The course of thyroid nodules and thyroid volume over a time period of up to 10 years: a longitudinal analysis of a population-based cohort

Design and sample

This is a longitudinal analysis of the population-based SHIP cohort (14). The study comprises data on demography, anthropometry, standardized thyroid laboratory measurements, thyroid ultrasound, self-reported data from a computer-assisted interview and billing data. The analysis is based on data of the first (SHIP-1, 2002–2006), the second (SHIP-2, 2008–2012) and the third follow-up (SHIP-3, 2014–2016). Only subjects who participated throughout all three follow-ups were included in this analysis (Fig. 1). Data from SHIP-0 (baseline, 1997–2001) were not considered, because billing data are only available for the period 2002–2016.

Flow chart of the study population selection.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Flow chart of the study population selection.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

Flow chart of the study population selection.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

Billing data (ICD-10-GM diagnoses and billing codes (German modification of the 10th revision of the International Classification of Diseases)) were provided by the Association of Statutory Health Insurance Physicians Mecklenburg-Vorpommern. The billing data were individually linked with SHIP data for the period 2002–2016. Participants with a history of thyroid surgery (SHIP-0) or thyroid surgery during the observation period (SHIP-1, SHIP-2, SHIP-3) (according to interview data and medical examination), with radioiodine therapy (SHIP-0 to SHIP-3, according to interview data and medical examination), with missing thyroid ultrasound data and/or billing data, without written informed consent for the use of billing data, with thyroid carcinoma 5 years prior to SHIP-2 examination (ICD-10-GM C73) and with lithium, amiodarone or thiamazole treatment (according to medication review) during the observation period were excluded from the analysis (Fig. 1). We excluded participants treated with lithium, amiodarone or thiamazole, because these medications are associated with goitre.

Measurements

Thyroid ultrasound was performed using the device Sonotron VST-Gateway from 2002 to 2005 and the portable device Vivid™i (General Electrics, Frankfurt a.M., Germany) with a 13 MHz linear array transducer from 2006–2016. Furthermore, ultrasound was performed by six different examiners throughout the observation period. The change in thyroid volume and number of thyroid nodules was observed from SHIP-1 to SHIP-3. Data on nodule size was only available in SHIP-2 and SHIP-3. Thyroid volume was calculated as length × width × thickness × 0.479 (mL) for each lobe. The average error of this equation to the true thyroid volume is about 16% (15). Goitre was defined as a thyroid volume > 18 mL in women and > 25 mL in men (16). Nodule numbers were documented in categories of 0, 1, 2 and more than 2 nodules. Nodules were only recorded in the right and left lobe, not in the isthmus. Also, only up to five nodules per lobe were documented. Nodule size was documented in two categories as nodules < 1 cm and nodules ≥ 1 cm.

The examiners had to complete training with at least 50 ultrasounds under the supervision and were then certified to perform ultrasounds for SHIP. Certification procedures were conducted in 15 volunteers before each examination wave and all 6 to 12 months during data collection. All intra- and interobserver variabilities for thyroid volume were < 5% (s.d. < 25%). To further reduce the observer and reader variation, the measurement error for thyroid volume calculated by Ittermann et al. (17) was considered in our analysis (s.d. 0.53).

Statistical analysis

In order to improve population representativeness, sampling weights were calculated and the dropout of participants between baseline and the second follow-up was considered by inverse probability weighting (ipw). To calculate ipw, logistic regression was performed with study participation as the outcome and dropout predicting variables from baseline investigation (age, sex, school years, household income, BMI, in a partnership, diabetes, smoking status, alcohol consumption g/day, HDL cholesterol, medication intake within the last 7 days, doctor visit within the last 12 months, systolic and diastolic blood pressure, antihypertensive medication, cancer) as independent variables. The participation probability for follow-up examinations was assigned to each subject, and the reciprocal of this probability was considered in the analysis.

