Comparative Effectiveness of Therapies for Graves' Hyperthyroidism: A Systematic Review and Network Meta-Analysis

Abstract

Context:

Several treatment options are available for Graves' disease (GD), including antithyroid drugs (ATDs), radioactive iodine (RAI), and thyroidectomy.

Objective:

The primary outcome was to determine the relapse rates of various treatment options. The secondary outcome was to present data regarding adverse effects of ATDs.

Data Sources:

We searched multiple databases through March 2012.

Study Selection:

Eligible studies were randomized clinical trials and comparative cohort studies in adults that included 2 or more treatment options for GD.

Data Extraction:

Two reviewers independently selected studies, appraised study quality, extracted outcome data, and determined adverse effect profiles.

Data Synthesis:

We found 8 studies with 1402 patients from 5 continents. Mean follow-up duration in months was: ATDs, 57; RAI, 64; and surgery, 59. Studies were at moderate to high risk of bias. Network meta-analysis suggested higher relapse rates with ATDs (52.7%; 352 of 667) than RAI (15%, 46 of 304) (odds ratio = 6.25; 95% confidence interval, 2.40–16.67) and with ATDs than surgery (10%; 39 of 387) (odds ratio = 9.09; 95% confidence interval, 4.65–19.23). There was no significant difference in relapse between RAI and surgery. Examination of 31 cohort studies identified adverse effects of ATDs in 692 of 5136 (13%) patients. These were more common with methimazole, mainly owing to dermatological complications, whereas hepatic effects were more common with propylthiouracil use.

Conclusion:

We confirm the relatively high relapse rate of ATD therapy in comparison with RAI or surgery, along with a significant side effect profile for these drugs. These data can inform discussion between physicians and patients regarding the choice of therapy for GD. The limited quality of the evidence in the literature underlines the need for future randomized clinical trials in this area.

Graves' disease (GD) is an autoimmune condition defined by overproduction of thyroid hormones due to unregulated stimulation of the thyroid by circulating TSH receptor antibodies (1, 2). It is the most common form of hyperthyroidism in the United States (3), and if left untreated results in increased morbidity and mortality (4) mainly due to cardiovascular (atrial fibrillation, heart failure, pulmonary hypertension, angina pectoris, and stroke) (5) and skeletal (osteoporosis) complications. Therefore, timely management of overt hyperthyroidism is of utmost importance. In addition, GD has a negative long-term influence on the quality of life, due to either the disease process itself or its treatment (4).

Treatment options for GD are aimed at inducing permanent hypothyroidism (radioactive iodine [RAI] therapy and thyroidectomy) followed by thyroid hormone replacement or restoring euthyroidism while awaiting resolution of the autoimmune process and disease remission (antithyroid drugs [ATDs]—methimazole [MMI], propylthiouracil [PTU], carbimazole [CBZ]). Beta blockers are used in combination with any of the above therapies to ameliorate the symptoms of hyperthyroidism. Selection of therapy is challenging for both patient and physician because each of the 3 modalities has been established as an effective treatment strategy with unique individual features. The recently published guidelines for the management of hyperthyroidism stress the importance of active discussion between patient and physician regarding the risks, benefits, and logistics of the various treatment options, taking into consideration the values and preferences of the patient (6). A better understanding of the favorable and unfavorable characteristics of each option will facilitate this discussion and lead to high-quality shared decision-making. We therefore performed the first systematic review and network analysis of randomized and observational studies to summarize the available evidence in adults with GD undergoing treatment and to determine the quality of evidence available in the literature supporting the efficacy of the available therapies. In addition, we report the adverse effect profile of ATDs extracted from cohort studies evaluating the use of ATDs alone.

Materials and Methods

This systematic review complies with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) statement (7).

Study eligibility

Studies eligible for this review were randomized clinical trials (RCTs) and observational cohort studies. There was no language restriction. We searched publications that involved adult patients (18 y and older) with the diagnosis of GD and compared treatments with at least two of the following interventions: 1) ATDs (MMI, CBZ, or PTU); 2) RAI; and 3) thyroidectomy.

