Serum thyroglobulin, a biomarker for iodine deficiency, is not associated with increased risk of upper gastrointestinal cancers in a large Chinese cohort

Abstract

Iodine concentrates in gastric tissue and may act as an antioxidant for the stomach. We previously showed that self-reported goiter was associated with significantly increased risk of gastric noncardia adenocarcinoma (GNCA) and non-significantly increased risks of gastric cardia adenocarcinoma (GCA) and esophageal squamous cell carcinoma (ESCC) in a prospective case-cohort study in a high-risk population in China. Negatively correlated with iodine levels, serum thyroglobulin (Tg) is a more sensitive biomarker of iodine deficiency than goiter. This study aimed to determine whether baseline serum Tg was also associated with development of GNCA, GCA, and ESCC in the same cohort, the Linxian General Population Nutrition Intervention Trial. Sera from approximately 200 subjects of each case type and 400 non-cases were tested for serum Tg concentration using appropriate assays. Tg was modeled as sex- and assay-specific quartiles in Cox regression models adjusted for age, smoking, alcohol, Helicobacter pylori status, pepsinogens I/II ratio, family history, and commune of residence. In the final combined analysis, participants in the highest quartile of serum Tg, compared to those in the lowest quartile, had adjusted Hazard Ratios of 0.88 (95% confidence interval 0.50–1.52), 1.14 (0.63–2.05), and 0.78 (0.47–1.31) for GNCA, GCA, and ESCC, respectively. Using serum Tg, a sensitive biomarker of iodine deficiency, we found no association between serum Tg concentrations and risk of these upper gastrointestinal (UGI) cancers in the study population. Our results do not support the hypothesis that iodine deficiency, as assessed by serum Tg, is associated with an increased risk of UGI cancers.

Introduction

An estimated 2 billion people, mostly in South Asia and sub-Saharan Africa, have insufficient iodine intake, and about 50% of Europeans have mild iodine deficiency.¹ Frequently, iodine deficiency occurs in mountainous regions where food is grown in iodine-poor soil or remote inland areas where no marine foods are eaten. Iodine deficiency can give rise to multiple health problems, including developmental delays and goiter. The highest iodine concentrations in the body are found in the thyroid, but iodine also concentrates in gastric tissue; therefore, iodine has been hypothesized to act as a local antioxidant which may reduce the risk of gastric cancer.²

UGI cancers, including stomach and esophageal cancers, together accounted for approximately 1.5 million diagnoses and 1.2 million deaths worldwide in 2007.³ Developed countries have seen significant decreases in stomach cancer incidence over the past decades, but many developing countries have not. Better sanitation has contributed to a decrease in the prevalence of infection with Helicobacter pylori (H pylori), a major risk factor for stomach cancer.⁴ In addition, dietary improvements, such as more access to fresh fruits and vegetables and a reduced need for cured, salted, and pickled foods due to refrigeration, may have contributed to the decreased risk in economically-developed areas.⁵

Stomach cancer incidence in the US has been steadily declining since the 1930s,⁶ and one factor thought to be contributing to this decline has been the introduction of iodine-fortified salt in the US in 1924. Epidemiologic studies have been limited and have produced mixed results. A 2007 study found that stomach cancer incidence rates in Poland also decreased after the introduction of obligatory household salt iodization.⁷ In a 2005 Turkish study, patients with stomach cancer had significantly more goiter and autoimmune thyroid diseases than control subjects, suggesting an association between stomach cancer and thyroid disorders.⁸ However, a small hospital-based case-control study in Italy found no relation between benign thyroid diseases and stomach cancer risk.⁹ One study suggested that urine iodine concentration and serum protein-bound iodine levels may be higher in patients with gastric cancer compared to healthy controls,¹⁰ whereas another study found that gastric cancer tissues contain lower iodine levels compared to surrounding normal tissue.¹¹

