Ablation rate after radioactive iodine therapy in patients with differentiated thyroid cancer at intermediate or high risk of recurrence: a systematic review and a meta-analysis

Purpose We performed a systematic review and a meta-analysis to investigate the successful ablation rate after radioiodine (RAI) administration in patients with differentiated thyroid cancer (DTC) at intermediate-high risk of recurrence. Methods A comprehensive literature search of the PubMed, Scopus, and Web of Science databases was conducted according to the PRISMA statement. Results The final analysis included 9 studies accounting for 3103 patients at intermediate-high risk of recurrence. In these patients, the successful ablation rates ranged from 51 to 94% with a 71% pooled successful ablation and were higher in intermediate (72%) than in high (52%)-risk patients. Despite the rigorous inclusion standards, a significant heterogeneity among the evaluated studies was observed. Higher administered RAI activities are associated with a lower successful ablation rate in the whole population and in the subgroup of high-risk patients. Furthermore, pooled recurrence rate in intermediate-risk patients achieving successful ablation was only 2% during the subsequent 6.4-year follow-up while the pooled recurrence rate was 14% in patients who did not achieve a successful ablation. Conclusion In a large sample of 3103 patients at intermediate-high risk of persistent/recurrent disease, 71% of patients achieved a successful ablation. In these intermediate-risk patients, the probability of subsequent recurrence is low and most recurrence occurred in those with already abnormal findings at the first control. Supplementary Information The online version contains supplementary material available at 10.1007/s00259-021-05440-x.


Introduction
After total thyroidectomy with or without lymph node dissection for differentiated thyroid cancer (DTC), radioiodine (RAI) may be administered for 3 main goals [1]: remnant ablation (to facilitate the detection of recurrent disease by destroying post-operative remnants of non-tumoral thyroid tissue and to permit initial staging with a whole-body scan), adjuvant therapy (to decrease the risk of recurrence by destroying suspected, but unproven persistent disease), or therapy (to treat known persistent disease). Successful ablation (SA) can be assessed some months later by an undetectable serum thyroglobulin (Tg) level on L-T4 using a sensitive method and unremarkable ndings at neck ultrasound [2][3][4].
The indication for post-operative RAI administration takes into account the American Thyroid Association (ATA) threetiered risk system [1] that classi es patients as low, intermediate or high-risk of recurrence. Whereas post-surgical RAI therapy is usually not indicated in low-risk patients, it is generally recommended in intermediate and high-risk patients. Intermediate-risk group includes patients with microscopic extra-thyroid extension, aggressive histology, vascular invasion, > 5 lymph node metastases (N1) with all N1 < 3 cm, multifocal papillary microcarcinoma with ETE or BRAFV600E mutation and RAI-avid metastatic foci in the neck on the rst whole-body RAI scan. High-risk patients show macroscopic ETE, incomplete tumor resection, biochemical or structural evidence of distant metastatic disease, any N1 > 3 cm or follicular thyroid cancer with extensive vascular invasion. Although literature offers a huge armamentarium of data on survival bene t of RAI in intermediate-high risk patients, to our knowledge a metaanalysis on SA rate in these patients [5,6] has not yet been performed. Therefore, we performed a systematic review and a meta-analysis to investigate the SA rate after RAI administration in patients with DTC at intermediate-high risk of recurrence.

Search strategy
This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (see Supplementary material for PRISMA Checklist) [7], and registered as 242409 in the PROSPERO database (University of York, UK; http://www.crd.york.ac.uk/PROSPERO/). An English literature search was performed using the PubMed and Embase databases to identify articles published from 2010 until June 2020. This search was restricted to data obtained in adults and was conducted using the following key words: "differentiated thyroid cancer" OR "DTC", "thyroid neoplasm", "prognosis", "outcome", "follow-up", "radioactive iodine therapy" OR "RAI therapy", "I-131 ablation", "thyroglobulin" OR "Tg".

Study selection
The title and abstract of potentially relevant studies were screened for appropriateness before retrieval of the full article by two reviewers (L.P. and F.V.), and disagreements were resolved by consensus.
The selected full-published reports were retrieved and the same reviewers independently performed a second-step selection based on the eligibility criteria; disagreements were resolved by consensus.
In addition, the bibliographies of retrieved articles were manually reviewed for potential additional citations.

