Treatment and Outcomes in Differentiated Thyroid Cancer: A Retrospective Exploration in Three UK Centres That Provide Different Advice on Low Iodine Diets

Background International guidelines on the treatment of differentiated thyroid cancers promote low iodine diets (LID) before radioiodine remnant ablation. Evidence that the LID ultimately improves treatment success is inconsistent. This study aimed to determine if there is a difference in ablation success rates according to provision of advice to follow a LID. Methods Retrospective study of patients with differentiated thyroid cancer treated with total thyroidectomy and radioiodine remnant ablation between 01/01/2015 and 31/12/2016 in 3 centres advising: no LID (C1: n=108); LID for 1-week before (C2: n=50); LID for 2-weeks before and 48 hours (C3: n=59) after RRA. Response to treatment was determined by adapted American Thyroid Association Dynamic Risk Stratication Score, stratied as excellent, indeterminate, or incomplete response. Results In total, 217 patients were included in the analysis. We found differences in preparation for radioiodine remnant ablation and in the assessment of outcomes between centres. Furthermore, although there was little difference in staging between centres there was a difference in the percentage of patients receiving 1.1GBq vs higher administered activities (15% in C1, 22% in C2 and 44% in C3, p<0.001). An excellent response was recorded for 49% in C1, 48% in C2 and 36% in C3 (p=0.61). With C1 as reference group, the odds ratios (OR) for an excellent response were C2 OR: 0.96 (95% CI 0.46,2.00) and C3 OR: 0.62 (95% CI 0.29,1.30), p=0.40. Conclusions We found no evidence that advice to follow a low iodine diet for 1 or 2 weeks before radioiodine remnant ablation impacts on ablation success but differences between centres means the results should be regarded as exploratory. There is no immediate need to change practice regarding the LID, but a prospective multi-centre study with a more homogenous approach to patient management or a randomised controlled trial will provide more denitive recommendations.


Background
Globally, over 255,000 new cases of thyroid cancer were recorded in 2017, with 70% in women (1). A third of cases live in higher-income countries, but the age standardised incidence rate is increasing fastest in middleincome countries. In 2017 in the UK there were 1068 new cases of thyroid cancer in men and 2756 in women compared with 310 in men and 815 in women in 1993 (2). Increasing incidence of thyroid cancer is a global trend, although mortality is, in most regions, falling (1).
Differentiated thyroid cancer (DTC) accounts for 95% of cases and has 10-year survival rates of 80-90%(3). DTC is usually treated with total thyroidectomy, which may be followed by treatment with radioactive iodine ( 131 I) to destroy residual thyroid or cancerous tissue. 'Radioiodine remnant ablation' (RRA) denotes destruction of residual thyroid tissue whereas 'radioiodine therapy' refers to 131 I administration for known residual local or metastatic disease (3). The decision on whether to proceed with 131 I, and the administered activity, is determined, in part, by staging and histological features. Depending on these factors, patients are classi ed into three groups 1) de nite indications for RRA (large tumour size or extensive extra thyroidal extension or distant metastases) 2) uncertain indications 3) no indication (small tumour size, no extra thyroidal extension, minimally invasive, welldifferentiated). The decision to proceed or not is individualised for patients in group 2, taking into account factors such as tumour size, histology, lymph node involvement and full extent of any extra thyroidal extension and after discussion of risks and bene ts of RRA with the patient(3).
Thyroid stimulating hormone (TSH) is raised to facilitate uptake of 131 I. This used to be achieved through thyroid hormone withdrawal (THW) but, since approval in the mid-2000s, recombinant human TSH (rhTSH) is routinely used (4). Non-radioactive iodine competes with 131 I for uptake, which may reduce e cacy of RRA (5). In relation to iodine status at the time of RRA, the European Association of Nuclear Medicine Therapy Committee recommends aiming for a urine iodine concentration (UIC) of < 100mcg/l and postponing RRA if UIC > 150-200mcg/l (5). International guidelines state that iodinated contrast agents should be avoided in the 8 weeks before RRA, to avoid RRA if amiodarone has been taken within the previous 12 months and, based on expert opinion, recommend a low iodine diet (LID) for 1 to 2 weeks prior to RRA (3)(4)(5)(6). LID advice reduces iodine intake and iodine status in patients with high iodine intake (7,8). However, no randomised controlled trials have been conducted on whether LIDs improve ablation success, and evidence from cohort and retrospective studies is mixed (9,10), particularly in countries with low dietary iodine, such as the UK (11).
Despite acknowledging the lack of evidence for LIDs, British Thyroid Association guidelines advise an LID prior to RRA(3). In the UK, many treatment centres use the UK Low Iodine Diet Working Group diet sheet and advise restricting sh, seafood, seaweed, dairy produce, eggs, iodised salt (although salt is not routinely iodised in the UK) and a wide variety of foods containing these ingredients as well as nutritional supplements and cough mixture (12). However, different treatment centres give differing advice, particularly regarding length of dietary restriction and when to restart usual diet (13). The effect of LID advice on treatment success in the UK has not been examined. This is a retrospective study of patients with DTC comparing response to RRA in three UK centres that give different LID advice prior to RRA. The aim is to determine if there is a difference in response according to provision of advice to follow a LID.

