Hemithyroidectomy versus total thyroidectomy for well differentiated T1–2 N0 thyroid cancer: systematic review and meta‐analysis

Background Evidence for limiting the extent of surgery in patients with low‐risk thyroid cancer is lacking. Methods A systematic search was performed according to the PRISMA and MOOSE guidelines to assess the effect of total thyroidectomy (TT) with or without radioactive iodine (RAI) treatment versus hemithyroidectomy (HT) on recurrence and overall mortality in patients with differentiated (papillary or follicular) T1–2 N0 thyroid cancer. PubMed, Embase and Cochrane databases were searched, and two authors independently assessed the articles. Results A total of ten eligible articles were identified. All were observational cohort series, representing a total of 23 134 patients, of which 17 699 were available for meta‐analysis. Six studies included patients who had TT followed by RAI treatment. The pooled recurrence rate after TT ± RAI and HT was 2·3 and 2·8 per cent respectively (odds ratio (OR) 1·12, 95 per cent c.i. 0·82 to 1·53; P = 0·48). The pooled 20‐year overall survival rate after TT ± RAI was 96·8 per cent, compared with 97·4 per cent for HT (OR 1·30, 0·71 to 2·37; P = 0·40). Overall, higher complication rates were found in the TT ± RAI group. Conclusion Recurrence rates after HT for treatment of well differentiated T1–2 N0 thyroid cancer were similar to those after TT ± RAI, with a lower incidence of treatment‐related complications.


Introduction
The incidence of thyroid cancer has been increasing over the past three decades, affecting both sexes and all ages 1,2 . The standard of care for well differentiated thyroid cancer is total thyroidectomy (TT) and adjuvant radioactive iodine (RAI) treatment, resulting in reported 10-year overall survival rates of 96-98 per cent 3 . Apart from a lifelong need for thyroid hormone supplementation, complications such as recurrent laryngeal nerve damage, bleeding and hypoparathyroidism contribute to treatment-related morbidity and decreased quality of life 4,5 .
Large non-comparative national database cohort studies [6][7][8][9] , however, show neither survival benefit nor difference in recurrence rates between TT followed by RAI and hemithyroidectomy (HT) for well differentiated, low-risk thyroid cancer. These low recurrence rates and good long-term survival explain an increased interest in de-escalation of treatment for these patients 10 . Recently, the American Thyroid Association 11 recommended consideration of HT for selected patients with low-risk, well differentiated thyroid cancer.
In selected patients, less extensive surgery for subcentimetre papillary thyroid microcarcinoma (PTMC) is supported by data from large databases 12,13 , but it is unclear whether HT will result in similar treatment outcomes in patients with low-risk differentiated thyroid cancers larger than 1 cm. This systematic review and meta-analysis Additional records identified through other sources n = 0 investigated HT as the definitive treatment for patients with low-risk, well differentiated (papillary or follicular), node-negative thyroid cancer of 4 cm or less (T1-2 N0), with a particular focus on recurrence and overall survival.

Methods
The study was registered in the International Prospective Register of Systematic Reviews (PROSPERO) (registration number CRD42018115971 After removal of duplicates, articles were initially screened by title and abstract to exclude irrelevant reports. Of the remaining articles, full texts were screened for eligibility by two authors. Reference lists of relevant articles were searched. Any discrepancy regarding article selection was resolved by consensus.

Inclusion and exclusion criteria
Any English-language original reports comparing HT with TT ± RAI for the treatment of 0-4-cm, well differentiated, low-risk thyroid cancer were included. Low risk was defined as node-negative (N0), T1 and T2 cancers 16 . Studies describing recurrent disease only, those including patients under 18 years of age, letters, reviews, editorials, case series with fewer than ten patients, and conference abstracts were excluded.

Data extraction
The following characteristics were extracted from the included studies: patient demographics (number of patients, age and sex), pathology, treatment characteristics, tumour size and definition of recurrence. The primary endpoint was recurrence rate and time to recurrence. Secondary endpoints were overall survival and perioperative morbidity, including laryngeal nerve damage, hypoparathyroidism and hypocalcaemia.

Statistical analysis
For the meta-analysis, outcome data for recurrence and overall survival were pooled using Review Manager version 5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark) and presented as forest plots. Heterogeneity was assessed by calculating the I 2 index. For low heterogeneity (I 2 below 25 per cent), a fixed-effect model was used for meta-analysis. For intermediate heterogeneity (I 2 ranging from 25 to 75 per cent), a random-effects model was used for meta-analysis.
I 2 above 75 per cent was considered substantial, and no meta-analysis was performed. Where available, data from matched-pair analysis of low-risk patients were used for the meta-analysis.

Quality assessment and risk of bias
The Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool 17 was used to score the overall risk of bias. Funnel plots were made to assess publication bias.

