Analysis and interpretability of machine learning models to classify thyroid disease.

Thyroid disease classification plays a crucial role in early diagnosis and effective treatment of thyroid disorders. Machine learning (ML) techniques have demonstrated remarkable potential in this domain, offering accurate and efficient diagnostic tools. Most of the real-life datasets have imbalanced characteristics that hamper the overall performance of the classifiers. Existing data balancing techniques process the whole dataset at a time that sometimes causes overfitting and underfitting. However, the complexity of some ML models, often referred to as "black boxes," raises concerns about their interpretability and clinical applicability. This paper presents a comprehensive study focused on the analysis and interpretability of various ML models for classifying thyroid diseases. In our work, we first applied a new data-balancing mechanism using a clustering technique and then analyzed the performance of different ML algorithms. To address the interpretability challenge, we explored techniques for model explanation and feature importance analysis using eXplainable Artificial Intelligence (XAI) tools globally as well as locally. Finally, the XAI results are validated with the domain experts. Experimental results have shown that our proposed mechanism is efficient in diagnosing thyroid disease and can explain the models effectively. The findings can contribute to bridging the gap between adopting advanced ML techniques and the clinical requirements of transparency and accountability in diagnostic decision-making.

Automatic differentiation of thyroid scintigram by deep convolutional neural network: a dual center study.

BACKGROUND: 99mTc-pertechnetate thyroid scintigraphy is a valid complementary avenue for evaluating thyroid disease in the clinic, the image feature of thyroid scintigram is relatively simple but the interpretation still has a moderate consistency among physicians. Thus, we aimed to develop an artificial intelligence (AI) system to automatically classify the four patterns of thyroid scintigram. METHODS: We collected 3087 thyroid scintigrams from center 1 to construct the training dataset (n = 2468) and internal validating dataset (n = 619), and another 302 cases from center 2 as external validating datasets. Four pre-trained neural networks that included ResNet50, DenseNet169, InceptionV3, and InceptionResNetV2 were implemented to construct AI models. The models were trained separately with transfer learning. We evaluated each model's performance with metrics as following: accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), recall, precision, and F1-score. RESULTS: The overall accuracy of four pre-trained neural networks in classifying four common uptake patterns of thyroid scintigrams all exceeded 90%, and the InceptionV3 stands out from others. It reached the highest performance with an overall accuracy of 92.73% for internal validation and 87.75% for external validation, respectively. As for each category of thyroid scintigrams, the area under the receiver operator characteristic curve (AUC) was 0.986 for 'diffusely increased,' 0.997 for 'diffusely decreased,' 0.998 for 'focal increased,' and 0.945 for 'heterogeneous uptake' in internal validation, respectively. Accordingly, the corresponding performances also obtained an ideal result of 0.939, 1.000, 0.974, and 0.915 in external validation, respectively. CONCLUSIONS: Deep convolutional neural network-based AI model represented considerable performance in the classification of thyroid scintigrams, which may help physicians improve the interpretation of thyroid scintigrams more consistently and efficiently.

Thyroid and heart, a clinically relevant relationship.

BACKGROUND: Thyroid disorders, both overt and subclinical, are highly prevalent conditions in the general population. Although a clear relationship between overt thyroid dysfunctions and cardiovascular complications has long been established, data regarding subclinical thyroid dysfunction are by far more controversial. PURPOSE: The present review will be aimed at providing a summary of most recent evidence coming from meta-analyses regarding the complex relationship between thyroid dysfunction and cardiovascular disease. CONCLUSIONS: The review will summarize, in the first part, the physiopathological link between thyroid hormone imbalances and the cardiovascular system. In the second part the review will outline the evidence coming from meta-analyses regarding the cardiovascular risk related with both overt and subclinical thyroid dysfunctions. Particular attention will be put towards studies showing data stratified for patient's age, TSH levels and pre-existing cardiovascular disease. Finally, an overview regarding the effects of specific therapy for subclinical thyroid diseases in terms of amelioration of cardiovascular outcomes will be included.

Classification of Thyroid Dysfunction in Pregnant Women Differs by Analytical Method and Type of Thyroid Function Test.