Sites
We recruited three UK treatment centres that routinely administer RRA for thyroid cancer giving different LID advice. Centre 1 (C1) did not advise patients to follow a LID, Centre 2 (C2) advised a LID for one week prior to RRA and Centre 3 (C3) advised a LID for two weeks prior to RRA and 48 hours post.

Patients
Clinical records of patients diagnosed with DTC who underwent RRA for thyroid cancer at the three sites between 01/01/2015 and 31/12/2016 were reviewed. Eligible patients were those treated with one stage or two stage total thyroidectomy and diagnosed with papillary or follicular thyroid cancer by histology. Included patients were those with stages N0 to N1b and treated with any administered activity who had been prepared with either rhTSH or THW protocol. Records were excluded if patients had distant metastases, were experiencing recurrence, or had two radioiodine therapies planned from the outset due to residual disease or local involvement of surrounding tissue. Patients diagnosed during 2016 but treated with RRA during 2017 were excluded. Identi cation of thyroid cancer during treatment for other cancers was not an exclusion criterion.

Results
A total of 289 records were screened, 217 were eligible (C1 = 108, C2 = 50, C3 = 59). Distant metastasis (n = 30 [42%]) and recurrence (n = 18 [25%]) were the most common reasons for ineligibility. Differences were observed between centres regarding preparation for RRA and tests used to assess treatment success. C2 and C3 prepared patients using rhTSH, assessed success with neck US and measured Tg using Beckman Tg-IA for all cases (17).
In contrast only 9 patients in C1 were assessed through neck US and Tg-IA measures (Roche E170). However, C3 was less likely to use supressed Tg measures than C1 and C2. These differences were unexpected and the initial objective to determine if there was a difference in treatment success according to provision of LID advice was Outcome Response to treatment was determined by a three tier system advised by UK Guidelines for the Management of Thyroid Cancer(3), adapted from the American Thyroid Association Dynamic Risk Strati cation Score (ATA Score) (14) which uses results from neck ultrasound (US) and thyroglobulin (Tg) tests at 9 to 12 months post ablation. Responses are classi ed as excellent response (low risk of recurrence), indeterminate response (intermediate risk of recurrence) and incomplete response (high risk of recurrence). Where neck US was not available, results from diagnostic whole-body scan (DxWBS) was used (3). Thyroglobulin is secreted by normal and cancerous thyroid cells and detectable Tg post-thyroidectomy indicates remnant thyroid tissue or the presence of residual or recurrent tumour (3). Tg may be measured using immunometric assays (Tg-IA) or radioimmunoassays (Tg-RIA). Tg antibodies (TgAb) can interfere with the measurement of Tg, producing falsely low results and masking disease. Both Tg-IA and Tg-RIA are subject to false negatives, but no TgAb detection cut-off exists that eliminates false negatives and false positives (15). Patients with no evidence of disease on neck US or DxWBS, supressed or stimulated Tg < 1ug/l and any detectable TgAb were classi ed as an indeterminate response, in strict accordance with the ATA Score.

Data extraction
Anonymised data on demographics, details of surgery, tumour, node and metastasis (TNM) staging, 131 I dose, post ablation assessment and evaluation of treatment success at 9 to 12 months post RRA were extracted. Data extraction was undertaken and checked by clinical staff or trained coders.

Statistical analysis
Overall comparisons were made between centres with respect to patients' ages, gender, grouped TNM stage (7th edition)(16), 131 I administered activity, and ATA Score, using chi-squared tests or one-way ANOVA as appropriate. These were followed by within-centre comparisons of these variables between patients with excellent, indeterminate, or incomplete outcomes. Across all centres, numbers of people experiencing an incomplete response were small so indeterminate/incomplete responses were combined. The main analysis used logistic regression to compare centres in respect of excellent response, adjusting successively for age and sex and then administered activity of 131 I and TNM stage; coding for response was incomplete/indeterminate = 0 vs excellent = 1, therefore higher odds were associated with better outcomes. A sensitivity analysis was undertaken with a graded response, using ordered logistic regression (proportional odds), coding the response as incomplete = 1, indeterminate = 2, excellent = 3. For both the main analysis and sensitivity analysis C1 was the comparison group. hampered by this. The results provided below are therefore considered exploratory in nature, and any differences observed between centres may not be due to differences in LID advice. Table 1 shows between-centre comparisons in respect of age, gender, TNM stage, 131 I administered activity and the results of post ablation whole body scans (PaWBS). C1 had fewer patients with TNM stages III/IV than the other centres, but there was little statistical evidence of a difference between centres (p = 0.14). A difference was observed between 131 I administered activity above 1.1GBq; fewer patients in C3 received higher activities (P < 0.001). PaWBS indicated that four (4%) patients in C1 and four (7%) in C3 showed uptake in cervical nodes.