Descriptive statistics was performed for the change in the number of thyroid nodules according to the categories, nodule size, goitre and thyroid medication. Changes in the number of thyroid nodules are shown on an individual level and categorized into decreased, unchanged and increased for a follow-up of 5 years and 10 years. The distributions of the thyroid volume at each SHIP examination, separately for men and women, are illustrated using boxplots. Linear mixed-effects models were used to assess the within subject difference in thyroid volume over time (random effects: intercept for subjects). Thyroid volume in mL was used as an outcome variable (continuous variable considering measurement error). The main predictor was the categorical variable time, representing SHIP-1, SHIP-2 and SHIP-3. Additionally, models were adjusted for age at SHIP-1 and BMI (SHIP-1 to SHIP-3) and stratified by sex. As a sensitivity analysis, participants with thyroid medication at any point in time during the observation period (SHIP-1 to SHIP-3) were excluded, and the same linear mixed-effects models, as well as the descriptive analysis on change of number of nodules, nodule size and prevalence of goitre, were performed (Supplementary Tables 1 and 2, see section on supplementary materials given at the end of this article). Furthermore, linear mixed-effects models were calculated, stratified by participants with goitre at baseline (SHIP-1) and participants with 'normal' thyroid volume according to the sex-specific cut-offs established by Gutekunst et al. (16) (Supplementary Tables 3 and 4). Data analysis was performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA).

Participant characteristics

A total of 1270 participants (53% female, median age at SHIP-1 51 years) were included in the analysis (Fig. 1 and Table 1). Of those, 66% (839/1270) had a good self-reported health status and 11% rated their general health as fair or poor. About 15% of all individuals reported a thyroid disease in the interview. Of those, more than half reported one or more thyroid nodules (53%, 100/188) and 11% goitre (20/188).

Changes in frequency of morphologic thyroid disorders and prescribed thyroid medication

The number of patients with at least one thyroid nodule increased from 34.9% at SHIP-1 to 39.6% at SHIP-2 (5-year follow-up) and to 47.5% at SHIP-3 (10-year follow-up) (Table 2). Throughout the entire observation period (SHIP-1 to SHIP-3), 44% (515/1270) had no nodules. There was an increase in the number of nodules over time. The highest increase was observed in the category > 2 nodules from 10% at baseline to 18% at 10 years follow-up. The number of participants with at least one nodule size ≥ 1 cm increased from SHIP-2 to SHIP-3 by 7%. In total, goitre was prevalent in 35% at SHIP-1 and 37% at SHIP-3 (Table 2). Women (32.5%) had a lower prevalence of goitre than men (38%) at SHIP-1. The prevalence of goitre in men decreased by 4% during the 5-year follow-up but then increased again by 5% during the 10-year follow-up (Table 2).

The number of participants with prescribed thyroid medication increased over time from 6% at SHIP-1 to 11% after 10 years. A total of 165 participants had a thyroid medication at one point in time during the observation period. Of those, 77.6% (128/165) were women. When excluding participants with prescribed thyroid medication at any point in time, the results did not differ substantially (Supplementary Table 1).

Change of number of nodules at an individual level

The majority of participants in the nodule category 0–2 (69.5%) had an unchanged number of nodules during a 5-year follow-up (Table 3). Participants with no nodules throughout the entire observation period (44%) are included in this analysis (Table 3). The proportion of participants with an unchanged number of thyroid nodules (category 0–2) reduced to 61.5% during the 10 years of follow-up.

In 31% of participants with > 2 nodules at SHIP-1 (n = 155), the number of nodules decreased at 5-year follow-up (Table 3). At 10-year follow-up, this proportion reduced to 24%. Yet, the majority in this category (>2) stayed unchanged and/or increased during the observation period.

Change in thyroid volume

The distribution in thyroid volume in both men and women remained constant over the observation period (not considering the outliers) (Figs 2and 3). The median volume in men was 22.8 mL (Q₁: 18.4 mL; Q₃: 28.4 mL) in SHIP-1 and 22.7 mL (Q₁: 18.5 mL; Q₃: 28.0 mL) in SHIP-3. The median volume in women was 16 mL (Q₁: 12.4 mL; Q₃: 19.8 mL) in SHIP-1 and 16.2 mL (Q₁: 12.6 mL; Q₃: 20.3 mL) in SHIP-3.

Thyroid volume in mL for men at the SHIP examinations, measurement error considered, >25 mL is defined as goitre established by Gutekunst et al. (15), marked with the bold line.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Thyroid volume in mL for men at the SHIP examinations, measurement error considered, >25 mL is defined as goitre established by Gutekunst et al. (15), marked with the bold line.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

Thyroid volume in mL for men at the SHIP examinations, measurement error considered, >25 mL is defined as goitre established by Gutekunst et al. (15), marked with the bold line.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

Thyroid volume in mL for women at the particular SHIP examination, measurement error considered, >18 mL is defined as goitre established by Gutekunst et al. (15), marked with the bold line.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Thyroid volume in mL for women at the particular SHIP examination, measurement error considered, >18 mL is defined as goitre established by Gutekunst et al. (15), marked with the bold line.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

Thyroid volume in mL for women at the particular SHIP examination, measurement error considered, >18 mL is defined as goitre established by Gutekunst et al. (15), marked with the bold line.