CONTEXT: Physiological alterations challenge the assessment of maternal thyroid function in pregnancy. It remains uncertain how the reference ranges vary by week of pregnancy, and how the classification of disease varies by analytical method and type of thyroid function test. DESIGN: Serum samples from Danish pregnant women (n = 6282) were used for the measurement of thyrotropin (TSH), total and free thyroxine (T4), total and free 3,5,3'-triiodothyronine (T3), and T-uptake using "Method A" (Cobas 8000, Roche Diagnostics). TSH and free T4 were also measured using "Method B" (ADVIA Centaur XP, Siemens Healthineers). MAIN OUTCOME MEASURES: Pregnancy week- and method-specific reference ranges were established among thyroid antibody-negative women (n = 4612). The reference ranges were used to classify maternal thyroid function, and results were compared by analytical method and type of thyroid function test. RESULTS: The reference ranges for TSH showed a gradual decrease during pregnancy weeks 4 to 14, a gradual increase was observed for total T4, total T3, and T-uptake, whereas free T4 and free T3 showed less variation. When TSH and free T4 were used, Method A classified 935 (14.9%) with abnormal thyroid function, Method B a total of 903 (14.4%), and the methods agreed on 554 individuals. When TSH and total T4 were used, 947 (15.1%) were classified with abnormal thyroid function, and classifications by either total T4 or free T4 agreed on 584 individuals. CONCLUSIONS: Even when pregnancy week- and method-specific reference ranges were established, the classification of maternal thyroid dysfunction varied considerably by analytical method and type of thyroid function test.

Thyrotoxic periodic paralysis: A retrospective, observational study from India.

Background & objectives: Thyrotoxic periodic paralysis (TPP) is an endocrine emergency presenting with acute-onset flaccid paralysis in a patient having thyrotoxicosis accompanied by hypokalaemia. This study was conducted to evaluate the clinical profile of patients with TPP presenting to three centres in India. Methods: This retrospective, observational study was conducted at three tertiary care Armed Forces medical centres, located at Lucknow, Kolkata and Delhi. The history, clinical features, treatment details and outcomes were evaluated. Results: Of the 244 patients with thyrotoxicosis, 15 were diagnosed with TPP and included in the study. These 15 patients (14 male and 1 female) had 32 episodes of TPP which were analyzed. The mean age was 30.2±6.2 yr (range: 21-39), and overt thyrotoxicosis was seen in all patients except one who had subclinical hyperthyroidism. Graves' disease was the most common cause of thyrotoxicosis (13/15) and the remaining two patients had subacute thyroiditis and gestational thyrotoxicosis. Hypokalaemia (serum potassium <3.5 mmol/l) was seen in 12 patients, and the mean serum potassium was 3.2±0.9 mmol/l (range: 2.1-4.9). All patients had flaccid weakness, predominantly involving the lower limb with no bulbar, respiratory or cranial nerve involvement. The average duration of paralysis was 10.6±5.7 h (range: 3-28 h). Interpretation & conclusions: Our study demonstrated an early age of presentation and presence of clinical and biochemical thyrotoxicosis in majority of patients with TPP. Hypokalaemia may not always be evident in patients with TPP.

Association between benign thyroid disease and breast cancer: a single center experience.

BACKGROUND: The relationship between breast cancer (BC) and thyroid disease (TD) is still controversial. The aim of the study was to investigate the possible coexistence of TD in patients with newly diagnosed BC and its correlation with BC clinical presentation with regard to menopausal status and stage of disease. METHODS: This is a retrospective cohort study of all patients treated for primary BC between 2014 and 2016 at the Breast Unit of Trieste University Hospital. Clinical charts and reports were reviewed for coexisting thyroid disorders (i.e. hyperthyroidism, hypothyroidism, benign TD, thyroid cancer, thyroid autoimmunity) and menopausal status at the time of BC diagnosis. Biomolecular profile, stage, and grading of BC were also evaluated. RESULTS: A total of 786 women and 7 men were included in the study. Co-presence of TD was found in 161(20.3%) cases: of these, 151(19.4%) patients presented benign TD and 10(1.3%) patients presented thyroid carcinoma. Thyroid autoimmunity was found in 51(32%) patients. Regarding thyroid function, 88(55%) patients had hypothyroidism, 19(12%) hyperthyroidism, and 54(33%) normal thyroid function. No statistically significant correlation was found between age and TD (p = 0.16), although TD was more common in women aged ≥60 years. Women with BC diagnosed at pre-menopausal age were more likely to have thyroid autoimmune diseases (45% vs. 29%, p = 0.05). No association was detected among BC molecular profiles with either thyroid autoimmunity (p = 0.26) or altered thyroid function (p = 0.63). High-grade BC was more frequent in women with hyperthyroidism (52.9%, p = 0.04), but the grading was independent from the presence of thyroid autoimmune disease (p = 0.87). BC stage was related to both thyroid autoimmunity (p = 0.04) and thyroid function (p < 0.001), with 55.2% of women affected by benign TD presenting with stage I BC and more aggressive BCs found in hypothyroid patients. CONCLUSIONS: According our study results, patients with primary BC present a greater incidence of autoimmunity disorders, especially when diagnosed in the pre-menopausal setting. However, further prospective studies are required to definitively prove causality.