Descriptive comparison between centres
Within-centre differences in response failed to show overall differences between subgroups (except possibly for C3 (TNM stage and administered activity), but numbers were small) (Supplementary tables).

Discussion
This retrospective study of patients with DTC compared treatment response rates at 9 to 12 months post RRA from three centres in the UK that gave different LID advice prior to RRA. We found differences between centres in both preparation and determination of treatment success, which was not expected at the outset of this study.
Comparison between centres was thus hampered, and ndings should be treated as exploratory.
We found no evidence for a difference in ablation success at 9 to 12 months assessment between the centres, adjusting for age, sex, TNM stage and administered activity 131 I. Treatment response for C3 (2-week LID) appeared worse, with a 40% reduction in the odds of experiencing an excellent response compared to that of experiencing an excellent response at C1 (no LID). However, the con dence interval was wide and included no difference and when an outcome is common the odds ratio will tend to overestimate underlying risk. There was no difference in effect size between C1 (no LID advice) and C2 (1-week LID).
Only 45% of included patients were classi ed as experiencing an excellent response at 9 to 12 months post RRA, which appears low. Ablation success was determined using the ATA Score combining results of neck US or DxWBS with stimulated or supressed Tg measurements (3). Patients with TgAb were not removed from the analysis but, in accordance with the guidelines, were classi ed as having an 'indeterminate' response. However, results from DxWBS or neck US alone indicate that no abnormality was detected for 84% of patients (data not shown), which is comparable with success rates reported in other studies, including the Hi Lo trial (18).

Comparison with other LID studies
No randomised controlled trials have examined whether LID advice prior to RRA or radioiodine therapy improves treatment success rates in thyroid cancer. In 1983 Maxon et al (19) observed that use of a LID increased 131 I administered activity to the tumour during diagnostic scans (LID = 19 vs normal diet = 21), but evidence that the LID improved treatment success was lacking. Since then, retrospective examinations of patient records have been conducted to determine whether use of LID improves success rates or whether iodine status is associated with treatment success. Table 3 summarises relevant studies. Only Pluijmen et al (20) found that there was a difference in success rates between patients advised to follow a LID and those who were not. However, between studies there is variation in preparation, 131 I administered activity, how response to treatment is assessed and whether patients with detectable TgAb were excluded or not.  (21) did not observe a similar reduction. Non-randomised (7,8,23,24) and randomised (25,26) studies have shown that advice to follow an LID for one to two weeks can lower UIC to < 100mcg/l. However, iodine status is not routinely measured prior to RRA in the UK(13) and we were unable to assess whether patients reduced their iodine status or whether iodine status itself was associated with treatment success. It is also unknown whether patients in C1 reduced high iodine foods despite being given no speci c LID advice. There is readily available information online about the LID(27) which patients could have accessed. We have conducted qualitative work that suggests patients not given advice to follow a LID may still reduce iodine intake (28).

Other strengths and limitations
This is the rst study conducted in the UK to investigate whether advice to follow a LID prior to RRA affects treatment success in DTC. Ideally, the only difference in treatment across centres would have been in the dietary advice given. However, unexpectedly, there was evidence for a difference between the centres in 131 I administered activity with fewer patients in C3 having higher administered activity. The fully adjusted model corrected for activity but there were also differences in preparation and assessment methods between centres that we were not aware of when designing the study. Although we do not consider these to be factors that would substantially affect treatment success (29), this meant we were not comparing identical practices. Given that evidence from other studies seems to indicate that following a LID prior to RRA does not confer substantial bene ts in terms of outcomes, the differences between centres may have masked any small bene ts from a LID.

Recommendations for research and practice
This study forms part of a larger overall project examining advice to follow a LID prior to ablation. We have conducted a qualitative study investigating the impact of the advice on patients(28) and a survey of practice regarding the use of the LID in the UK (13). This retrospective study indicates that routinely collected data cannot be used in the UK to determine whether advice to follow an LID has an impact on treatment success due to centre-level differences on top of differences in LID advice. In the UK, it is unclear whether current LID advice successfully lowers iodine intake or iodine status and, globally, it remains unclear as to whether low iodine status improves treatment success. The aws in our retrospective study indicate that there is a need for an RCT or, given that the LID is widely used in clinical practice, a large well-controlled prospective observational study to determine 1) whether LID advice lowers iodine status and 2) whether lowering iodine status has an impact on treatment success rates.

Conclusion
We conducted this retrospective review to determine if following a low iodine diet for 1 or 2 weeks prior to RRA for differentiated thyroid cancer contributed to treatment success. In common with other retrospective studies, we found no evidence that such advice impacted on success rate. However, differences in preparation and assessment meant we were not comparing identical practices across centres and the results should be treated as exploratory. There is no immediate need to change practice regarding the LID in the centres giving different dietary advice, but there is a need for greater harmonisation in overall treatment, drawing on evidence from recent trials (30). We believe that adequately powered well-controlled prospective studies and, potentially, an RCT, are required to con rm the role of a LID prior to RRA on the outcomes of treatment.