The Relationship between Body Mass Index and Thyroid Cancer Pathology Features and Outcomes: A Clinicopathological Cohort Study

Abstract

Background: Obesity has been implicated as a predisposing and disease-modifying factor in cancer. Epidemiological studies suggest that obesity is associated with an increased risk of thyroid cancer; however, the relationships between obesity and thyroid cancer stage or behavior are uncertain. We hypothesized that a higher body mass index (BMI) would be associated with aggressive thyroid cancer features and a higher incidence of persistent/recurrent disease.

Methods: Two hundred fifty-nine consecutive patients with thyroid cancer were enrolled in this retrospective cohort study. Histopathological tumor features, stage at diagnosis, and disease status during and at the end of the study were determined based on chart review. BMI was calculated at the first clinical visit to our institution. The relationships between BMI and these parameters were assessed.

Results: Mean follow-up time for the group was 6.2 yr (0.11–46 yr). No positive associations were identified between BMI and T, N, or M stage at diagnosis, vascular invasion, or recurrent or persistent disease on univariate or multivariate analyses. The absence of an association was also demonstrated on analysis by BMI quartiles. An unexpected inverse association was identified between BMI and nodal metastasis and tumor invasion on both univariate and multivariate analyses, suggesting that obesity may be associated with less aggressive tumor features, a finding that requires confirmatory studies.

Conclusion: Although obesity has been associated with increased thyroid cancer incidence, a higher BMI was found not to be associated with more aggressive tumor features or a greater likelihood of recurrence or persistence over the analyzed time period.

Obesity is not associated with aggressive tumor features or a greater likelihood of recurrence or persistence in differentiated thyroid cancer.

Thyroid cancer incidence is rising at a rate that is among the fastest of all malignancies (1,2,3,4,5,6,7); it is estimated that about 37,000 new cases of thyroid cancer were diagnosed in the United States in 2009 and this disease accounted for about 1,600 deaths (8). Although the causes for this increase in incidence are incompletely defined, enhanced detection of early-stage tumors through the use of neck ultrasound and ultrasound-guided biopsy in the past 2–3 decades likely has a major impact. For example, Davies and Welch (2) reported a doubling of the incidence of thyroid cancer from 1973 to 2002 in the Surveillance, Epidemiology, and End Results database, with 87% of cases related to tumors less than 2 cm. However, it is also notable that a similar but smaller trend of increased thyroid cancer incidence began before the widespread use of diagnostic ultrasound (9), and a recent analysis of the Surveillance, Epidemiology, and End Results data from 1980 to 2005 also identified an increase in incidence of large (>5 cm) papillary thyroid cancers (10), raising the possibility that there may be other contributing factors.

During this same time period, it has been demonstrated by the National Health Examination Survey and the National Health and Nutrition Examination Survey that obesity rates nearly doubled in the United States (11). A potential role for obesity as a precipitating factor in cancer development has been reported. For example, one recent metaanalysis of several studies in which the relationship between body mass index (BMI) and cancer was reported demonstrated an association between obesity and several types of cancer (12). Individual epidemiological surveys have demonstrated an increased risk of adenocarcinoma of the esophagus, colon, kidney, endometrium, and malignant melanoma with increasing BMI, suggesting that the obesity epidemic may impact cancer incidence (13). In addition to cancer risk, obesity has also been associated with aggressive tumor pathological features and worse outcomes. For example, a higher BMI has been associated with an increased risk of recurrence and progression of breast and prostate cancers by several groups (14,15,16,17,18,19,20,21,22). Potential mechanisms include interactions between adipokines and cancer cells (23).