Subcategorisation of AUS/FLUS thyroid lesions as per the 2017 Bethesda System for Reporting Thyroid Cytopathology: a retrospective study from a tertiary care centre analysing risk of malignancy (ROM) of the different subcategories.

AIMS: The 2017 Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) recommends subclassification of atypia of undetermined significance (AUS)/follicular lesion of undetermined significance (FLUS) into six subcategories. The present study evaluates the risk of malignancy (ROM) and risk of neoplasm (RON) among these. METHODS: All thyroid aspirates reported as AUS/FLUS over a 4.5-year period, with available histology, were reviewed and subclassified as per TBSRTC. ROM and RON were calculated and compared. RESULTS: Of 2554 thyroid aspirates, 281 (11.0%) were AUS/FLUS. Eighty-one with available histology were evaluated. ROM was 51.8%. Cytologic and architectural atypia (AUS-C&A) was the most prevalent (62.9%), followed by Hürthle cell type (19.6%), AUS-A (11.1%), AUS-not otherwise specified (NOS) (7.4%), cytologic atypia (AUS-C) (4.9%) and atypical lymphoid cells (1.2%). Papillary thyroid carcinoma (PTC) and adenomatous goitre (AG) were the most common histological diagnoses (27% each). On histology, AUS-C had 2/4 PTC and 2/4 AG on histology. AUS-A had 4/9 follicular neoplasm (FN) and 2/9 non-invasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) while AUS C&A had 18/51 PTC, 13/51 AG, 11/51 NIFTP and 5/51 FN. ROM and RON were similar across subcategories, ROM was the highest for AUS-C&A (58.8%), AUS-C (50%) and AUS-NOS (50%). NIFTP reclassification as non-malignant reduced ROM to 35.8% (absolute reduction of 16% and a relative decrease of 31%) with the greatest relative decrease seen in AUS-A (50%), followed by AUS-C&A (37%), and none in others. CONCLUSIONS: AUS/FLUS subcategorisation helped to indicate risk for the more likely neoplasm, whether PTC or FN. ROM was the highest for cases with cytological atypia but did not differ significantly across different subcategories. NIFTP changed the ROM of AUS-A and AUS-C&A, since both NIFTP and FN have microfollicles.

Differentiating between benign and malignant thyroid nodules: An evidence-based approach in general practice

BACKGROUND: The widespread use of imaging techniques has led to more frequent detection of thyroid nodules, and while the majority are benign, the risk of malignancy in an adult ranges from 7% to 15%. General practitioners (GPs) must be able to evaluate thyroid nodules and refer cases when appropriate. OBJECTIVES: The aim of this article is to bring GPs up to date on the evidence-based management of thyroid nodules, with specific focus on neoplastic nodules, while highlighting significant changes in the 2015 American Thyroid Association guidelines. DISCUSSION: Thyroid nodules frequently occur in the general population. Differentiating between a benign and malignant nodule can be challenging, and community guidelines have standardised investigation, management and follow-up procedures. The key tests for risk stratification of thyroid nodules include serum thyroid-stimulating hormone testing, ultrasonography and fine-needle aspiration. GPs should be aware of the latest evidence-based recommendations for the appropriate management of a thyroid nodule.

Exploring Systemic Autoimmunity in Thyroid Disease Subjects.

