A Case–Control Study of Levothyroxine and the Risk of Colorectal Cancer

Abstract

Levothyroxine is a synthetic T₄ hormone commonly used to treat thyroid disease. Increased incidence of mostly autoimmune thyroid disease has been associated with breast and other malignancies, and thyroid hormone levels might also be associated with risk of colorectal cancer (CRC). In this population-based matched case–control study (2566 pairs) of CRC in northern Israel, use of levothyroxine for at least 5 years was assessed using structured interviews and validated by prescription records. The analysis included use of statins, aspirin, and hormone replacement therapy; CRC family history; physical activity; vegetable consumption; ethnicity; age; and sex. All statistical tests were two-sided. The use of levothyroxine was associated with a statistically significantly reduced relative risk of CRC (odds ratio = 0.59, 95% confidence interval = 0.43 to 0.82, P = .001). This association remained statistically significant after adjustment for age, sex, use of aspirin and statins, sports activity, family history of CRC, ethnic group, and level of vegetable consumption (odds ratio = 0.60, 95% confidence interval = 0.44 to 0.81, P = .001). No statistically significant interactions were seen between use of levothyroxine and aspirin, statins, or hormone replacement therapy.

CONTEXTS AND CAVEATS

Prior knowledge

Thyroid disease may be associated with elevated risk of breast, stomach, and pancreatic cancers, but its relationship with colorectal cancer (CRC) is unknown. Levothyroxine is a synthetic T₄ hormone used to treat hypothyroidism.

Study design

The association between use of levothyroxine and occurrence of CRC was studied in a population-based case–control study of incident CRC patients and control subjects in the general population in northern Israel.

Contribution

The use of levothyroxine for at least 5 years was associated with a statistically significantly reduced risk of CRC. The negative association was stronger for women than for men, particularly postmenopausal women not using hormone replacement therapy.

Implications

Along with aspirin and statins, levothyroxine may be an effective agent against CRC. However, the effect may be moderated in women by the use of hormone replacement therapy.

Limitations

The use of retrospective data may introduce recall bias. The slightly unbalanced participation rates of case patients and control subjects could also result in selection bias. The study was underpowered to test the association in premenopausal women.

From the Editors

Colorectal cancer (CRC) is the second most commonly diagnosed cancer in Israel, with approximately 3200 new cases and 1500 deaths in 2006 (1). Thyroid disease may be associated with elevated risk of breast (2–9), stomach (10), and pancreatic cancers (11), but its relationship with CRC is unknown. Increased risk of cancer has been associated with elevated antithyroid antibodies (2,3,5,11,12), possibly reflecting impaired immunoregulation caused by malignant cells. Hypothyroidism, use of thyroid medications, and low to normal levels of the thyroid hormone T₄ have been associated with increased risk of postmenopausal breast cancer (4). In addition, major pathways in the development of CRC have been shown to be negatively regulated by thyroid hormone (13). Levothyroxine is a synthetic T₄ hormone that is used to treat patients with hypothyroidism. We studied the possible association between use of levothyroxine and occurrence of CRC in a population in northern Israel.

The Molecular Epidemiology of Colorectal Cancer (MECC) study is a population-based case–control study of incident CRC patients in northern Israel and control subjects in the general population matched by year of birth, sex, residence (defined by primary clinic location), and ethnic group (Jewish vs non-Jewish). All subjects were interviewed in person to obtain information about their personal and family history of cancer, reproductive history, medical history, medication use, and health habits, including a previously described dietary questionnaire (14). The variables included in this analysis were use of levothyroxine for at least 5 years; use of statins for at least 5 years; daily aspirin use for at least 3 years; ever use of hormone replacement therapy (HRT); first-degree family history of CRC; physical activity, yes or no (15); vegetable consumption of at least five servings per day (16); ethnicity; age; and sex. All variables are known risk factors for CRC (see Supplementary Methods, available online). We matched self-reports of levothyroxine use against prescription records when available. Unconditional and conditional stepwise backward elimination logistic regression models were used. Statistical analyses were performed using SPSS v 15.0 (SPSS, Inc, Chicago, IL). All statistical tests were two-sided. Institutional review boards at Carmel Medical Center and the University of Michigan approved all procedures.

Data from 2648 patients and 2566 control subjects (2566 matched pairs) were available for analysis. The response rate was 70% of all eligible case patients and 59% of eligible control subjects. Prescription records were validated for 177 (93.7%) of the 189 self-reported levothyroxine users for whom computerized data were available, with similar agreement for case patients and control subjects.

As expected, Ashkenazi Jews were overrepresented among case patients compared with non-Ashkenazi Jews (17) and Arabs, in keeping with known CRC rates in Israel (18). Family history (first-degree relatives), sports participation, vegetable consumption, and use of aspirin, statins, and HRT differed statistically significantly between case patients and control subjects (Table 1).

