medications or RAI. Even though studies demonstrate more than a 95% cure rate, surgery remains the least chosen option for treatment.5-11 Studies worldwide have shown a significant 80%–90% decrease in the number of surgeries for GD throughout recent decades.12, 13 Common reasons include the concern for a neck scar, permanent hypothyroidism and possible complications such as hypoparathyroidism and recurrent laryngeal injury. With the development of remote access thyroidectomies and intraoperative nerve monitoring of the recurrent laryngeal nerve (RLN), combined with current knowledge of possible risks associated with RAI or failure of ATDs, revaluation of the benefit to harm ratio of surgery in the treatment of GD is warranted.

The aim of this updated review is to discuss possible reasons for the low proportion of surgery in the treatment of GD, emphasizing an evidence-based approach to clinicians’ preferences for surgical referrals and surgical indications for GD.

2 METHODS

We searched Medline and Google Scholar using various combinations of terms: ‘Graves’ disease’ ‘quotes’, ‘indication’, ‘preferences’, ‘referral’, ‘cost-effectiveness’, ‘costs’, ‘thyroidectomy’, ‘surgery’ and ‘complications’. We gave preference to high-quality observational studies, randomized controlled trials, systematic reviews published in the past 10 years and global presentation of different series and findings. Articles were selected and approved by all authors. No ethical approval was required due to the nature of the article.

To further support the evidence base of this review, a systematic literature review was conducted by the first author (O.C.) according to the Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines (http://www.prisma-statement.org/). Outlined literature search and results are presented in Supporting Information 1 (checklist) and Figures S1–S3. For ‘Indications for Surgery in the Treatment of GD’ section, the search included the words ‘Graves disease’ AND (indication OR referral) AND (surgery OR thyroidectomy); For ‘Referral for Surgery in the Treatment of GD’ section, the search included the words ‘Graves disease’ AND (preference OR survey) and for the Healthcare Costs section the search included the words ‘Graves disease’ AND (‘cost effectiveness’ OR ‘cost’). Studies were eligible if they were in the English language, original articles and published in peer-reviewed journals. Exclusion criteria included publications without full text available and duplicated publications. All publications were initially screened by titles and/or abstracts. For articles deemed relevant, assessment of eligibility of the full-text articles was done by evaluating the main findings and supplementary data according to the inclusion and exclusion criteria. Measures were presented in numbers and percentages when applicable, presented in tables and text in appropriate sections.

3 POSSIBLE SURGERIES IN THE TREATMENT OF GD

Surgeries indicated for GD include either total thyroidectomy, which intend to remove all of the thyroid tissue; bilateral subtotal thyroidectomy; or hemithyroidectomy plus contralateral subtotal thyroidectomy, also known as Dunhill procedure.14, 15 A 2015 Cochrane review was conducted to assess the optimal surgical technique for GD and GD ophthalmopathy (GO).15 Total thyroidectomy was found to be more effective at preventing recurrent hyperthyroidism, even though there were greater risks for potentially causing hypoparathyroidism. The review found a neutral effect of the various surgical techniques on permanent recurrent laryngeal nerve palsy and on the deterioration of GO. Yet, total thyroidectomy has become the procedure of choice by many institutions as it is more effective and just as safe as a subtotal thyroidectomy.16-19

4 INDICATIONS FOR SURGERY IN THE TREATMENT OF GD

The American Thyroid Association (ATA) guidelines, as well as the American Association of Endocrine Surgeons (AAES), recommend surgery for the following indications: (1) Women planning a pregnancy in <6 months provided thyroid hormone levels are normal; (2) symptomatic compression or the presence of a large goitre (≥80 g); (3) relatively low uptake of RAI; (4) suspicious thyroid malignancy and (5) patients with moderate to severe active ophthalmopathy.3, 20 In addition to the aforementioned indications, the AAES Guidelines have included social considerations (infants, small children in the home, unable to follow RAI safety measures), airway compromise necessitating urgent resolution and central neck pathology requiring surgery (primary hyperparathyroidism, large size [>4 cm] or multiple benign thyroid nodules).20 In the 2018 European Thyroid Association guideline for the management of GD, surgery is indicated as first-line therapy only following the personal decision of the patients, while recognized indications as a thyroid nodule, goitre or active ophthalmopathy are stated for second line alone.21