INTRODUCTION: Individuals with one autoimmune disease are at risk of developing a second autoimmune disease, but the pathogenesis or the sequential occurrence of multiple autoimmune diseases has not been established yet. In this study, we explored the association and sequential occurrence of antibodies in thyroid disease and systemic autoimmune disease subjects. We evaluated thyroid hormones, thyroid-stimulating hormone (TSH), free thyroxine (FT4), thyroid autoantibodies, anti-thyroperoxidase (anti-TPO), and anti-thyroglobulin (Tg) to comprehend the association with systemic autoimmune autoantibodies, anti-nuclear antibodies (ANA), and autoantibodies to extractable nuclear antigens (ENA) in subjects with thyroid-related symptoms. METHODS: A total of 14825 subjects with thyroid-related symptoms were tested at Vibrant America Clinical Laboratory for thyroid markers (TSH, FT4, anti-TPO, and anti-Tg) and an autoimmune panel (ANA panel and ENA-11 profile) from March 2016 to May 2018. Thyroid-positive (based on TSH and FT4 levels), anti-TPO-positive, and anti-Tg-positive subjects were assessed for the prevalence of ANA and anti-ENA antibodies. A 2-year follow-up study was conducted to assess the sequential order of appearance of autoimmune markers in thyroid and systemic autoimmune diseases. RESULTS: In the retrospective analysis, 343/1671 (20.5%), 2037/11235 (18.1%), and 1658/9349 (17.7%) of thyroid+, anti-TPO+, and anti-Tg+ subjects were found to be seropositive for ANA. Anti-ENA was detected in a higher prevalence than ANA with 475/1671 (28.4%), 3063/11235 (27.3%), and 2511/9349 (26.9%) in the same groups of subjects, respectively. Our results are found to be much higher than the reported prevalence of anti-ENA in general population. During the 2-year follow-up study, anti-TPO appeared significantly earlier than ANA and anti-ENA in an average of 253 (±139) and 227 (±127) days, respectively. CONCLUSIONS: A high prevalence of anti-ENA and ANA was found to be coexisting with autoimmune thyroid disease subjects, with anti-TPO occurring prior to the onset of ANA and anti-ENA. Therefore, frequent follow-ups and evaluation of ANA and anti-ENA in subjects with anti-TPO positivity would be beneficial in early detection of other systemic autoimmune diseases.

Differences in Clinical Features Between Subcategories of "Atypia/Follicular Lesion of Undetermined Significance".

Within the Bethesda System for Reporting Thyroid Cytopathology (TBSRTC), category III (atypia or follicular lesion of undetermined significance (AUS/FLUS)) comprises specimens with heterogeneous features and the need for subcategorization has been reported. We compared the clinical features of two subgroups from within the category of AUS/FLUS to determine precise guidelines for clinicians who explain the results of fine-needle aspiration (FNA) to patients. Retrospective review was performed including data from all patients who underwent FNA with results reported as AUS/FLUS at a single tertiary hospital from January 2010 to August 2014. The results of repeated FNA, core needle biopsy, and diagnostic surgery were analyzed. Of the 903 patients with results categorized as AUS/FLUS, 500 were subcategorized as cellular atypia and 160 as architectural atypia. On repeated biopsy, diagnostic results were obtained for 78.9% patients of the cellular atypia group, compared to only 54.3% of the architectural atypia group (p < 0.0001). The rate of neoplasm or malignancy was also significantly higher in the cellular atypia group compared to the architectural atypia group (51.8 vs. 25.2%, p < 0.0001, 48.2 vs. 14.2%, p < 0.0001). Therefore, clinicians can inform patients with architectural atypia that the risk of malignancy is low. However, the high rate of non-diagnostic results on repeated biopsy makes clinical decisions difficult. The data of the present study revealed the necessity of subcategorization of category III of TBSRTC in the future.

AMERICAN ASSOCIATION OF CLINICAL ENDOCRINOLOGISTS AND AMERICAN COLLEGE OF ENDOCRINOLOGY POSITION STATEMENT ON THYROID DYSFUNCTION CASE FINDING.

Hypothyroidism and hyperthyroidism can be readily diagnosed and can be treated in a safe, cost-effective manner. Professional organizations have given guidance on how and when to employ thyroid-stimulating hormone testing for the detection of thyroid dysfunction. Most recently, the United States Preventive Services Task Force did not endorse screening for thyroid dysfunction based on a lack of proven benefit and potential harm of treating those with thyroid dysfunction, which is mostly subclinical disease. The American Association of Clinical Endocrinologists (AACE) is concerned that this may discourage physicians from testing for thyroid dysfunction when clinically appropriate. Given the lack of specificity of thyroid-associated symptoms, the appropriate diagnosis of thyroid disease requires biochemical confirmation. The Thyroid Scientific Committee of the AACE has produced this White Paper to highlight the important difference between screening and case-based testing in the practice of clinical medicine. We recommend that thyroid dysfunction should be frequently considered as a potential etiology for many of the nonspecific complaints that physicians face daily. The application and success of safe and effective interventions are dependent on an accurate diagnosis. We, therefore, advocate for an aggressive case-finding approach, based on identifying those persons most likely to have thyroid disease that will benefit from its treatment.