Study eligibility and data extraction
Each study was initially identi ed considering journal, authors, and year of publication. To harmonize the predictors of interest, a study was considered eligible if all of the following criteria were met: 1) Data were available on age, gender and administered RAI activity, histopathology and extent of surgery. 2) the study presented data of adult subjects with differentiated thyroid cancer at intermediate or high risk of recurrence after RAI therapy; we excluded studies on low-risk patients only and we excluded the low-risk patients in studies that included both low and intermediate/high-risk patients [1]. Investigations that considered different risk classi cation, such as the AJCC/TNM staging system [8] were included in the nal analysis only if a detailed description of clinical and histopathological characteristics of patients was provided, allowing patient categorization with the ATA risk classi cation basis (e.g. studies only including patients having microscopic extra-thyroidal extension which are considered intermediate ATA risk or studies clearly providing data on patients with pT3-pT4 tumor or metastasis who were considered ATA high risk); 3) the study included at least 100 subjects; 4) follow-up after RAI therapy for at least 1 year; 5) the study provided data on SA after RAI therapy de ned as absence of abnormal ndings at neck ultrasonography and undetectable serum Tg in the absence of anti-Tg antibodies [2,3]. In case of multiple studies reported from the same research group, potential cohort duplication was avoided by including only the study on the largest number of patients.

Assessment of the methodological quality of studies
All studies were assessed for methodological quality using Joanna Briggs Institute (JBI) Prevalence Critical Appraisal Tool [9]. The criteria address the following issues: Ensuring a representative sample, ensuring appropriate recruitment, ensuring an adequate sample size, ensuring appropriate description and reporting of study subjects and setting, ensuring data coverage of the identi ed sample is adequate, ensuring the condition was measured reliably and objectively, ensuring appropriate statistical analysis, ensuring confounding factors/subgroups/differences are identi ed and accounted for. These questions can be answered with four possible responses: yes, no, unclear or not applicable [9]. Two reviewers (V.C. and R.G.) evaluated the risk of bias in each eligible study. and disagreements between reviewers were resolved by consensus.
The two reviewers completed the screening process independently. Disagreement in the process of answering questions was discussed until consensus was reached. A nal decision of yes (favorable scenario, "+"), no (unfavorable scenario, "−"), or unclear (mixed scenario, "+/−") was made by the reviewers after systematic discussion. If the answers to all the signal problems were "yes", a low risk of bias was attributed to the study; if the answers to all the signal problems had one or more "no" or "unclear" values, an unclear risk of bias was used; if the answers to all the signal problems contained at least one "no" but no "yes" answers, a high risk of bias was attributed.

Statistical analysis
We used a systematic analytic approach to compute the pooled SA rate after RAI therapy in patients with DTC from all eligible studies. Metaprop was used to perform meta-analysis of proportions close to or at the margins, 0% or 100%. Metaprop pools proportions and presents a weighted subgroup and overall pooled estimates with inversevariance weights obtained from a random-effect model [11]. Heterogeneity of the included studies was examined by using the I-squared (I 2 ) statistic, to re ect the percentage of total variation across studies [11], assigning adjectives of low, moderate, and high to I 2 values of 25%, 50%, and 75%. According to the Cochrane handbook, I 2 > 50% re ects a substantial heterogeneity [12]. Therefore, a random effect model is used to combine data in the meta-analysis [13]. The possibility of publication bias in the present study was examined by using Egger's test [14]. Publication bias was graphically examined by the funnel plot and also formally assessed with the regression test of asymmetry described by Egger et al [14]. When statistical heterogeneity was substantial, meta-regression analysis was performed to assess if study-level variables such as age, gender and RAI administered activity were associated with pooled SA rate. When it was feasible, we evaluated the pooled SA rate separately in DTC patients at intermediate and high risk of recurrence. We also evaluated the pooled recurrent disease rate in DTC patients who achieved SA when data were available. All analyses were performed using Stata, version 15.1 (StataCorp, College Station, TX). Two-sided P values < 0.05 were considered statistically signi cant.