Data sources and search strategies

We conducted a comprehensive search of databases from database inception to March 2012. The following databases were included: Ovid Medline In-Process & Other Non-Indexed Citations, Ovid MEDLINE, Ovid EMBASE, Ovid Cochrane Database of Systematic Reviews, Ovid Cochrane Central Register of Controlled Trials, and Scopus. The search strategy was jointly designed by an experienced librarian (L.J.P.) and two of the investigators (J.P.B., V.S.). Controlled vocabulary supplemented with keywords was used to search for the concepts GD, thyrotoxicosis, RAI, thyroidectomy, MMI, CBZ, PTU, and ATD effectiveness and adverse effects, limited to controlled trials and cohort studies. The detailed research strategy is available upon request.

Study selection

Two reviewers (J.P.B., V.S.) working independently screened titles and abstracts. Full-text articles obtained from initial screening were retrieved for second-stage screening. Chance-adjusted inter-reviewer agreement was substantial (κ statistic = 0.9), and discrepancies were resolved through discussion with R.S.B. and M.N.S.

Data extraction

The main outcome of interest was relapse of GD, which was defined as the recurrence of hyperthyroidism after at least 12 months of treatment with ATDs, one dose of RAI, or first thyroidectomy. Disease recurrence was chosen as the main outcome because its avoidance would represent to the physician effective therapy and “cured” disease. The two reviewers (J.P.B., V.S.) independently reviewed each eligible study (each study included in the analysis was reviewed in duplicate). A pilot-tested form was used to extract the following information: demographic information, goiter size, type of treatment, treatment duration for ATD, follow-up duration, type of surgery (total, subtotal, or near-total thyroidectomy), number and dose of RAI treatment, and number of patients with relapse. In addition, we reviewed all cohort studies of ATD therapy identified in our initial search of the literature to identify those listing adverse effects.

Quality assessment

Two reviewers (J.P.B., V.S.) working in duplicate assessed the methodological quality of studies selected for the meta-analyses. The Newcastle-Ottawa scale (8) was used for observational studies, and elements from the Cochrane risk of bias tool (9), including allocation concealment, blinding, and loss of follow-up, were utilized for RCTs.

Data synthesis

Random effects meta-analysis models were constructed to pool odds ratios (ORs) from direct comparisons using the method of DerSimonian and Laird (10), with the estimate of heterogeneity being taken from the Mantel-Haenszel model. We used the I² statistic and Cochran's Q test to assess heterogeneity across individual studies. We then conducted network meta-analyses to combine direct and indirect evidence, using Lumley's generalized linear mixed models (11). A network meta-analysis approach provides estimates of effects sizes for all possible pairwise comparisons, whether or not they have been compared in previous trials (12). We evaluated the agreements of indirect comparisons, also called “incoherence,” and incorporated incoherence in the calculation of confidence interval (CI) of the pooled OR. Data from direct evidence and network meta-analyses were presented together and compared for consistency. Statistical analyses were completed using STATA version 12 (StataCorp) and R version 2.15.0 (R Foundation for Statistical Computing).

Results

Eight studies were eligible for the main outcome analysis: 1 RCT, and 7 comparative cohort studies (Figure 1). These studies involved a total of 1402 patients (667 in the ATD group, 314 patients in the RAI group, and 419 patients in the surgical group). The dose of RAI used was included in 6 of 8 studies and was a mean of 8.5 mCi (range, 6.8 to 12.6). Mean follow-up duration was 57, 64, and 59 months for ATDs, RAI, and surgery, respectively. The overall relapse rate was 52.7% (352 of 667) for ATD, 15% (46 of 304) for RAI, and 10% (39 of 387) for surgery. The studies were conducted in the following countries: United States, 1; Chile, 1; Ethiopia, 1; Ireland, 2; Sweden, 2; and Turkey, 1 (Table 1).

Figure 1.

Study selection process.

Table 1.