We previously found that self-reported goiter was associated with a significantly increased risk of gastric noncardia adenocarcinoma (GNCA); self-reported goiter was also associated with nonsignificantly increased risks of gastric cardia adenocarcinoma (GCA) and esophageal squamous cell carcinoma (ESCC) in the Linxian General Population Nutrition Intervention Trial (NIT) cohort in Linxian, China, an area with high rates of all of these cancers.¹² While these findings were consistent with the hypothesis that iodine deficiency may be associated with an increased risk of stomach cancer, self-reported goiter has low sensitivity as a measure of iodine deficiency. In addition, the reported association could have arisen because of confounding from other goitrogenic exposures that may also be stomach carcinogens. Therefore, to further test the association of iodine deficiency with risks of GNCA, GCA, and ESCC, we assessed serum Tg, a sensitive biomarker of iodine exposure,^13–14 in a case-cohort study nested in the same NIT cohort.¹⁵

Materials and Methods

Study cohort

Subjects were selected from the cohort of all participants in the Linxian General Population Nutrition Intervention Trial. The people of Linxian, China, have very high rates of ESCC and GCA and moderately high rates of GNCA; approximately 20% of Linxian residents die of these UGI cancers.¹⁵ Previous reports have provided a detailed description of the design, conduct, and results of the NIT intervention and follow-up.^15–17 Briefly, participants were 29,584 healthy adults aged 40–69 years, randomized to test the ability of four vitamin/mineral combinations taken daily for 5.25 years to affect esophageal and gastric cancer incidence and mortality. In 1985, one year prior to the start of the intervention, informed consent was obtained from all participants, and they were interviewed, given a physical examination, and had 10 ml of blood drawn. Serum specimens were separated, aliquoted, and stored at −80°C for future analyses. Throughout the intervention, local healthcare workers recorded cancer incidence and mortality data at monthly intervals, and in the 10 years after the trial, subjects were contacted monthly. To verify and confirm case status, American and Chinese experts reviewed pathology slides and/or x-rays available for 85% of the cases; Chinese experts reviewed cases without diagnostic materials and deaths due to other causes. Outcomes for this study were based on follow-up data collected from March 1986 to May 2001; <1% of the subjects were lost to follow-up.

Cases and subcohort

We recently completed a case-cohort study of serum pepsinogens which included approximately 300 GNCA, 600 GCA, and 300 ESCC cases that occurred during the same 15 years of follow-up and a subcohort of approximately 1000 randomly selected NIT subjects.¹⁸ Serum samples from 183 GNCA, 187 GCA, and 157 ESCC cases (with approximately equal numbers of males and females for each cancer) and 436 subcohort members were available for thyroglobulin measurements. Because the subcohort was a random sample of the cohort, some members developed cancer during follow-up. Thus, there were 192 GNCA cases (183 selected cases and 9 subcohort cases), 192 GCA cases (187 selected cases and 5 subcohort cases), and 193 ESCC cases (157 selected cases and 36 subcohort cases).

Serological assays

Because the presence of circulating Tg autoantibodies (TgAb) may cause falsely high or low Tg concentration measurements,¹⁹ all serum samples were first screened for the presence of TgAb using a highly sensitive, quantitative radioimmunoassay (RIA)system (Kronus, San Clemente, CA, performed at the University of Southern California Endocrine Services Laboratory, Los Angeles, CA). The RIA system had a detection limit of 1.0 IU/mL, and samples with TgAb concentrations more than 1.0 IU/mL were considered antibody positive.

For TgAb-positive samples, serum Tg concentrations were measured using a Tg RIA assay, employing CRM-457 standardization, with a functional sensitivity of 0.5 ng/mL, and a reference range for euthryoid subjects of 3–40 ng/mL.²⁰ This assay has been reported to give clinically useful results in TgAb-positive subjects.²¹ For TgAb-negative samples, serum Tg concentrations were measured using an immunoradiometric assay (IMA) system (Kronus, San Clemente, CA), with a functional sensitivity of 0.35 ng/mL and a reference range for euthyroid subjects of 0–60 ng/mL.

Fifty-six quality control samples, aliquoted from a single large serum pool of cohort members, were distributed among the assays. On the basis of these quality control samples, the coefficients of variation were 12.1% and 7.7% for the RIA and IMA assays, respectively. We previously published descriptions of the generation and analysis of data on H pylori serology²² and circulating pepsinogen I (PGI) and pepsinogen II (PGII)¹⁸ in these subjects.