Table 1.

Comparison of major study variables between case patients and control subjects in the Molecular Epidemiology of Colorectal Cancer study in Israel^*

Variable	Unpaired			Paired
	Case patients (n = 2648)	Control subjects (n = 2566)	P†	Case patients (n = 2566)	Control subjects (n = 2566)	P†
	No. (%)	No. (%)		No. (%)	No. (%)
Sex			.99			1.00
Men	1342 (50.7)	1301 (50.7)		1301 (50.7)	1301 (50.7)
Women	1306 (49.3)	1265 (49.3)		1265 (49.3)	1265 (49.3)
Age, mean, y	69.9	70.6	.06	69.8	70.6	.03
Ethnicity			.39			.30
Non-Jews	348 (13.1)	317 (12.4)		342 (13.3)	317 (12.4)
Jews	2300 (86.9)	2249 (87.6)		2224 (86.7)	2249 (87.6)
Jewish ethnicity			<.001			<.001
Ashkenazi (% of Jews)	1793 (78.0)	1607 (71.5)		1734 (78.0)	1607 (71.5)
Sephardi (% of Jews)	458 (19.9)	602 (26.8)		444 (20.0)	602 (26.8)
Mixed (% of Jews)	49 (2.1)	40 (1.8)		46 (2.1)	40 (1.8)
Levothyroxine use for ≥5 y	77 (3.0)	129 (5.1)	<.001	74 (3.0)	129 (5.1)	<.001
Family history of CRC among first-degree relatives	304 (12.0)	204 (8.1)	<.001	295 (12.0)	204 (8.1)	<.001
Sports activity	769 (29.9)	1063 (41.5)	<.001	750 (30.1)	1063 (41.5)	<.001
Vegetable intake of ≥5 servings per day	1402 (56.5)	1552 (60.9)	.002	1359 (56.5)	1552 (60.9)	.002
Aspirin use daily for ≥3 y	371 (14.5)	584 (22.9)	<.001	364 (14.6)	584 (22.9)	<.001
Statin use for ≥5 y	173 (6.8)	293 (12.5)	<.001	168 (6.8)	293 (12.5)	<.001
Ever HRT use (% of all women)	70 (6.6)	126 (10.8)	.001	69 (6.8)	126 (10.8)	.001

^*Case patients were incident CRC patients and control subjects were members of the general population matched by year of birth, sex, residence (defined by primary clinic location), and ethnic group (Jewish vs non-Jewish). The 3- and 5-year cut points for use of aspirin, levothyroxine, and statin were selected to capture long-term use. Percentage discrepancies are because of missing values in some of the variables. CRC = colorectal cancer; HRT = hormone replacement therapy.

^†P values for comparison between case patients and control subjects calculated using the two-sided χ² test and the t test for age.

Levothyroxine use was reported more by control subjects than by case patients (5.1% vs 3.0%, respectively, P < .001) and was associated with a statistically significant reduction in risk of CRC (odds ratio [OR] = 0.59, 95% confidence interval [CI] = 0.43 to 0.82, P = .001). Levothyroxine use differed statistically significantly between male and female control subjects (2.0% vs 8.2%, respectively, P < .001), reflecting known differences in hypothyroidism by sex (19). The negative association between levothyroxine use and CRC in men was not statistically significant (OR = 0.75, 95% CI = 0.42 to 1.36, P = .35), whereas the negative association in women was strong (OR = 0.54, 95% CI = 0.38 to 0.75, P < .0001). Estrogen and/or progestin replacement treatments in peri/postmenopausal women have been shown to be associated with reduced risk of CRC (20–24). HRT use in our study was associated with a statistically significantly reduced risk of CRC (OR = 0.64, 95% CI = 0.50 to 0.83, P = .001). In a univariate analysis, levothyroxine use was associated with a statistically significantly reduced risk of CRC in postmenopausal women (OR = 0.53, 95% CI = 0.37 to 0.74, P < .001) overall; this effect remained statistically significant among non-HRT users (OR = 0.49, 95% CI = 0.33 to 0.73, P < .001) but not among HRT users (OR = 0.90, 95% CI = 0.35 to 2.36, P_interaction = .19). A statistically significant interaction was previously found between use of HRT and aspirin with respect to the risk of CRC (24) and was included in the final model in this study.

After adjustment for potential confounders, the association between levothyroxine use and reduced risk of CRC remained statistically significant (OR = 0.60, 95% CI = 0.44 to 0.81, P = .001) (Table 2). The crude and adjusted odds ratios for levothyroxine use were similar in the conditional and unconditional analyses. In the fully adjusted model for postmenopausal women, levothyroxine use was associated with a statistically significant reduction in risk of CRC (OR = 0.60, 95% CI = 0.41 to 0.87, P = .008) (Table 2).