Detailed Description of Included Studies

Study	Country	ATDs				RAI				Surgery
		Total, n	Relapse, n (%)	Duration of Therapy, mo	Follow-Up Duration, mo	Total, n	Relapse, n (%)	Mean Dose of RAI, mCi	Follow-Up Duration, mo	Total, n	Relapse, n (%)	Follow- Up Duration, mo	Type of Surgery
Alizadeh, 1979 (21)	United States	NR	NR	NR	NR	11	0 (0)	8.56	36.6	11	4 (36)	43	NR
Sugrue, 1980 (22)	Ireland	272	153 (56)	24	60	43	1 (2)	7.45	60	266	16 (6)	60	NR
Berglund, 1991 (23)	Sweden	83	36 (43)	20	21	106	5 (5)	NR	36	23	2 (9)	17	Subtotal thyroidectomy
Mengistu, 1992 (24)	Ethiopia	47	3 (6)	24	11	NR	NR	NR	NR	6	1 (17)	36	NR
Torring, (old cohort) 1996 (13)	Sweden	35	12 (34)	18	30	39	15 (38)	6.8	48	37	3 (37)	42	Subtotal thyroidectomy
Torring (young cohort) 1996 (13)	Sweden	25	10 (40)	18	30	NA	NA	NA	NA	28	1 (4)	48	Subtotal thyroidectomy
Pineda, 1998 (25)	Chile	88	58 (66)	6–48	NR	70	6 (9)	12.6	NR	37	2 (5)	NR	NR
Leary, 1999 (26)	Ireland	74	50 (68)	18–24	NR	38	10 (26)	7.5	NR	5	1 (20)	NR	Subtotal thyroidectomy
Tutuncu, 2006 (27)	Turkey	43	15 (35)	18	52	7	0 (0)	8.1	75	6	0 (0)	40	Subtotal thyroidectomy

Abbreviations: NR, not reported; NA, not applicable.

Methodological quality

The 7 observational studies included in our analysis were of low quality and subject to high risk of bias (Table 2). The primary limitation of these studies was a lack of comparability of cohorts regarding goiter size, gender, age, and degree of hyperthyroidism. In the only RCT (13), intention to treat analysis was not performed but allocation concealment was properly described. Due to the limited number of studies included, we were unable to test publication bias (14).

Table 2.

Quality Assessment of Included Studies

First Author, Year (Ref)	Study Design	Representative of Cohorts	Comparability of the Cohorts				Assessment of Outcome	Sufficient Follow-Up for Outcomes to Occur
			Goiter Size	Gender	Age	Degree of Hyperthyroidism
Alizadeh, 1979 (21)	Historical cohort	No	Unclear	Unclear	Unclear	Yes	Records linkage	Yes
Sugrue, 1980 (22)	Historical cohort	No	Unclear	Unclear	Unclear	Unclear	Records linkage	Yes
Berglund, 1991 (23)	Historical cohort	No	Unclear	No	No	Unclear	Records linkage	Yes
Mengistu, 1992 (24)	Concurrent cohort	Yes	Unclear	Unclear	Unclear	Unclear	Records linkage	Yes
Pineda, 1998 (25)	Historical cohort	Yes	No	Yes	No	Yes	Records linkage	Yes
Leary, 1999 (26)	Historical cohort	Yes	Unclear	Unclear	Unclear	Unclear	Records linkage	Yes
Tutuncu, 2006 (27)	Concurrent cohort	No	Unclear	Unclear	Unclear	Unclear	Records linkage	Yes
Torring, 1996 (13)	Randomized clinical trial	NA	Blinding	Intention to treat analysis		Allocation concealment	Funding
			No	No		Yes	Nonprofit

Abbreviation: NA, not available.

Network analysis

Eight studies were included in the analysis (Table 2). For the primary outcome, both direct and indirect estimates suggested higher relapse rates with ATDs than RAI (OR = 6.25; 95% CI, 2.40–16.67; I² = 81%) and with ATDs than surgery (OR = 9.09; 95% CI, 4.65–19.23; I² = 42%) (Table 3). There was no significant difference in relapse between RAI and surgery. We did not find large incoherence in the network (ω < 0.001). Forest plots depicting the results of random effects meta-analysis are presented in Figure 2.

Table 3.