Study selection
The complete literature search is presented in Fig. 1. The initial search identi ed 819 potentially eligible citations. Among these, 119 were identi ed as duplicates and thus removed, leaving 700 records. The reviewers, after the evaluation of their titles and abstracts of these studies removed 655 citations. Then, each investigator blindly reviewed the full text of the remaining 45 articles, and 36 articles were excluded. The remaining 9 articles included a total of 6675 patients, 3103 of whom at intermediate-high risk of recurrence and were the basis of the present metaanalysis.

Characteristics of the included studies
Demographic data and clinical characteristics of the 3103 patients included in the meta-analysis are detailed in Table 1 [15][16][17][18][19][20][21][22][23]. Study sample size ranged from 152 to 627 subjects. Mean age was 47 ± 1 years, with the proportion of women ranging from 66-96%. Mean follow-up was 4.7 ± 1.5 years. Seven of the 9 studies [15,[18][19][20][21][22][23] referred to the ATA risk classi cation and this permitted to de ne separate risk categories. One study [16] referred to AJCC/TNM staging system without mentioning the ATA risk classi cation. However, this investigation only included patients with small tumor size, extra-thyroidal extension, and no neck lymph node metastasis who were considered at intermediate risk in the separate analysis. In another study [17] only referring to the AJCC/TNM staging system, patients with pT3-pT4 tumors or metastasis were de ned as high-risk. Finally, although one investigation [22] referred to ATA risk categories, SA rates were only available when considering both intermediate and high-risk patients together and this study was only considered for the overall analysis. Values are expressed as mean ± standard deviation or median or as number (percentage) of subjects. TT total thyroidectomy, RAI radioactive iodine Assessment of the methodological quality of the included studies Figure 2 summarizes the quality assessment of the 9 included studies using Critical Appraisal tools for use in JBI Systematic Reviews. The risk of bias was considered low overall. The domain that showed an unclear risk of bias was "study subjects and setting". This result could be due to slightly different descriptions of patient characteristics among studies.

Successful ablation rate in the overall population
The SA rate reported in the 9 studies ranged from 51-94% (Fig. 3), the pooled SA rate was 71% (95% con dence interval, CI, 59-83) and the heterogeneity was 98.58% (P < 0.001). The funnel plot indicates no publication bias (P = 0.16) among these studies (see Supplementary material, Figure S1). At meta-regression analysis, there was a negative association between SA rate and mean RAI activity administered (coe cient = -0.002; standard error = 0.27; t = 3.51; P < 0.05). As shown in Fig. 4, higher RAI activities administered for more clinical severe disease was associated with a lower SA rate. With a mean RAI activity of 83.7 mCi, the SA rate was 93.7%, while with a mean RAI activity of 151 mCi the SA rate was 57.5%.

Successful ablation rate in intermediate risk patients
The SA rate in intermediate risk patients was reported in 7 studies [16,17,19,[20][21][22]24], including 1939 patients. It ranged from 51-94%, the pooled SA rate was 72% (95% CI 59-86) and the heterogeneity was 98.4% (P < 0.001) (see Supplementary material, Figure S2). The funnel plot indicates no publication bias (P = 0.14) among these studies (see Supplementary material, Figure S3). At meta-regression analysis did not nd any association between SA and the variables analyzed. Of note, the 350 patients at intermediate and high-risk evaluated by Avram et al. [22] were not included in any subgroup analysis because data could not be evaluated in separate categories but only as a single group.

Successful ablation rate in high-risk patients
The SA rate according to separate high-risk patients' category has been reported in only 4 studies [15,17,18,23] including 814 patients and ranged from 18-78%. The pooled SA rate was 52% (95% CI 31-74) and the heterogeneity was 96.5% (P < 0.001) (see Supplementary material, Figure S4). The funnel plot indicates no publication bias (P = 0.08) among these studies (see Supplementary material, Figure S5). At meta-regression analysis, a negative association between lower SA rate and higher mean RAI activity administered was found (coe cient = -0.001; standard error = 0.47; t = 4.02; P < 0.05) (see Supplementary material, Figure S6). For example, with a mean RAI activity of 105 mCi the SA rate was 84%, while with a mean RAI activity of 138 mCi the SA rate was 53%.
Recurrent disease rate at in patients who achieved successful ablation A late follow-up (mean 6.4 ± 1.4 years) was available in 4 studies [16,19,20,21] including 656 intermediate-risk patients ( Table 2). In one study [21] recurrence was de ned as new evidence of biochemical or structural disease after any disease-free period, in the other 3 studies [16,19,20] recurrence was cytologically or histologically proven. Recurrence rate after SA ranged from 0-7% and the pooled recurrent disease rate was 2% (95% CI 0-5) (Fig. 5). The heterogeneity was 78.5% (P < 0.001) and the funnel plot indicated no publication bias (P = 0.08) among these studies (see Supplementary material, Figure S7). At meta-regression analysis no association between recurrent disease rate and all variables analyzed was found. In contrast, recurrences were observed in 18/121 intermediate risk patients who did not achieve a SA at the rst control, with a pooled rate of 14%. Values are expressed as number of subjects