Citation: European Journal of Endocrinology 185, 3; 10.1530/EJE-21-0610

• Download Figure
• Download figure as PowerPoint slide

In women, the linear mixed effect models showed a mean increase of 1.18 mL (95% CI: 0.57; 2.42) after 5 years and 1.75 mL (95% CI: 0.84; 3.66) after 10 years (Table 4). In women, after adjusting for age and BMI, the 5-year and 10-year mean increase were 1.04 mL (95% CI: 0.51; 2.14) and 1.46 mL (95% CI: 0.70; 3.04), respectively. In men, there was a mean increase of 0.18 mL (95% CI: 0.07; 0.48) after 5 years and 0.57 (95% CI: 0.21; 1.52) after 10 years. After adjusting for age and BMI, the mean increase after 5 years was 0.15 mL (95% CI: 0.05; 0.39) and after 10 years 0.47 mL (95% CI: 0.18; 1.26). BMI had a stronger influence on thyroid volume than age at SHIP-1.

When excluding those participants with prescribed medication at any point in time, the mean increase for women was 0.90 mL (95% CI: 0.67; 1.23) after 5 years and 1.59 mL (95% CI: 1.18; 2.16) after 10 years. After adjusting for age and BMI, this decreased to 0.81 mL (95% CI: 0.59; 1.10) and 1.35 mL (95% CI: 0.98; 1.85) after 5 years and 10 years, respectively (Supplementary Table 2). The mean increase in men was 0.31 mL (95% CI: 0.19; 0.49) after 5 years and 0.90 mL (95% CI: 0.57; 1.42) after 10 years, when excluding participants with prescribed medication. After adjusting for age and BMI, this decreased to 0.25 mL (95% CI: 0.16; 0.40) and 0.74 mL (95% CI: 0.46; 1.18) after 5 and 10 years, respectively.

There was a mean increase of 1.46 mL (95% CI: 1.16; 1.85) after 5 years and 2.51 mL (95% CI: 1.99; 3.17) after 10 years, adjusted for age and BMI, only considering participants with a 'normal' thyroid volume at baseline (SHIP-1, Supplementary Table 3). When only considering participants with an enlarged thyroid at baseline (SHIP-1), the mean increase was 0.07 mL (95% CI: 0.04; 0.13) after 5 years and 0.18 mL (95% CI: 0.10; 0.35), after adjusting for age and BMI (Supplementary Table 4).

Strengths and limitations

This is the first study in Germany, using thyroid ultrasound measurements in a population-based cohort to investigate the course of structural thyroid abnormalities over one decade. To our knowledge, there are only a few international studies on the natural course and development of thyroid nodules or size over a long-time period. To account for sampling and attrition bias, we used sampling weights and inverse probability weights. Only a small number of participants was excluded due to missing ultrasound measurement. Our results are, however, limited by the definition of thyroid nodule number in categories (0; 1; 2; >2), which does not allow to assess changes of thyroid nodule number in participants of > 2 nodules (Table 3). A maximum of five nodules per lobe was documented, excluding nodules in the isthmus, which are not common (18, 19).

Nodule size was only measured at the second and third follow-ups. However, we hypothesize that a similar 5-year increase of 7% in the proportion of participants with at least one nodule ≥ 1 cm can be expected during a 5-year follow-up (Table 2). Clinical symptoms associated with thyroid disease were not documented in the study, and we did not have information on the exact treatment duration for thyroid medications. However, a minor increase in thyroid volume is unlikely to cause dyspnoea or dysphagia.

Over the long study period, two different ultrasound devices were used and ultrasound was performed by different examiners. To account for interobserver variation and observer bias, examiners underwent rigorous training and certification. We additionally included the measurement error for thyroid volume in our analysis.

Clinical meaning of the results and comparison with literature

Structural abnormalities of the thyroid gland, including nodules and goitre, are very common and often lead to repeated testing, including functional test, ultrasound follow-up and scintiscan, which may induce invasive procedures. Previous analyses of the SHIP cohort have shown that 6.8% (378/5552) of subjects had at least one ultrasound and 2.6% (146/5552) had at least one scintiscan within 1 year prior to the study examination based on billing data (20). The majority of structural abnormalities are asymptomatic, clinically irrelevant and malignancy developed in only a very small proportion (6, 7). Therefore, this is an area of potential overtesting and overdiagnoses, leading to increased use of medical resources and costs.