When the indications for surgery in GD patients are reviewed, great heterogenicity was observed between the different indications. A literature review was conducted to evaluate the proportion of different known indications among different operated GD patients’ cohorts worldwide. A summary of indications for surgical treatments in different large case series representing different centres worldwide is presented in Table 1. Selection of the series was based on the availability of data (indications were not published in all series), quality of data, date of publication and country of origin. Overall, great heterogenicity was observed between the different indications. Patients’ preferences ranged 3.4%–49.3% while an indication of a goitre ranged 7.2%–65%. From the patients’ perspective, Grodski et al.30 found that 30% of the patients who elected surgery as definitive management did so in the absence of a specific, known indication, further reflecting the difficulties to establish uniform criteria and indications for surgery. The most commonly stated reasons were personal preference to avoid radioiodine treatment, desire to pursue a career in the armed forces and cosmesis for a prominent goitre. A similar percentage of patients selecting surgery without a formal indication was noted in another small case series (n = 32).31 Moreover, Jin et al.32 have shown that surgery is significantly more chosen among uninsured patients in the United States, implying that ‘preference’ may be a result of social and economic burden. Yip et al.23 were able to show that the surgical indications for GD thyroidectomy significantly changed with two decades investigated and compared, further stressing the need for a continuous reevaluation of surgical indications as they were shown to be both time and place dependent. The distribution of the indications may impact surgical outcomes and possible complications, as will be discussed in the ‘complications’ section.

Table 1. Summary of indications for thyroidectomy for Graves' disease (GD) Authors, year N Country Patient Preference Other treatment failures (%)a Goitre (%)b Malignancy (%)c Orbitopathy (%) Pregnancy (%) Other (%) Witte et al. (2000)14 150 Germany NA 55.4 25.3 NA 19.3 NA Lal et al. (2005)16 103 US 26d 15 18 24 20 Seed Young age—14 Efremidou et al. (2009)17 247e Greece NA 47 42e NA 11 Wilhelm et al. (2010)18 136 US 49.3 16.9 30.9 7.3 2.2 NA Tamatea et al. (2012)22 117 New Zealand 3.4 58 NA 1.7 12 NA Yip et al. (2012)23,f 346 China 47 3.2 7.2 6 21.7 5.2 Young age—4 Phitayakorn et al. (2013)24 300 US 13 35.6 10.3 17.6 20 1.6 Hyperparathyroidism—1.6 Shinall et al. (2013)25 165 US 38 21.6 22 NA 17 6 Bojic et al. (2015)26 1432 Serbia NA 15 65 10 10 NA Kwon et al. (2016)27,g 189 Korea 13.7 16.4 7.9 62.4 NA NA Swegal et al. (2017)28 52 US NA 10 11 2 58 6 Desire to avoid RAI—13 Cipolla et al. (2019)29,h 594 Italy 36.2 57.2 30.8 NA 14.6 2.3 Abbreviation: NA, not addressed in the article. 5 REFERRAL FOR SURGERY IN THE TREATMENT OF GD

The management of GD depends upon institutional practice, the surgeon concerns and the pre-eminence of endocrinologists or nuclear physicians. Similarly, referral for surgery seems to be affected by cultural, economic, and general preference either by the patient or the physician. Fundamental differences exist between Europe and the United States regarding the treatment of GD. RAI is being used more frequently, and surgery is less commonly used in European centres.33 Therefore, even ‘accepted’ guidelines, including indication categories, remain controversial and cannot be regarded as universal.33 GD is managed by three different disciplines—endocrinology, nuclear medicine, and surgery (Otolaryngology—Head and Neck Surgery or a General Surgeon with an endocrine interest). The choice for treatment for primary, persistent, or relapsing disease is usually taken with the attending endocrinologist, while a direct discussion between the patient and the surgeon remains limited. Hence, the management preference of the physician may have a great impact on the patient's decision. In a 2011 multinational survey of 730 members of the Endocrine Society, ATA, and American Association of Clinical Endocrinologists (AACE), Burch et al.5 found that surgery was the preferred management option for uncomplicated GD in only 0.6%, compared with 45% of RAI and 53.9% of ATD. Following surveys from Europe, Middle East and Oceania have confirmed surgery as the least desirable first-line therapy, not reaching 5% of cases.5-10 Furthermore, even in cases of relapses, surgery was found least recommended compared with ATD and RAI, which was chosen for 20%–25% of the patients,5-8 reaching 33% in others.10 In a recent study by Brito et al.,11 of the 4661 patients with GD, 60% received ATD (n = 2817), 33% received RAI (n = 1549), and only 6% received surgery (n = 295). The success for surgical treatment was 99%, 93% for RAI (treatment failure was defined as a requirement of subsequent treatment in the 2 years after their initial treatment [either RAI or surgery]), and only 50% for ATD (treatment failure was defined as either receiving RAI/surgery or a break of >90 days in ATD use followed by reinitiation of ATD at any time during follow-up). This demonstrates a correlation between the surveys and the treatments implemented.