Results
The initial search resulted in a total of 7203 records. After removal of 2080 duplicates and screening of all titles and abstracts, a total of 68 papers were assessed in full text for inclusion, after which 58 were excluded, resulting in a total of ten studies 18-27 for analysis (Fig. 1). All ten articles described observational cohorts, representing a total of 23 134 patients, predominantly women: 19 069 women (82⋅4 per cent) and 4065 men (17⋅6 per cent). Some 17 699 patients with low-risk, well differentiated thyroid cancer were included in the meta-analysis. No prospective randomized studies were available ( Table 1). The reported duration of follow-up ranged from 57⋅3 to 141⋅6 months. To assess outcome, five studies 20,21,23,26,27 performed a matched-pair analysis based on patient characteristics. Patient characteristics are summarized in Table 2. The median age ranged from 43 to 49 years. From seven studies 19,20,[22][23][24][25][26] that explicitly reported histopathological characteristics, 13 412 patients with papillary thyroid cancer (PTC) and 40 with follicular thyroid cancer (FTC) were included. One study 18 reported on patients with PTC and FTC, but did not provide separate data on histopathological characteristics for those in the low-risk treatment group, and two studies 21,27 failed to report histopathological characteristics. Tumour size ranged from less than 0⋅5 to 1⋅93 cm. RAI treatment following TT surgery was reported in six studies [21][22][23][24]26,27 and varied from 71⋅7 to 93⋅9 per cent of the cohorts.

Complications
Complications were reported in four studies 21,23,25,26 ( Table 3). Overall, higher complication rates were found in the TT ± RAI group. Bleeding was reported in two studies 23,26 . In the HT group, bleeding rates were 0 per cent 26 and 0⋅7 per cent 23 , compared with rates of 0 per cent 26 (P = 1⋅0) and 1⋅0 per cent 23 (P = 0⋅6) in the TT ± RAI group. Regarding laryngeal nerve injury, TT ± RAI had a higher rate than HT, with an incidence from 0 per cent 21,26 to 1⋅6 per cent 25 for HT and from 0⋅6 per cent 23 to 9⋅4 per cent 25 for TT ± RAI. Permanent hypoparathyroidism was reported in three studies 21,23,26 . Although no patients with hypoparathyroidism were reported following HT, in the TT ± RAI groups the rate Kim et al. 22 Song et al. 27 Choi   ranged from 1⋅7 per cent 23 (P < 0⋅001) to 7⋅7 per cent 21 .

Quality assessment and risk of bias
Quality of included studies according to the ROBINS-I tool is presented in Table S5 (supporting information). The overall quality of the included studies was considered moderate to critical, based mainly on the observational nature of included studies. Funnel plots showed an even distribution of studies over possible outcomes, suggesting that publication bias was minimal (Figs S2-S5, supporting information).

Discussion
This systematic review with meta-analysis has shown that, for patients with low-risk, well differentiated thyroid cancer, low recurrence rates and high survival can be achieved with both HT and TT ± RAI treatment. Reported recurrence and 20-year survival rates were similar. These data correspond to survival rates from large national databases 12,13 . The updated American Thyroid Association guideline 11 recommending HT for selected patients with 1-4-cm well differentiated PTC is supported by the results of the present meta-analysis. As the present results indicated that HT was a non-inferior alternative to TT ± RAI for well differentiated T1-2 N0 thyroid cancer, possible complications of both treatments, such as recurrent nerve damage, hypoparathyroidism and hypothyroidism, need to be taken into account. Hypoparathyroidism, with rates ranging from 1⋅7 to 7⋅7 per cent, was seen exclusively in the TT ± RAI group. These rates are comparable with those reported for temporary and permanent postoperative hypoparathyroidism after TT, which have been described with a prevalence of 1⋅6-53⋅6 and 0⋅2-9⋅3 per cent respectively 28,29 . Hypoparathyroidism after thyroidectomy has been shown to correlate with a decrease in overall survival, even in patients who had surgery for benign thyroid disease 30 . In addition, with HT a considerable proportion of patients could be prevented from having to take lifelong oral thyroid hormone substitution. Although after TT there is a 100 per cent risk of hypothyroidism, and supplementation is always needed, a small number of patients do need thyroxine supplements after HT 31 .
The follow-up strategy is also important. Clearly, active follow-up is warranted after HT so that, if recurrence or metastasis is identified, completion thyroidectomy can performed followed by RAI treatment if necessary. There are no guidelines outlining the duration and method of such follow-up.
There are limitations when interpreting the results of this meta-analysis. The absence of any randomized comparison is important, and the selection biases inherent in cohort studies are acknowledged. Importantly, it has to be noted that a large proportion of included patients in both HT and TT ± RAI groups had PTMC, limiting the evidence on treatment with HT for tumours sized 1-4 cm. Rates of recurrence and overall survival were, nevertheless, comparable for HT and TT ± RAI in the patients with 1-4-cm tumours. Two of the included studies 18,24 involved patients from a considerable time ago, which may have resulted in an underestimation of the clinical N and M status as modern ultrasound and CT scans were not available at that time. Despite expecting that this might negatively affect the oncological outcome of patients who had HT, there was still no difference in oncological outcome in the pooled results of the meta-analysis. Some of the studies included both patients with PTC and those with FTC 18,19,24 , and the prognosis for FTC depends mainly on the degree of invasiveness rather than primary tumour size. Unfortunately, data from studies reporting on both PTC and FTC were not presented separately. Additional limitations include the different definitions of both recurrence and complications used in the studies, and only two studies reported survival data.

Disclosure
The authors declare no conflict of interest.