There are only a few current international clinical guidelines with diagnostic and follow-up recommendations on thyroid nodules, mainly based on histological assessment (1, 4). Once a thyroid nodule is detected, a cascade of follow-up examinations with moderate sensitivity and even worse, low specificity, is recommended to exclude thyroid cancer. In clinical practice, ignoring findings or systematic follow-ups is common (20). Once follow-up examinations are started, there is uncertainty when to discontinue. Guidelines do not recommend a time scope for discontinuing follow-up in asymptomatic patients without clinical changes. A retrospective study from Cappelli et al. 2020 showed that the majority of benign nodules remained stable in size over a 10-year follow-up (83%), and only 3 out of 75 patients with nodule growth (4%) had thyroid cancer according to the FNA. Based on these findings, the authors suggest ultrasound follow-up after 2 years and then after 5 years following cytological evaluation. Afterwards, they recommend a follow-up every 4 to 5 years (21).

The lack of clear recommendations on larger follow-up intervals or decisions to stop follow-up despite the low incidence of malignancies is partly due to a knowledge gap on the (natural) course of structural thyroid abnormalities. Large studies based on biopsy results are not feasible due to ethical and practical reasons. Therefore, epidemiological data from large population-based cohort studies using ultrasound assessment and billing data to assess medical treatment and outcomes can shorten the gap and enhance future guideline recommendations. In our study, nearly 70% of participants had an unchanged number of nodules after 10 years, although for participants with > 2 nodules, we were only able to assess a proportion of up to 25% with a decreased nodule number after 10 years. These results are in line with other studies, which reported that only 9% of subjects developed new thyroid nodules after 5 years (6). In that study, the size of thyroid nodules increased only in 11% of the cases and the cancer incidence rate of original nodules was 0.3%.

For individuals with at least one nodule ≥ 1 cm, current guidelines recommend fine-needle aspiration (1). From our study, patients with malignancy were excluded, but in the total SHIP cohort, only four cases of thyroid cancer, based on billing data, were identified. In an observational study by Medici et al., the authors followed up 1254 patients with cytologically benign nodules sonographically and reported no malignancies or thyroid cancer-related deaths, if the interval for the first follow-up was extended from 3 to 4 years (22). Although there is a degree of clinical uncertainty, the magnitude according to current data is low, considering the prevalence of malignancy and the favourable long-term outcomes.

We argue that epidemiological data on the (natural) course of morphological changes and corresponding risks should be explicitly communicated to patients. Our results support a more restrictive approach regarding thyroid imaging and follow-up examinations in asymptomatic patients with diagnosed thyroid nodules, based on informed shared decision-making. An example of such an approach is the work up for incidental thyroid nodules detected at initial imaging recommended by the American College of Radiology (23, 24). For example, for nodules not suspicious on CT, MRI or ultrasonography, no further evaluation is recommended in patients younger than 35 years with nodules < 1 cm and in older patients with nodules < 1.5 cm, including for patients with limited life expectancy (24).

In our study, thyroid volume in both women and men remained stable over the 10-year period. The minor observed changes (<1.5 mL) were even smaller when adjusted for age at baseline and BMI and are not clinically relevant (25). Although our analyses focus on volume change over time, our data also indicate a large proportion of the study population to have goitre (38% men, 33% women). Healthy thyroid volume was defined according to current standards in ultrasound assessment, which in fact date back to cross-sectional German data from 1988 (16). The distribution of thyroid volume in our data indicates that a substantial proportion of subjects has a volume above the established reference volume due to iodine deficiency. We hypothesize that a minor increase in thyroid volume above reference volumes might have increased goitre prevalence substantially. It is likely that a rigid definition of thyroid volume reference values for goitre leads to overmedicalization and subsequent follow-up diagnostics and, in some cases, medical treatment even in the absence of functional abnormalities. Clinical significance of goitre arises from its association with nodules and functional abnormalities and/or clinical symptoms. In the absence of these, deviation of thyroid volume around a singular reference value will provide little meaningful information. We argue that centile curves for thyroid volume (age/BMI) might provide a more useful tool to quantify thyroid volume deviation and will better reflect the distribution of 'normal' volume in the population. However, reference values reflect regional iodine supplementation status, and changes over time will reflect changes in iodine supplementation. Therefore, the use of reference values is appropriate for epidemiological studies but should not be used for clinical decision-making or defining illness.