For differentiated thyroid cancer, exposure to ionizing radiation and dietary deficiency of iodine are established environmental risk factors for tumor development. Obesity has also been reported to be associated with an increased incidence of thyroid cancer in some cohorts. Engeland et al. (24) demonstrated an increased relative risk of thyroid cancer in women above age 50 yr with a BMI greater than 30 kg/m² in a Norwegian cohort [relative risk (RR) of 1.31 (1.09–1.57)]. In a French Polynesian cohort, being overweight or obese at age 18 yr was associated with an increased incidence of thyroid cancer later in life, particularly in women (25). In a cohort of Korean men, Oh et al. (26) identified an increased incidence of papillary thyroid cancer in men with higher BMI. Conversely, Iribarren et al. (27) did not identify an association between obesity and thyroid cancer. In a pooled epidemiological analysis of several studies, increasing BMI was reported to account for a moderate increase in thyroid cancer risk in females in the highest BMI tertile vs. those in the lowest tertile, and a statistically significant association between BMI and a thyroid cancer diagnosis in men was also identified (28). More recently in a metaanalysis of epidemiological publications reporting on the relationship between BMI and cancer risk for many cancer types, an increase in weight of 5 kg/m² was associated with an increased risk of a thyroid cancer diagnosis in both men [RR 1.33 (1.04–1.70; P = 0.02)] and women [RR 1.14 (1.06–1.23; P = 0.001)] (12). Thus, although the data are not entirely consistent, an association between obesity and an increased incidence of thyroid cancer has been reported in the majority of the analyzed populations.

Whereas these studies identified a link between obesity and thyroid cancer, they have not addressed whether obesity is associated with more aggressive thyroid cancer. Because this relationship has been reported for other obesity-associated malignancies, we sought to examine the potential relationship between obesity and aggressive pathological features and clinical course in a cohort of patients from whom extensive clinical data are available using modern methods of surveillance.

Patients and Methods

Design and data sources

Three hundred fourteen patients were recruited for entry into the Institutional Review Board-approved Endocrine Neoplasia Repository database during their appointment in the thyroid neoplasia clinic at the Ohio State University Medical Center between 2006 and 2008. All patients seen in the clinic were offered enrollment in the study and informed consent was obtained. The enrolled individuals completed an extensive medical history questionnaire that was subsequently reviewed with a medical professional and had serum and tissue banked for further study. Only subjects with documented differentiated follicular cell-derived thyroid cancer were included in this report (n = 259). Further detailed chart review, pathology review when slides were available, review of radiographic, and biochemical testing was performed to classify the stage of disease at diagnosis and to determine clinical status during the course of follow-up as described in detail below.

BMI calculations

BMI was calculated using measurements of weight and height obtained at the first clinical visit to The Ohio State University. BMI groupings were based on standardized categories set by the World Health Organization as follows: underweight or normal weight (<25 kg/m²; n = 79), overweight (25–29.99 kg/m²; n = 79), obese class 1 (30–34.99 kg/m²; n = 53), and obese classes 2 and 3 (>35 kg/m²; n = 48). In 48% of our patients, BMIs were recorded within 6 months of their diagnosis, whereas others were done at later times. Because of this variability, we explored the effects of timing of BMI measurement on associations in our analysis (see below).

Primary end points

BMI associations with clinicopathological features of thyroid cancer at the time of diagnosis were made with tumor size, nodal status, and evidence of metastasis based on the American Joint Committee on Cancer tumor node metastasis staging system, sixth edition (29). Additional primary end points included the presence or absence of tumor and vascular invasion at the time of initial diagnosis of thyroid carcinoma. For tumor invasion, any evidence of tumor invasion on pathology report was considered presence of invasion. For the statistical analysis, nodal status was designated as the presence (N1a or N1b), absence (N0), or unknown [no nodes removed at surgery (NX)] of nodal metastasis. Distant metastasis at the time of diagnosis required treatment with I-131 and was determined by radiographic evidence of disease on whole-body I-131 scanning within a year of diagnosis. In some cases, computed tomography, magnetic resonance imaging, or positron-emission tomography was also performed. If these results were interpreted as clear evidence of distant metastases by the treating clinician (R.T.K. or M.D.R.) based on image results combined with a markedly elevated thyroglobulin, biopsy result, or concordance with I-131 scans, they were considered to have metastatic disease at diagnosis. Uncertain cases were considered to have unknown metastatic status (MX) at diagnosis. In addition, patients not treated with I-131 were also considered MX at diagnosis (n = 28), including nine individuals treated by hemithyroidectomy only. Because of the reported associations between tumor invasion beyond the thyroid capsule and vascular invasion and thyroid cancer outcomes (30,31,32), these two variables were analyzed separately from staging in which tumor invasion is included with other features as part of T categories based on its extent.