Estimates of Direct and Indirect Comparisons

Comparisons	Direct			Network
	OR	95% CI	P Value	OR	95% CI	P Value
ATD:RAI	6.33	2.40, 16.67	<.01	6.13	3.27, 11.49	.01
ATD:surgery	9.43	4.65, 19.23	<.01	9.80	5.13, 20.00	.01
RAI:surgery	1.53	0.64, 3.65	.34	1.60	0.81, 3.18	.27

Figure 2.

Random effect meta-analysis of the included studies comparing risk relapse rates among the 3 interventions. Vertical lines indicate no risk difference; squares and horizontal lines indicate OR and associated 95% CI for each study; diamonds indicate pooled OR.

A search of observational studies describing adverse effects of ATD therapy yielded 31 reports in which 660 of 5136 (13%) patients experienced events (MMI, 505 of 3969 patients [14.9%]; and PTU, 68 of 983 patients [6.9%]; OR = 2.3; 95% CI, 1.8- 3.06). The predominant adverse effect of MMI was rash (6%; 239 of 3969). Hepatic involvement was more common with PTU (2.7%; 27 of 983) (Figure 3). Twelve events among 184 patients were not specified as to the type of ATD used.

Figure 3.

Adverse effects documented in studies reviewed, with percentage of total attributed to individual adverse effects.

Discussion

We conducted a systemic review and network analysis of the three therapeutic options for the treatment of Graves' hyperthyroidism to inform the patient-physician discussion regarding choice of therapy and to determine the quality of evidence available in the literature supporting the efficacy of these therapies. Clinical experience demonstrates that all two modalities successfully eliminate hyperthyroidism. However, the comparative effectiveness of these treatments, as characterized by relapse rates, is demonstrated in the literature only by low-quality evidence. As expected, we found ATDs to have a higher relapse rate than either RAI therapy or thyroidectomy, with the latter two therapies having no significant difference in relapse rates. It is important to note that the most “effective” therapy from the physician's perspective both eliminates hyperthyroidism and prevents its recurrence. In fact, the goal of both RAI and thyroidectomy is to render the patient hypothyroid such that lifelong thyroid hormone replacement is necessary. In contrast, patients may well desire a treatment that has the potential to allow their thyroid to resume normal functioning. In this sense, ATDs carry an advantage over surgery and RAI that was not captured in our comparative effectiveness data.

The mean dose of RAI was 8.5 mCi (range, 6.8 to 12.6). Dosing was mainly based on goiter size and RAI uptake. In addition, 1 study incorporated age, gender, and clinical severity in dose determinations, whereas in another, dosing was empirical. The relatively low doses of RAI used in these studies may be explained by the inclusion of patients from iodine-deficient countries that tend to have higher RAI uptake, by advice given to some patients to follow a low-iodine diet before RAI therapy or by the goal of RAI in some practices being to achieve euthyroidism rather than to render the patient hypothyroid as recommended in recent guidelines (6).

Full understanding of the consequences of the two treatments includes a discussion between physician and patient regarding the adverse effects of each modality. Unfortunately, adverse effects were not reported consistently across the included studies. Therefore, we selected other observational studies of ATDs to estimate the adverse effects of these agents and used landmark studies to discuss the adverse effects of RAI therapy and thyroidectomy. We found a significant rate of adverse effects for ATDs (13%) reported in 31 observational studies involving medical therapy alone. The predominant adverse effect of MMI was rash (6%), and that of PTU was hepatic involvement (2.7%) (Figure 3). Individual studies in the literature suggest that the most common adverse effect of RAI therapy is new or worsened Graves' ophthalmopathy, which may develop in 15–33% of patients, particularly smokers (15, 16). Another complication of RAI therapy is radiation thyroiditis, occurring in about 1% of the patients (17). Potential complications of thyroidectomy (either total or subtotal) include hypoparathyroidism (temporary, 22.5%; permanent, 1.8%), recurrent laryngeal nerve injury (temporary, 3.3%; permanent, 1.24%), and immediate postoperative bleeding (1%) (18). Four among the 8 studies in our analysis specified the type of surgery performed, which was subtotal thyroidectomy in every instance. Because total thyroidectomy has become the standard of care in the last decade (6), it would be most meaningful to discuss with the patient the potential complications of total thyroidectomy alone. The same meta-analysis (18) also reported the complications of total thyroidectomy to be hypoparathyroidism (temporary, 32.5%; and permanent, 2.6%), recurrent laryngeal nerve injury (temporary, 3.43%; and permanent, 1.46%), and immediate postoperative bleeding (<1%). Because these complications are operator dependent, individual surgeons should discuss their own complication rates while counseling their patients.