Thyroid disease as a risk factor for cerebrovascular disease.

BACKGROUND: Thyroid disease is the medical condition impairing function of the thyroid. Among this disorder category, hyperthyroidism is that the thyroid gland produces excessive amounts of thyroid hormones whereas hypothyroidism is that the thyroid gland does not produce enough thyroid hormone. Various studies have supported the comorbid association between thyroid disease and cardiovascular disorder. However, there is insufficient evidence to prove the relationship between cerebrovascular disease (CVD) and thyroid disease. METHODS: In this study, we tried to verify that thyroid disease increases the risk of CVD development employing a population-based database, National Health Insurance Research Database of Taiwan. A total of 16,808 hyperthyroidism cases and 5793 hypothyroidism patients with corresponding control subjects were studied, respectively. Hazard ratio (HR) by the Cox regression was used to quantify risk of CVD in different groups of subjects, that is, case patients versus matched controls. Further stratification studies for risk factors of CVD were performed to evaluate the comorbid association between CVD and hyperthyroidism/hypothyroidism. RESULTS: Evaluation results have shown that hyperthyroidism increased 38% of the hazard of developing follow-up CVD (adjusted HR, 1.38) whereas hypothyroidism increased even higher the risk (adjusted HR, 1.89). Further stratification studies for risk factors of CVD suggested that the comorbid association between hypothyroidism and CVD was comparable to those influences from cardiac risk factors, such as diabetes mellitus, hyperlipidemia, hypertension, or renal failure and so forth. CONCLUSIONS: Thyroid disease may predispose to onset of CVD. Advanced analysis is required to investigate the pathologic mechanism underlying the association between CVD and thyroid disease.

Reducing indeterminate thyroid FNAs.

BACKGROUND: The Bethesda System recommends limiting the percentage of cases diagnosed as indeterminate in thyroid fine-needle aspirations (TFNAs). However, studies are lacking that document how to decrease the rate of indeterminate TFNAs and its effects. METHODS: The authors modified the Bethesda criteria to reduce the rate of indeterminate TFNAs ("atypical cells of undetermined significance" and "suspicious for a follicular/Hurthle cell neoplasm") by reassigning cases that may have been categorized as indeterminate to "suspicious for papillary carcinoma" or "favor benign" for an 18-month period and correlated this with histologic follow-up. RESULTS: The indeterminate rate decreased from 14.4% (Bethesda) to 8.2% (modified Bethesda; P < .001). There was no significant difference in the risk of malignancy for the 2 groups (23.6% vs 17.3%, respectively; chi-square P = .39). There were 6 false-positive diagnoses in the Bethesda group compared with none in the modified Bethesda group (P = .06). In the Bethesda group, but not in the modified Bethesda group, the indeterminate rate was correlated with the overall rate of malignant and suspicious for papillary carcinoma diagnoses. CONCLUSIONS: Modifying the Bethesda criteria can decrease indeterminate TFNAs diagnoses without a significant decrease in the risk of malignancy. The modified Bethesda criteria, but not the Bethesda criteria, are independent of the malignancy rate.

The Bethesda System for Reporting Thyroid Cytopathology is applicable to frozen section diagnosis in children.

The Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) offers standardized and widely understood diagnostic categories for reporting thyroid cytology diagnoses. We compared the utility of TBSRTC categories in pediatric cytology diagnoses and pediatric intraoperative frozen section diagnoses. We examined the experience of our primary and referral care center over a 20-year period. This included 182 thyroidectomy patients who underwent 64 preoperative fine-needle aspirations and 91 intraoperative frozen section evaluations, including 38 patients evaluated sequentially by each method. All diagnoses were retrospectively reclassified into TBSRTC categories and correlated with the final thyroidectomy diagnoses. For each sampling method, malignant final diagnoses were observed at similar frequencies to rates predicted by TBSRTC. Malignant final diagnoses following fine-needle aspiration or frozen section diagnoses in TBSRTC categories other than malignant or suspicious for malignancy most often resulted from difficulty in detecting papillary carcinoma, including difficulty detecting the nuclear characteristics of papillary carcinoma in frozen sections. The limitations of needle biopsy and frozen section evaluations differ, yet serial utilization of these procedures was rarely informative. Based on the experience of our institution, classification of cytology and frozen section diagnosis by TBSRTC predicts a risk of malignancy similar to the guidance offered by TBSRTC. We recommend including a TBSRTC category when reporting either thyroid cytology or frozen section diagnoses in children.