A possible reason for the dominance of ATD in the primary and relapse treatment of GD is the chance to avoid permanent hypothyroidism. Many patients may prefer antithyroid drugs due to understandable concerns regarding permanent hypothyroidism, weight gain, or reduced quality of life (QoL) on levothyroxine. The use of the latter has been shown to result in high frequency of abnormal thyroid function, iatrogenic thyrotoxicosis and hypothyroidism.34 This may also be a wish of the endocrinologist to ‘spare’ the patients this inevitable consequence of surgery. While a recent review by El Kawkgi et al.34 has supported the use of prolonged ATD treatment over definitive treatment, Sjolin et al.35 investigated the long-term follow-up for GD patients (6–10 years). They found only half of the patients who started ATD for initial treatment would avoid definitive treatment and only 40% would not need thyroid medication to achieve a euthyroid state. Moreover, studies focusing on QoL have all reported improved QoL following surgery.36-38 Scerrino et al.36 compared the QoL results of 57 patients undergoing surgery for GD to a control group of patients using ATD. Patients following surgery demonstrated significant improvement in all fields assessed in the questionnaire. Compared with control, the patients who underwent surgery had a better QoL than those who received prolonged medical therapy. Hamilton et al.37 that used the Glasgow Benefit Inventory to assess QoL, have also found improvement in QoL. Results on QoL outcomes between the treatment modalities have been conflicting. While some have shown that GD long-term QoL was not affected by treatment modality,39 a recent large-scale population-based study in Sweden has shown that RAI treatment resulted in lower QoL compared with surgery and ATD, while no difference was found between surgery and ATD.40 Thus, it seems that surgery offers the highest cure rates, with at least non-inferior QoL compared with RAI and ATD, yet is significantly underutilized compared with the alternatives.

In a recent, multicenter study conducted in the Netherlands, van Kinschot et al.41 performed a discrete choice experiment regarding the three management options for primary and first relapse GD presentation. The study included 286 patients and 61 clinicians. Considerations for each treatment arm were based on type of treatment, remission rate, severe side effects, permanent voice change and hypocalcemia. Remission rate was the most important determinant of treatment choice. ATDs were the most preferred treatment option by both patients and clinicians. In contrast to clinicians, patients preferred treatment with surgery over RAI, which exhibits the difference in approach to treatment and the role of the primary consultant has in the determination of treatment.

The reasons why surgery remains the least chosen option by endocrinologists have not been investigated up to date. A better understanding of the causes would improve the dialogue between different disciplines and may improve our ability to provide patients with an accurate, contemporary knowledge of the risk and benefits of surgery. Obvious reasons include a finite number of operative resources for treating a common medical problem, the invasiveness of surgery, and the fact that not all patients want surgery or are eligible for thyroid surgery. A shortage of high-volume trained thyroid surgeons may also decrease willingness to refer patients for surgery, just as low availability of nuclear medicine would increase the rates of surgery. Common disadvantages of surgery cited in reviews include neck scar, hypocalcemia (transient and permanent) and higher costs.42-44 In the following paragraphs, we will review recent evidence-based literature on the concerns and address recent developments in these fields.

6 POSSIBLE CLINICAL CONCERNS REGARDING SURGICAL MANAGEMENT OF GD 6.1 Neck scar

Up until recently, neck scar was considered an inevitable consequence of thyroid surgery. Recent developments in remote access thyroidectomies have allowed for scarless procedures for benign and malignant disease, including GD patients.45 Kwon et al.27 compared 44 GD patients who underwent bilateral axillo-breast robotic thyroidectomy to 145 GD patients who underwent open thyroidectomy and found comparable success and complications rate.