Statistical methods

Because the serum Tg measurements between the two assays were not directly comparable and because the males and females were not equally represented among the two assays, we created exposure quartiles based on the distributions of serum Tg from subcohort members divided into sex- and assay-specific groups: 1) female, IMA; 2) female, RIA; 3) male, IMA; and 4) male, RIA. Those in the lowest quartile from each of the four groups were combined as the lowest overall quartile, and those in the other three quartiles from each of the four groups were similarly combined. Cases were then assigned a quartile based on the cutoffs from the serum Tg measurements of one of the four groups to which the case was classified (i.e., female, RIA). Iodine is inversely correlated with serum Tg concentrations, so those in the highest Tg quartile have presumably the lowest levels of iodine. The overall quartiles were used in the primary risk analyses; multivariate Cox proportional hazards models were used to estimate crude and adjusted hazard ratios (HRs) and 95% confidence intervals (95%). For each outcome, the models were fit using the case-cohort Cox models in Epicure Software (Hirosoft, Seattle, Washington, USA). All p-values reported are two-sided, and p-values less than 0.05 or confidence intervals that excluded 1.0 were considered statistically significant. The proportional hazards assumption was graphically examined and tested using time interaction terms, and no violations were detected.

Consistent with previous studies of UGI cancers in Linxian, age (years), history of smoking (yes or no), alcohol consumption (yes or no in the past 12 months), H pylori seropositivity, and the PGI/II ratio (>4, yes or no) were considered as potential confounders. Family history of UGI cancer and resident commune were also evaluated as potential confounders. Medians and interquartile ranges of the continuous variable (age) and numbers and percentages of the categorical variables (history of smoking, alcohol consumption, H pylori seropositivity, PGI/II ratio, family history, and commune) were calculated and reported for the subcohort and each cancer type.

Results

Table 1 shows the demographic characteristics and potential confounders in cancer cases and subcohort members. Compared to the subcohort, all subgroups of cancer cases were older on average.

Table I.

Characteristics of subcohort and cases of the Linxian General Population Nutrition Intervention Trial cohort

Characteristic	Subcohort (n=436)	Gastric non-cardia adenocarcinoma (n=192)	Gastric cardia adenocarcinoma (n=192)	Esophageal squamous cell carcinoma (n=193)
Age, in years, median (Q1-Q3)	52 (44–60)	57 (51–62)	56 (50–61)	56 (49–62)
Males, n (%)	228 (52)	96 (50)	96 (50)	97 (50)
BMI (kg/m2), median (Q1-Q3)	22 (20–24)	21 (20–23)	22 (20–23)	21 (20–23)
Diastolic blood pressure, median (Q1-Q3)	80 (75–90)	80 (75–90)	80 (75–90)	80 (70–90)
Systolic blood pressure, median (Q1-Q3)	130 (115–140)	130 (120–140)	130 (110–140)	130 (115–145)
Tobacco smokers, n (%)1	180 (41.4)	72 (37.5)	57 (29.8)	72 (37.3)
Alcohol drinkers, n (%)2	114 (26.2)	37 (19.3)	43 (22.5)	42 (21.8)
Family history of UGI cancer, n (%)	118 (27.1)	60 (31.3)	70 (36.8)	68 (35.2)
Helicobacter pylori seropositive, n (%)	317 (73.0)	143 (78.1)	160 (85.6)	146 (78.9)
Pepsinogen I:II ratio, median (Q1-Q3)	8.6 (6.2–13.3)	6.3 (4.5–10.4)	6.5 (4.8–10.3)	8.5 (5.7–14.2)
Tg-Ab+ status, n (%)	62 (14.2)	30 (15.6)	27 (14.1)	30 (15.5)

¹Ever smoked cigarettes for 6 or more months.

²Ever drink any alcoholic beverage in the past 12 months.

BMI, body mass index; Q1, first quartile; Q3, third quartile.

Based on the subcohort, 14.2% of the population was TgAb-positive. More women than men were TgAb-positive; 51 (25%) of the 208 females and 11 (5%) of the 228 males in the subcohort were TgAb-positive. Based on World Health Organization (WHO) guidelines,²³ 29.6% of the subcohort was iodine deficient; more females than males were iodine deficient (35.6% versus 24.1%, respectively).