Discussion
The present investigation refers to 9 studies accounting for 3103 patients at intermediate or high-risk of recurrence.
The main nding of our meta-analysis is a 71% pooled SA rate after RAI treatment and this was indeed higher in intermediate (72%) than in high (52%) risk patients. This is well below the reported SA rates achieved in low-risk patients [24,25], even with the use of relatively low activities (30mCi) following rhTSH [25,26].
Higher administered activities in high-risk patients were associated with a lower SA rates, suggesting that more extensive disease observed in these high-risk patients are also less sensitive to RAI treatment. This might be related to an impairment in iodine metabolism as documented in patients with a BRAFV600E and TERT mutations [28].
Despite the rigorous inclusion standards, a signi cant heterogeneity among the evaluated studies was observed. The presence of heterogeneity in a meta-analysis is an expected issue [29]. This result could be explained by different speci c end-points considered within each single investigation. It should be also taken into account that, as shown in Table 1, the overall population assessed in every single study referred to a more extensive patient group with heterogeneous clinical characteristics, while the evaluated population included in our analysis matched strict criteria suggesting that the investigations included in the current meta-analysis suffer from different degrees of bias across studies. Nevertheless, the meta-analysis method remains a powerful option to interpret multiple data coming from literature.
A late follow-up was available in 4 studies [16,19,20,21] only considering intermediate risk patients. In patients achieving SA at early follow-up there was a pooled recurrence rate of 2%. In contrast, the pooled recurrence rate was 14% in patients who did not achieve a SA at early follow-up. These results are in close agreement with the prognostic value of the ATA classi cation, with the majority of recurrences occurring in patients with some abnormal ndings at the rst control [27]. Yet, data on high-risk patients were not available to perform a similar analysis in this category.

Conclusion
The present meta-analysis demonstrates that overall SA rate after RAI treatment in patients at intermediate-high risk of recurrence was 71% and it was higher in intermediate (72%) than in high (52%) risk patients. Further, pooled recurrence rate in patients achieving SA was only 2% further highlighting the very low probability of recurrence of disease once a SA has been obtained after RAI therapy. These data indeed underline the importance of using the concept of ongoing risk assessment that allows to modify the individual prognosis and follow-up strategy according to the results of each control.

Declarations
Compliance with ethical standards Con ict of interest 29. Higgins JP. Commentary: Heterogeneity in meta-analysis should be expected and appropriately quanti ed. Int J Epidemiol. 2008;37:1158-60. Figure 1 PRISMA owchart illustrating the study selection process.

Figure 2
Methodological quality of the included studies assessed with JBI tool for risk of bias and applicability concerns. The green circle represents low risk of bias, the yellow circle unclear risk of bias and the red circle high risk of bias.

Figure 3
Forest plot for the successful ablation rate after RAI therapy. Horizontal lines represent 95% con dence interval of the point estimates. The diamond represents the pooled estimate (size of the diamond = 95% con dence interval). The solid vertical line represents the reference of no increased risk and the dashed vertical line represents the overall point estimate.

Figure 4
Meta-regression analysis between SA and the mean RAI activity administered. Bubble size for each study is proportional to the inverse of the variance.

Figure 5
Forest plot for the persistent/recurrent disease rate at late follow-up in patients who achieved SA. Horizontal lines represent 95% con dence interval of the point estimates. The diamond represents the pooled estimate (size of the diamond = 95% con dence interval). The solid vertical line represents the reference of no increased risk and the dashed vertical line represents the overall point estimate.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.