In a retrospective review of 97 GD patients where open thyroidectomy (n = 49) was compared to transoral endoscopic thyroidectomy vestibular approach (n = 46), similar outcomes were found in terms of recurrent laryngeal nerve injury, hypoparathyroidism and blood loss demonstrating the safety of remote access surgery. Significantly improved pain control was noted for the transoral endoscopic thyroidectomy vestibular approach group compared with the open thyroidectomy group.46 Therefore, accumulating evidence suggests that for patients who are concerned with having a cervical scar, remote access surgery is a safe approach for definitively addressing hyperthyroidism associated with GD. However, considerations such as an overly enlarged thyroid gland, the relation between critical structures in the neck, as well as preoperative hyperthyroidism state limit candidacy for this approach,47 which may be more suitable for selected cases. Moreover, availability of the aforementioned remote access approaches has yet to become universal, hence neck scars may be unavoidable. Surgical considerations such as concealment of the scar within the natural skin creases can increase patients' aesthetic satisfaction.48

The significance of a cervical scar is variable depending on age, gender and cultural considerations. While some reported scar to be the most commonly reported adverse event,49 with an adverse effect on the QoL,50 others reported no impact on QoL.51, 52 This was further verified by significant differences in results reported on patients’ willingness to undergo scarless procedures.53, 54 Women, younger patients, patients with a tendency to heal with hypertrophic scars or keloid and patients with benign thyroid pathology have all been reported to be more concerned with a neck scar and to favour remote access thyroidectomy.54, 55 Feasibility and safety of transoral thyroidectomy approach among the paediatric population was recently shown,56 allowing for future investigations on the safety of transoral approach among paediatric GD patients. As remote access thyroidectomies are expanding worldwide, studies on the expected candidacy rate among GD patients should be taken.

6.2 Complications of surgery

A 2020 nationwide population-based study by Brito et al.11 on the efficacy and safety of all treatment modalities reported a 24% adverse effect rate in the surgical arm, compared with 12% and 6% in the ATD and RAI arms, respectively. Surgical adverse effects included haemorrhage, seroma, infection, vocal cord paralysis, dysphonia and hypoparathyroidism. For comparison, RAI's adverse effects included radiation-induced thyroiditis and new GO, while ATD's adverse effects were all major such as agranulocytosis, vasculitis, and necrosis of the liver. This reflects the impact of definitions on the published complications (or adverse effect) rate, which in turn may cause a bias when deciding on surgery. To keep in line with complications related to RAI and ATDs, three major complications should be addressed: RLN injury, hypoparathyroidism and haematoma (which can result in reoperation). The latter prevalence is 1%–1.5% and is excluded from this section,14, 57 though a recent meta-analysis found an odds ratio of 1.58 (1.09–2.31) for GD surgery.58

There is a lack of uniformity in the definition of reported complications: RLN injury varies in the definition for transient versus permanent and methods of diagnosis (mandatory observation of the vocal cords vs. clinical complaints) between studies57, 59, 60; hypocalcemia was shown to have a significant variation in both definition (minimum level of calcium) and duration of symptoms which correlates with its subsequent incidence.61 This, in turn, undermines the reliability of systematic reviews analysing these complications. Therefore, it is pivotal for surgical reports to reach a consensus on the definition for these two major complications.

This section reviews recent reports on complications for total thyroidectomy performed for GD. Previous large retrospective case series have found a higher rate of complications for GD compared with other techniques to remove the thyroid.59, 62 However, these studies were exposed to significant bias (proportion of goitres among GD patients), lacked proper matching in many cases, and significant confounders were present (e.g., in the work of Thomusch et al.,59 the weight of specimen was significantly higher among lower volume centres, which were found to be an independent risk for complications) may have impacted reported results. This section aims to review matched series and recent prospective data.

Postoperative hypocalcemia after total thyroidectomy is one of the most common complications that can be temporary (less than 6 months) or permanent (more than 6 months).63 Welch and McHenry60 compared outcomes in 111 GD patients to 283 nontoxic goitres. No statistical difference was found in the complication rates except for higher transient hypocalcemia in the GD group, but this was no longer significant after matching for thyroid weight. For transient hypocalcemia, Pesce et al.64 did not find a significant difference between 55 randomly selected GD patients to randomized, non-matched controls in postoperative Day 1 hypocalcemia and 30 days following surgery. Kwon et al.62 demonstrated a higher rate of transient complications (both hypocalcemia and RLN injury) among GD patients, yet no difference was found in permanent complications. However, the thyroid weight of the GD group was significantly larger than the control group (approximately three times larger), exposing results for significant potential bias. As a higher rate of temporary complications was reported for goitre thyroidectomy regardless of GD,65 interpretations of non-matched for goitre or gland weight of GD patients should be done with caution, especially when considering that a goitre may be the indication for surgery in up to 65% of patients26 (Table 1).