The relationship between BMI and the likelihood of having persistent or recurrent thyroid cancer was also examined. For all subjects during clinical follow-up, disease status was categorized as free of disease, persistent disease, or recurrent disease. Disease-free status was defined as an undetectable TSH-stimulated thyroglobulin (Tg) with no radiographic evidence of disease, stable Tg if radioiodine ablation was not performed, or using radiographic evidence of disease alone if anti-Tg antibodies were present (n = 20). Persistent disease was defined as having a detectable Tg and/or radiographic or pathological evidence of disease (computed tomography, ultrasound, magnetic resonance imaging) in patients who did not achieve a complete remission after surgery and I-131 therapy. Recurrent disease was defined as having new evidence of disease after achieving disease-free status as defined above. Due to the duration of follow up of the study, patients may have had Tg levels measured using different assays over time with different lower limits of detection. Before March 2003 assays were performed using the Nichols chemiluminescence immunometric assay (Nichols Institute Diagnostics, catalog no. 60-4240, analytical sensitivity 0.07 ng/ml). The inter-assay coefficients of variation for this assay for four serum pool controls (1.49 to 2.29, 18.10 to 21.90, and 3.93 to 6.01 ng/ml) were 10.6%, 4.8% and 10.5%, respectively, and 24.8% for a Tg level of 0.44 ng/ml. The lowest reported detectable level with this assay was 0.5 ng/ml. Between March, 2003 and February, 2006, the Nichols Advantage thyroglobulin assay (Nichols Institute Diagnostics, San Clemente, CA; catalog no. 62-7035, analytical sensitivity ≤ 0.04 ng/ml) was used. The interassay coefficients of variation for this assay for four serum pool controls (0.68, 2.4, 18.2, and 54.4 ng/ml) were 14.7, 9.9, 9.9, and 8.4%, respectively. The lowest reported detectable value was 0.9 ng/ml. From February 2006 to the present, the Immulite thyroglobulin assay (Siemens Inc., Deerfield, IL; catalog no. PIL2KTY, analytical sensitivity < 0.2 ng/ml) was used. The interassay coefficients of variation for this assay for four serum pool controls (0.25, 1.8, 10.7, and 39.6 ng/ml) were 12, 4.7, 4.7, and 5.8%, respectively. The lowest reported detectable value was 0.2 ng/ml. To minimize the impact of the assay changes and the different lower detectable values, we defined the absence of disease as a Tg level below the lowest detectable value of the assay being used rather than choosing a specific cut point.

Statistical analysis

Logistic regression modeling was used to obtain BMI odds ratios for the binary clinical variables, which include nodal status, tumor invasion, vascular invasion, and metastasis status. BMI was used as a continuous variable in the initial models. First univariate estimates were obtained, and then estimates adjusted for potential confounders were obtained. The following potential confounders were used to adjust estimates: age, gender, smoking status (ever or never smoked), and diabetes status. We added an indicator for when BMI was recorded (within 6 months of diagnosis vs. later) to determine whether the timing changed the association substantially. We also performed an analysis using four categories (quartiles) of BMI according to World Health Organization (WHO) guidelines. One-sided tests of the adjusted odds ratios for BMI and the five clinical variables used Holm’s procedure to strongly control type I error at alpha = 0.05. For the persistence variable, we also used alpha = 0.05. With 259 patients we had approximately 80% power to detect odds ratios in the range of 1.8 to 2.5 (depending on the prevalence of the variable) for a 10-point increase in BMI. The small percentage of patients with distant metastasis (4.4%) made power inadequate for this variable. We included the results for this variable for exploratory purposes, but the confidence interval is very wide. To account for effect of socioeconomic status on the results, the percent of individuals living below the national poverty level in the ZIP code in which each patient resided at the time of study enrollment was obtained (city-data.com; 2008 data). These data were included in the model as both a continuous variable and a dichotomous variable (percentage in poverty either above or below the level in the state in which the patient resides). Ethnicity was also included in the multivariable model.