Our study has limitations, perhaps the primary one being the increased risk of bias inherent in the included cohort studies. We identified only a single RCT; the remaining studies were observational. In addition, the number of included studies and enrolled patients was small, affecting the precision of our estimates. We also had limited information concerning the size of thyroid gland, severity of hyperthyroidism, or other predefined factors (dose of ATD and extent of thyroid surgery) needed to conduct subgroup analysis and explore the effect of important covariates. Our strengths stem from the clinical relevancy of the question at hand, preplanned analysis, extensive literature search, reproducible duplicate data extraction and collection, and greater generalizability of the results because our studies originated from five different countries and six continents reflecting various races and ethnic population groups. Our network model also had low heterogeneity and appeared consistent.

Our search of the literature did not identify other existing network meta-analyses evaluating the comparative effectiveness of the three treatment options for Graves' hyperthyroidism in adults. The most recent systematic review focusing on the best definitive treatment for GD reported surgery to be 3.44 times more likely to be successful than RAI (P < .001) (19). In that review, total thyroidectomy was 95.45 times more successful than RAI, whereas subtotal thyroidectomy was only 2.33 times more successful than RAI (P < .001). The search strategy differed from our own in that it was limited to studies published between 2001 and 2011, it did not compare ATDs, and a network analysis was not performed.

The recently published guidelines for management of hyperthyroidism by the American Thyroid Association and the American Association of Clinical Endocrinologists have simplified the clinical decision-making process (6). Factors that favor a particular modality as treatment in the presence of limitations to the other therapies have been well detailed. For example, a 44-year-old executive concerned about insomnia with limited time for future visits will value a treatment that offers the fastest resolution of symptoms and might opt for thyroid surgery performed by a high-volume surgeon. A 35-year-old opera singer concerned about the possible postoperative damage to her voice and the need for lifelong thyroid replacement may opt for ATD therapy. Finally, the guidelines weave in 2 essential concepts: 1) the importance of careful discussion between patient and physician; and 2) the need to include the values and preferences of the patient in shared decision-making. The treating physician is advised to discuss the available treatment options in light of the individual patient's medical status and preferences. Issues to be discussed include advantages, drawbacks, potential adverse effects, expected time to recovery, local availability of expertise, financial implications, time away from work, impact on social life in the immediate post-treatment period, and longer-term quality of life issues. Once the patient understands the information and is able to participate in the decision-making process, the physician's best clinical judgment, coupled with the values and preferences of the patient, will allow for optimum treatment selection. In some instances, patients' hyperthyroid state will impair rational decision-making (20). In these cases, a good approach might be to treat the patient with ATDs until euthyroid and discuss the option of long-term ATD therapy or to pursue 1 of the definitive treatment strategies.

Decisions made by patients reflect information presented to them by their treating physician and perhaps also information obtained from the internet and family/friends. In a recent international survey of endocrinologists, the preferred treatment for an illustrative case of uncomplicated GD was ATDs (53.9%), RAI therapy (45%), and thyroidectomy (0.7%) (28).

These data suggest that individual physicians carry their own biases concerning optimum therapy and that there is no general consensus among physicians. Given this, it is particularly important for the physician to present the treatment options fairly and involve the patient in the decision-making process. An opportunity lies in the development of a state-of-the-art risk communication tool for GD that could be used by the physician to elicit the patient's values and preferences and by the patient for clarification of the risks and benefits of each option.

In conclusion, high-quality evidence quantifying and comparing the relapse rates and adverse effects of the various treatment options for GD is unavailable in the literature and warrants study in the setting of RCTs. The objective information gained from this meta-analysis can be used to facilitate discussions between physician and patient concerning optimum choice of therapy for Graves' hyperthyroidism.