The National German Prospective StuDoQ/Thyroid Registry published complications rate for 1808 thyroidectomies for GD, of which 93.4% were total thyroidectomies and 1.4% were minimally invasive (including invasive video-assisted thyroidectomy robotic-assisted transaxillary thyroidectomy). Intraoperative nerve monitoring (IONM) was used in 98.9% of the cases. The rate of unilateral transient RLN injury was 3.9%,57 compared with a 2.74% rate found on a meta-analysis for IONM in thyroid surgeries.66 This comparable rate may stress the importance of IONM in GD patients.

6.3 Possible measures to decrease complications rate 6.3.1 Referral for high-volume centres

Numerous studies, from both Europe and the United States, have found that high-volume thyroid centres have a lower complication rate, yet the cutoff for high-volume ranged from 25 to 50 surgeries per year.57, 67 The authors agree that GD patients should only be treated in dedicated centres that have a high-volume endocrine practice.

6.3.2 Preparation for surgery

Potassium iodide (given as SSKI or as Lugol's solution) given 7–10 days before surgery has been shown to reduce thyroid blood flow, vascularity and intraoperative blood loss.68, 69 However, its actual impact on complication rate remains debated,70, 71 and several publications questioned the true value of such preparation.61, 72 Supplementation of oral calcium, vitamin D or both preoperatively may reduce the risk of postoperative hypocalcemia following thyroidectomy, as shown by a meta-analysis of risk factors for postoperative hypocalcemia.73 Both preoperative vitamin D deficiency and GD were found to be risk factors for postoperative hypocalcemia.73 A matched case-control study by Oltmann et al.74 found that nontreated (treatment of 1 g of oral calcium three times a day, 2 weeks before surgery) GD had significantly higher rates of both biochemical and symptomatic hypocalcemia compared with euthyroid and treated GD patients.

6.3.3 Postoperative administration of calcium and PTH

Providing oral calcium calcitriol for patients at risk (all GD patients) has become standard of care, and was published in a recent dedicated statement by the ATA.75 In their randomized, Palermo et al.76 have shown that patients with intact PTH lower than 10 pg/ml 4 h after thyroidectomy can receive 20 mcg of teriparatide (PTH 1–34) since diagnosis and every 12 h till discharge resulted in a significant reduction of postoperative hypocalcemia among the treatment arm.76

6.4 Healthcare costs

Traditionally, the cost of thyroidectomy has been another argument against the choice of surgery for primary treatment or for ATD resistant GD.42-44 Our literature search found three articles on the cost-effectiveness of thyroidectomy in the treatment of GD, summarized in Table 2. In the United States, a more recent study by In et al.78 conducted a comprehensive cost-effective analysis for patients failing to achieve a euthyroid state after 18 months of ATD (relapse/persistent disease). Total thyroidectomy was found to be more cost-effective than RAI or lifelong ATD in these patients, when the cost of surgery was less than $19,300. Zanocco et al.77 compared cost-effectiveness of total/subtotal thyroidectomy with ATDs and RAI, and found total thyroidectomy to be more costly and less effective than RAI and subtotal thyroidectomy strategies. In their analysis for primary GD among the UK and Australian patients, Donovan et al.79 compared the cost-effectiveness of RAI, ATDs and total thyroidectomy. They concluded that RAI was the least expensive while total thyroidectomy was unlikely to be cost-effective. Great differences in the studies’ models prevent a direct comparison between the results (Table 2). In the absence of similar definitions for complication, length of follow-up and its quality, a simple conclusion that surgery carries high costs or is not cost-effective, as written in leading reviews,42-44 should be done with caution, as the studies in which it was based on may have been exposed to substantial bias. Moreover, differences in thyroidectomy costs between countries were shown to considerably affect cost-analysis considerations in other scenarios.80

Table 2. Summary of cost-effectiveness studies addressing treatment options for Graves’ disease (GD) Article Primary versus relapse/persistence GD Country Outcomes Cost calculations Modalities examined General factors included Surgery specific factors included Results Zanocco et al. (2012)77 Primary US QALY Medicare charge and reimbursement data

ATD