Serum Tg was analyzed in the Cox proportional hazards models as sex- and assay-specific quartiles. The Tg measurements from the subcohort were used to assign the subjects to a quartile; Table 2 lists the number of subjects and the serum Tg cut-offs within each quartile. Due to some identical measurement results, the quartiles among the four groups were uneven (Table 2).

Table II.

Serum Tg quartile designation among the subcohort of the Linxian General Population Nutrition Intervention Trial cohort¹

	Serum Tg quartile
Sex, assay	Q1	Q2	Q3	Q4
Female, IMA, n	33	33	33	34
serum Tg (ng/ml)	< 10.7	10.7 to < 15.7	15.7 to < 29.7	≥ 29.7
Female, RIA, n	16	16	24	19
serum Tg (ng/ml)	< 8.0	8.0 to < 14.0	14.0 to < 20.0	≥ 20.0
Male, IMA, n	48	51	50	51
serum Tg (ng/ml)	< 8.3	8.3 to < 13.2	13.2 to < 20.7	≥ 20.7
Male, RIA, n	5	8	7	8
serum Tg (ng/ml)	< 6.0	6.0 to < 9.0	9.0 to < 17.0	≥ 17.0
Total, n (%)	102 (23.4)	108 (24.8)	114 (26.1)	112 (25.7)

¹Due to identical Tg measurement results, the quartiles among the four groups were uneven

We found that increased quartiles of Tg, the proxy for iodine deficiency, were not significantly associated with increased risks of UGI cancers; both crude and adjusted hazard ratios are listed in Table 3. Compared to participants in the lowest quartile, those in the highest quartile of serum Tg had a crude Hazard Ratio (HR) of 0.99 (95% confidence interval [CI] 0.61–1.61) for GNCA, 1.07 (CI 0.64–1.79) for GCA, and 0.82 (CI 0.51–1.30) for ESCC; the p-trends for GNCA, GCA, and ESCC were 0.67, 0.43, and 0.66, respectively. After adjusting for age, sex, smoking, alcohol drinking, H pylori status, PGI/II ratio, family history, and commune of residence, we found that increasing quartiles of serum Tg were not associated with an increased risk of any of the three cancers. Compared to participants in the lowest quartile, those in the highest quartile of serum Tg had an adjusted HR of 0.88 (95% CI 0.50–1.52) for GNCA, 1.14 (CI 0.63–2.05) for GCA, and 0.78 (CI 0.47–1.31) for ESCC; the p-trends for GNCA, GCA, and ESCC were 0.23, 0.14, and 0.17, respectively. Similar risks were found in analyses with males only, females only, dichotomous evaluation of serum Tg based on the WHO definition for iodine deficiency, and dichotomous evaluation of serum Tg of those in the highest quartile versus all participants in the remaining three quartiles.

Table III.

Hazard ratios and 95% confidence intervals from Cox regression models for the association between serum thyroglobulin quartiles and risk of upper gastrointestinal cancers in the Linxian General Population Nutrition Intervention Trial cohort

Cancer type and exposure	n (%)	Crude model, HR (95% CI)	Adjusted model1, HR (95% CI)
Gastric non-cardia adenocarcinoma
Q1	50 (26.0)	ref	ref
Q2	38 (19.8)	0.65 (0.39–1.09)	0.56 (0.31–1.01)
Q3	52 (27.1)	0.95 (0.58–1.54)	0.94 (0.54–1.61)
Q4	52 (27.1)	0.99 (0.61–1.61)	0.88 (0.50–1.52)
p-trend		0.67	0.23
Gastric cardia adenocarcinoma
Q1	41 (21.3)	ref	ref
Q2	43 (22.4)	0.93 (0.56–1.56)	0.84 (0.47–1.53)
Q3	62 (32.3)	1.40 (0.86–2.29)	1.53 (0.88–2.65)
Q4	46 (24.0)	1.07 (0.64–1.79)	1.14 (0.63–2.05)
p-trend		0.43	0.14
Esophageal squamous cell carcinoma
Q1	54 (28.0)	ref	ref
Q2	39 (20.2)	0.65 (0.40–1.06)	0.56 (0.33–0.96)
Q3	53 (27.5)	0.90 (0.57–1.42)	0.90 (0.55–1.47)
Q4	47 (23.3)	0.82 (0.51–1.30)	0.78 (0.47–1.31)
p-trend		0.66	0.17

¹Models were adjusted for sex, age, family history, commune of residence, drinking alcohol, smoking, H pylori status, pepsinogen I/II ratio.