Histopathological patterns of thyroid disease in Al-Madinah region of Saudi Arabia.

OBJECTIVES: This study aimed to characterize the histopathological pattern of thyroid lesions among Saudi patients and to highlight the age and gender variations of these lesions as base line data. MATERIALS AND METHODS: We retrospectively analyzed the data from thyroid specimens received at the Department of Pathology, King Fahad Hospital, Madinah, Saudi Arabia from January 2006 to December 2013. RESULTS: The 292 thyroidectomy specimens received during the study period came from 230 (78.8%) females and 62 (21.2%) males giving a female: male ratio of 3.7:1. Age of the patients ranged from 14 to 95 years with a mean age 39.7 years. Two hundred and eleven (72.3%) cases were found to be non-neoplastic and 81 (27.7%) cases were neoplastic. The non-neoplastic group included: colloid goiter, including both diffuse and nodular goiter (170 cases; 58.2%), nodular hyperplasia (28 cases; 9.6%), Hashimoto/chronic lymphocytic thyroiditis (12 cases; 4.1%), and Grave's disease (1 case; 0.3%). In neoplastic lesions, there were 7 benign tumors and 74 malignant tumors. Among the benign tumors, 5 were follicular adenomas and 2 were Hurthle cell adenomas. Papillary carcinoma was the commonest malignant tumor accounting for 87.8% of all thyroid malignancies, followed by lymphoma, follicular carcinoma and medullary carcinoma. The size of papillary carcinoma was more than 2 cm in 40 cases (76.9%). CONCLUSIONS: Non-neoplastic thyroid lesions were more common than neoplastic ones. Colloid goiter was the most common lesion. Follicular adenoma was the commonest benign tumor and papillary carcinoma was the commonest malignant lesion. There appears to be a slightly increased trend of papillary carcinoma diagnosis, most being diagnosed at an advanced stage.

[Evaluation of long-term results of surgical treatment of thyroid's diseases].

It was analyzed the surgical treatment results of thyroid's diseases in 462 patients in terms 1-15 years. Relapse of the disease was defined in 39.4% of cases after organ-preserving operations. Relapse formation was through 5-10 years after surgery. There were nodes up to 1 cm in diameter after hemistrumectomy and maximum subtotal resection of thyroid. Large nodes more than 3 cm in diameter were after subtotal resection of thyroid. Good results after interstitial laser photocoagulation were achieved in all patients, after sclerotherapy - in 89.5% of cases. An average degree of node reduction after laser destruction was 55.5±3.0%, after sclerotherapy - 49.7±_5.6%.

Few items in the thyroid-related quality of life instrument ThyPRO exhibited differential item functioning.

OBJECTIVE: To evaluate the extent of differential item functioning (DIF) within the thyroid-specific quality of life patient-reported outcome measure, ThyPRO, according to sex, age, education and thyroid diagnosis. STUDY DESIGN AND SETTING: A total of 838 patients with benign thyroid diseases completed the ThyPRO questionnaire (84 five-point items, 13 scales). Uniform and nonuniform DIF were investigated using ordinal logistic regression, testing for both statistical significance and magnitude (∆R(2) > 0.02). Scale level was estimated by the sum score, after purification. RESULTS: Twenty instances of DIF in 17 of the 84 items were found. Eight according to diagnosis, where the goiter scale was the one most affected, possibly due to differing perceptions in patients with auto-immune thyroid diseases compared to patients with simple goiter. Eight DIFs according to age were found, of which 5 were in positively worded items, which younger patients were more likely to endorse; one according to gender: women were more likely to report crying, and three according to educational level. The vast majority of DIF had only minor influence on the scale scores (0.1-2.3 points on the 0-100 scales), but two DIF corresponded to a difference of 4.6 and 9.8, respectively. CONCLUSION: Ordinal logistic regression identified DIF in 17 of 84 items. The potential impact of this on the present scales was low, but items displaying DIF could be avoided when developing abbreviated scales, where the potential impact of DIF (due to fewer items) will be larger.