Results

BMI and clinical-pathological features of thyroid cancer

The relationship between BMI and clinicopathologic features of thyroid carcinoma at diagnosis and clinical status at the end of the study was evaluated in 259 patients. Patient demographics and BMI categories are shown in Table 1. In this cohort, the female to male ratio was 3.4:1, 90% (234 of 259) had papillary thyroid cancer (PTC) and 10% (25 of 259) had follicular thyroid cancer (FTC). Variants of FTC or PTC were included in the appropriate FTC or PTC category. The duration of follow-up ranged from 1.3 months to 46 yr, with a mean of 6.2 yr. In the population, 93% were Caucasian, 2% Asian, 2% black, 1% Hispanic, 1% multiracial, 0.5% Native American/Alaskan Native, and 0.5% of unknown ethnicity. Forty-two percent of subjects were followed up for more than 5 yr and 8% were followed up for less than 1 yr. The high percentage of patients with recurrent or residual disease likely reflects our institutional referral bias. For N stage, we had a larger amount of missing data compared with the other parameters due to the lack of removal of lymph nodes in many surgeries. Thus, only 178 of 259 patients are included for this variable. For other variables the data were almost all complete as seen in Table 2. There were some patients in whom tumor size or the presence or absence of invasion was not able to be discerned precisely from original pathology reports, and they were excluded from the analysis for that particular variable. For one subject smoking was missing so it is not included in the reported odds ratios. Socioeconomic data were available for all patients and ethnicity was available for all but one patient.

Table 1.

Clinical and pathological features of cohort

Feature	Frequency
Gender (female:male)	(3.5:1)
Age (median, range)	(46.6, 15.9–89.2)
BMI (median, range)	(27.8, 18–59)
PTC (n, %)	(234/259, 90%)
FTC (n, %)	(25/259, 10%)
T1 (n, %)	(97/247, 39%)
T2 (n, %)	(74/247, 30%)
T3 (n, %)	(52/247, 21%)
T4 (n, %)	(24/247, 10%)
N1 (n, %)	(105/178, 59%)
M1 (n, %)	(10/229, 4%)
Tumor invasion (n, %)	(64/248, 26%)
Vascular invasion (n, %)	(32/259, 12%)
Recurrent or persistent (n, %)	(119/259, 46%)

Two hundred fifty-nine patients with differentiated thyroid cancer are included in the cohort. Note that not all pathology variables were recorded for all patients, so the number of patients for which some features are evaluable is lower than the total cohort based on record and pathology review as noted in the denominators. Only patients in whom the features could be verified were included in the data analysis. 

Table 2.

Odds ratio of more severe disease with a 10-point increase in BMI

Odds ratio for disease severity indicators per 10 point increase in BMI (95% confidence interval)
	Tumor invasion	Vascular invasion	Distant METs	Tumor size	Nodal METs	Persistent vs. not
All patients	0.44	0.61	1.29	0.78	0.52	1.00
	(0.25, 0.77)	(0.32, 1.17)	(0.36, 4.64)	(0.49, 1.21)	(0.32, 0.84)	(0.69, 1.44)
	n = 247	n = 258	n = 228	n = 246	n = 178	n = 258
	P = 0.004	P = 0.14	P = 0.69	P = 0.26	P = 0.008	P = 0.98

The data are presented as the odds ratio and P value of the pathology or follow-up feature as a function of a 10-point increase in calculated BMI. The data are controlled for patient age, gender, smoking, diabetes, socioeconomic status, and ethnicity. The number in the groups differ because all information was not available for each patient. Significant P values are noted for tumor invasion and the presence or absence of nodal metastases (METs). The confidence intervals are below 1, meaning that the higher the BMI, the lower the risk of having that particular feature. 

Of the association tests performed between BMI and thyroid cancer severity, no significant results were obtained in the hypothesized direction (see Table 2 for all the covariate adjusted odds ratios). Although we include P values in Table 2, it is important to recognize that these P values are based on results opposite from what was hypothesized. For two of the variables, the presence of nodal metastases and tumor invasion, the P values are quite small but must be considered suggestive and require confirmation. For both variables, a higher BMI was associated with a lower risk of tumor invasion and nodal metastases. Unadjusted odds ratios also showed the same result with range values that are less than 1 at the upper end (Table 2). The time between BMI measurement and thyroid cancer diagnosis (<6 months vs. > 6 months) was used in a model to determine whether this timing of measurement impacted on the association (interaction effects between time and BMI). Recall from Table 1 that 48% of the patients had BMI recorded within 6 months of diagnosis. In this sensitivity analysis, all of the odds ratios for the subgroup of patients who had BMI measured within 6 months of diagnosis were close to what was found from all patients. In particular for this subgroup, the odds ratio for tumor invasion and nodal metastasis were 0.33 and 0.46, respectively, which are very close to the values reported in Table 2 based on all patients.