Discussion

The Linxian General Population Nutrition Intervention Trial was conducted in an area of China with very high rates of both esophageal and gastric cancer. At baseline, about 2% of the study population self-reported a history of goiter.¹² However, this may have been an underestimate of iodine deficiency in this population because mild iodine deficiency may not manifest as a clinical goiter. Nevertheless, we found that self-reported goiter was significantly associated with an increased risk of future GNCA in this cohort. Therefore, we decided to further examine the hypothesis that iodine deficiency is associated with an increased risk of UGI cancers by evaluating a more direct and sensitive measure of iodine deficiency^13–14 in this population. In iodine sufficiency, a small amount of Tg (less than 10 ng/ml) is secreted into circulation, and in areas of endemic goiter, serum Tg levels increase.¹ According to WHO, individuals with insufficient iodine intake have serum Tg concentrations above 20 ng/ml.²³ Based on this cut-off, approximately 30% of our study population was considered iodine deficient.

Our results are not consistent with the hypothesis that iodine deficiency, as reflected by serum Tg levels, is associated with an increased risk of gastric cancer. Both crude and adjusted associations were not significant, and addition of potential confounders in the final model had little effect on the hazard ratios.

Our study has several strengths. To our knowledge, this is the first prospective evaluation using serum Tg levels as a direct indicator of iodine deficiency to examine the association of iodine deficiency and gastric or esophageal cancer. We used a well-defined population with increased risks for UGI cancers and very complete follow-up. In addition, the nested case-cohort design estimated the exposure of iodine deficiency in the population. We first screened all subjects for TgAbs and then used the most appropriate methods for measuring Tg concentrations in TgAb-positive and TgAb-negative subjects. And finally, in our analyses we adjusted for H pylori infection, a major cause of gastric cancer, and PGI/II ratio, an indicator for gastric fundic atrophy, both of which might possibly confound the iodine – cancer associations.

Our study also has several weaknesses. An individual’s serum Tg concentration integrates several factors, including the mass of differentiated thyroid tissue, any inflammation or injury to the thyroid gland that would cause the release of Tg into the circulation, and the level of thyroid-stimulating hormone (TSH) receptor stimulation.¹⁹ Although participants in the NIT study were screened to be generally healthy at baseline, some people who had elevated serum Tg may have had benign thyroid conditions. Also, iodine deficiency status may not be easily captured by one indicator, so other methods for assessing iodine status, including urinary iodine, serum thyroxine (T4), and serum TSH might produce different results. In addition, our prospective follow-up time was only 15 years, and some cancers may require more time to develop. And finally, as in all observational epidemiologic studies, failure to appropriately adjust for unmeasured or unknown confounders may have obscured the true relationship between iodine deficiency and the cancer outcomes.

The previously reported association between goiter and risk of gastric cancer in this population and the lack of association between iodine deficiency and risk of gastric cancer in the current study suggest the possibility that other goitrogens may be present which can increase the risk of cancer. Goiter can occur in people with Keshan disease,²⁴ and this population has been shown to have low selenium status.²⁵ Also, drinking water contaminants such as nitrates and polycyclic aromatic hydrocarbons are known to be goitrogenic and may be associated with UGI cancers. Further work that examines these and other goitrogens may improve our understanding of the etiology of UGI cancers.

Abbreviations used

GNCA

gastric noncardia adenocarcinoma

GCA

gastric cardia adenocarcinoma

ESCC

esophageal squamous cell carcinoma

Tg

thyroglobulin

UGI

upper gastrointestinal

TSH

thyroid-stimulating hormone

H pylori

Helicobacter pylori

HR

hazard ratio

CI

confidence interval

WHO

World Health Organization

BMI

body mass index

IQR

interquartile range

RIA

radioimmunoassay

IMA

immunoradiometric assay

NIT

Linxian General Population Nutrition Intervention Trial