To further understand the unexpected relationship between BMI and the end points, the subjects were categorized into underweight, normal, overweight, and obese according to the WHO. Using these categories, the two highest BMI groups had a lower frequency of vascular invasion and nodal metastases (Fig. 1). The percentage of underweight or normal, overweight, and obese patients were similar to 2008 data from the Centers for Disease Control and Prevention (http://cdc.gov/brfss) for the eight-county region of Ohio that includes Columbus.

Figure 1.

Lymph node metastasis and invasion are less frequent in overweight and obese patients. Patients were separated into WHO quartiles of BMI. When normalized to normal or underweight individuals, patients with BMI in the 30–35 and greater than 35 kg/m² ranges were less likely to have nodal metastasis (Mets) or local tumor invasion, with both odds ratios and confidence intervals being less than 1.0 in the highest weight group. The numbers of patients (n) in the different weight groups from whom nodal metastasis (NM) or tumor invasion (TI) data were available are included below the graph.

BMI and clinical outcomes

Patients were characterized for their disease status over the course of the study on the basis of the presence or absence of a detectable Tg level performed during TSH suppression and after TSH stimulation (either after withdrawal of levothyroxine or administration of recombinant human TSH), radiographic imaging, and pathology evaluations as part of the routine clinical practice of two authors (R.T.K. and M.D.R.). Based on this information, patients were classified as having no evidence of disease, evidence of persistent disease, or evidence of recurrence (i.e. achieved the definition of no evidence or disease and later recurred). For statistical analysis, we dichotomized them into no disease vs. recurrence or persistence because the intensity of monitoring may not have been similar before referral vs. after referral to our institution. The odds ratio for BMI and this clinical outcome is provided in last column of Table 2. In this analysis, no significant association was found between recurrent or residual thyroid cancer and BMI. Because few patients died during follow-up, there was insufficient power to detect an effect for this outcome.

Discussion

Increased BMI has been associated with a higher cancer incidence for several malignancies, including thyroid cancer (12). In addition, obesity has been reported to be associated with poor pathological prognostic correlates and the development of recurrence and metastases for several cancer types, including breast, prostate, and colon cancers (17,18,21). Experimental models have suggested that adipokines such as adiponectin, leptin, and hepatocyte growth factor regulate cancer cell proliferation and invasiveness, and in vivo studies have demonstrated relationships between obesity, adipokines, and cancer progression (reviewed in Ref. 23). Because thyroid cancer has been associated with obesity in a number of studies, we examined the relationship between obesity, defined by BMI, and thyroid cancer aggressiveness and outcomes in a carefully characterized cohort of patients using sensitive Tg assays and imaging modalities (33).

In our cohort, a relationship between obesity and more aggressive histological features and tumor recurrence or persistence was not detected on either univariate or multivariate analyses. Moreover, although the study was not designed to detect an inverse relationship between obesity and histological features or outcomes, such a relationship was suggested by the statistical analysis. Interestingly, a lower risk of malignancy in obese patients with indeterminate cytology on fine-needle aspiration was recently reported (34). The reasons for possible tissue-specific difference in the relationships between obesity, cancer histology, and clinical course are uncertain. It is potentially relevant that the amount of adipose tissue in the thyroid is substantially lower than breast tissue. This might lead to a different impact of cancer-adipocyte interactions on tumor behavior in these two tissues. Moreover, the effects of circulating or locally produced adipokines, such as adiponectin, leptin, and visfatin, may be cell type specific. Increased expression of leptin and/or its receptor in thyroid cancer were recently reported by two groups (35,36), one of which also demonstrated an association with more aggressive tumor features and effects of biological behavior of thyroid cancer cell lines in vitro (36). However, no association between the increased expression of leptin or its receptor and BMI was identified (35). Thus, whereas leptin may be an important autocrine or paracrine factor affecting thyroid cancer biology, its relationship in thyroid cancer with systemic obesity is uncertain. Further studies to evaluate these relationships may help clarify the biological relevance of our findings.

There are several weaknesses of our study to be considered. First, although BMI determinations have been the most commonly cited measurement in the medical literature for this type of study, measurements such as waist to hip ratio, percent body fat, skin-fold thickness, and/or assessment of intraabdominal fat may be better measures of obesity. These measurements were not available for analysis in this retrospective study. However, such a wide range of BMIs were found in this study, and it is likely any inaccuracies likely would attenuate the associations only slightly.

Second, it was not possible to assess BMI in many patients before the thyroid cancer diagnosis; thus, the impact of prediagnosis obesity was not able to be assessed. We chose to consider the first BMI measurement at our institution to be consistent for each patient and to have greatest confidence in the measurements. The potential effect of the duration of time between the diagnosis of thyroid cancer and the BMI measurement was assessed statistically and no impact was identified. However, this analysis does not exclude an effect of long-standing obesity before thyroid cancer diagnosis on the behavior of the tumor after diagnosis. Indeed, a positive relationship between the duration of obesity and thyroid cancer incidence has been reported (25) in some but not all studies (24).

Third, because this was a retrospective study, the extent of surgery varied between patients, as documented by the absence of information regarding lymph node metastases in a substantial minority of patients. In addition, data regarding how the thyroid nodule or cancer was originally detected was not routinely documented; thus, it is not possible to assess the impact of mode of diagnosis on initial tumor stage in this study. It is possible, for example, that differences in the mode of diagnosis (physical examination vs. incidental on radiographic testing) might differ based on BMI. However, although these may affect the initial tumor staging, the absence of a higher frequency of tumor persistence in the more obese patients despite using the same monitoring paradigm for all patients suggests that any impact on the results is likely to be small.

Fourth, the duration of follow-up in the present study was relatively short for thyroid cancer. For this reason, we focused on tumor recurrence or persistence rather than tumor-specific survival as the primary end points. In addition, because all of these patients were actively followed up by ultrasound and sensitive Tg measurement in the same manner, regardless of BMI at one institution by a limited number of clinicians, it seems likely that the majority of recurrences would have been detected during the duration of follow-up in this study as has been reported (37,38). Nonetheless, we plan to follow up this cohort for a longer period of time to further account for a long-term influence of obesity on thyroid cancer progression over time.

Finally, prediagnosis TSH levels were not available from many patients; thus, it is not possible to assess a relationship between TSH levels and the stage of thyroid cancer at diagnosis.

Our initial hypothesis that obesity would be associated with worse histological tumor features and a higher rates of persistent or recurrent disease was not confirmed, and the result that obesity appears to be associated with better tumor features was unexpected. We did address the potential confounding factors of age, gender, smoking, diabetes, ethnicity, and socioeconomic status. In particular, we were concerned that the patients with the lowest BMI quartile might be younger, and because younger patients with thyroid cancer are reported to have a higher incidence of nodal metastases, that this might increase the frequency of lymph node metastases. In addition, one might predict that physical examination and imaging may be less sensitive in obese subjects, thereby leading to a later tumor stage diagnosis. This was not supported by our data. Whereas inclusion of ethnicity and socioeconomic status in the multivariate model did not alter the relationship between obesity and tumor stage, the population was largely Caucasian and represents data from a single institution; thus, it is possible that larger multicenter studies will uncover relationships between socioeconomic status or ethnicity with stage at detection and outcomes not found in this study. However, if the associations found here are confirmed in subsequent studies, mechanistic studies will be required to fully understand the relationship between lesser degrees of thyroid cancer invasion and nodal spread and obesity.

In summary, in this retrospective cohort study of a well-characterized group of thyroid cancer patients, obesity was not associated with aggressive histological features or behavior over a mean 6.2 yr follow-up duration. Thus, although obesity has been associated with an increased incidence of thyroid cancer, the results from this cohort study do not support an association between obesity and aggressive thyroid cancer features or behavior. Additional studies to confirm these findings and to establish potential relationships between obesity and thyroid cancer biology are warranted.