Table 2.

Final multivariable logistic regression models of the association of use of levothyroxine with colorectal cancer risk in the Molecular Epidemiology of Colorectal Cancer study in Israel^*

Variable	All participants (conditional) (n = 4008)	P	All participants (unconditional) (n = 4617)	P	Men (unconditional) (n = 2340)	P	Women (peri/postmenopausal) (unconditional) (n = 1982)	P
	OR (95% CI)		OR (95% CI)		OR (95% CI)		OR (95% CI)
Levothyroxine use, ≥5 y
No	1.00 (referent)	.001	1.00 (referent)	.001	†		1.00 (referent)	.008
Yes	0.59 (0.43 to 0.82)		0.60 (0.44 to 0.81)				0.60 (0.41 to 0.87)
Aspirin use, daily, ≥3 y
No	1.00 (referent)	<.001	1.00 (referent)	<.001	1.00 (referent)	<.001	1.00 (referent)	<.001
Yes	0.58 (0.49 to 0.69)		0.57 (0.49 to 0.67)		0.57 (0.46 to 0.70)		0.57 (0.44 to 0.75)‡
Statins use, ≥5 y
No	1.00 (referent)	<.001	1.00 (referent)	<.001	1.00 (referent)	.01	1.00 (referent)	<.001
Yes	0.61 (0.49 to 0.77)		0.62 (0.50 to 0.77)		0.70 (0.52 to 0.93)		0.53 (0.38 to 0.74)
HRT use ever
No	NR		NR		NR		1.00 (referent)	<.001
Yes							0.60 (0.43 to 0.84)‡
Family history of CRC, first-degree relatives
No	1.00 (referent)	<.001	1.00 (referent)	<.001	1.00 (referent)	.005	1.00 (referent)	<.001
Yes	1.56 (1.26 to 1.94)		1.57 (1.29 to 1.92)		1.50 (1.13 to 2.00)		1.75 (1.30 to 2.36)
Sports activity
No	1.00 (referent)	<.001	1.00 (referent)	<.001	1.00 (referent)	<.001	1.00 (referent)	<.001
Yes	0.60 (0.52 to 0.69)		0.62 (0.55 to 0.71)		0.68 (0.57 to 0.80)		0.57 (0.46 to 0.70)
Vegetables, ≥5 servings/d
No	1.00 (referent)	.003	1.00 (referent)	.001	1.00 (referent)	.001	†
Yes	0.81 (0.71 to 0.93)		0.81 (0.71 to 0.91)		0.75 (0.63 to 0.89)
Age, per year	NR		0.99 (0.99 to 1.00)	.003	0.99 (0.98 to 1.00)	.03	0.99 (0.98 to 0.99)	<.001
Sex
Men	NR		1.00 (referent)	.003	NR		NR
Women			0.84 (0.74 to 0.94)
HRT × aspirin	NR		NR		NR		2.72 (1.12 to 6.58)	.026

^*Odds ratios shown for variables in final models only. CI = confidence interval; CRC = colorectal cancer; HRT = hormone replacement therapy; NR = not relevant; OR = odds ratio.

^†Dropped out of final model.

^‡This model includes an interaction of HRT × aspirin; therefore, the aspirin effect is among non-HRT users and the HRT effect is among non-aspirin users.

Our data indicate a strong inverse association between risk of CRC and long-term use of levothyroxine. Although there are few reports relating thyroid hormones and CRC, thyroid dysregulation is reproducibly associated with risk of breast cancer (2–5), and its relationship to pancreatic (11) and gastric (10,25) cancers has been discussed.

Preclinical data (13) suggest that transactivation by the Bcat/tcf1 transcription factor of the Cyclin D promoter is negatively regulated by thyroid hormone. It also showed that the T3-bound receptor inhibits the transcription of cyclin D1through the Tcf/Lef-1 site, which is positively regulated by the Wnt-signaling pathway. Because the Wnt pathway is a central pathway in causation of sporadic CRC, as determined in humans (26), hormones that inhibit this pathway could potentially be considered as cancer prevention agents. Thyroid hormones have also been shown to inhibit estrogen-induced degradation of the estrogen receptor-α protein in breast cancer (27), induce formation of aberrant thyroid hormone receptors in hepatocytes (28,29), suppress pituitary tumor–transforming 1 (PTT1) activity (30), regulate cell cycle progression and proliferation in estrogen responsive breast ductal cell lines, and raise p53 levels (31). HRT (21,32,33) and thyroid hormones are associated with reduction in risk of CRC in our study, but both are suspected to be associated with an elevated risk of breast cancer (5–9,34), thus suggesting a common mechanism. In vitro studies (35,36) suggested cross talk between the estrogen receptor and thyroid hormone receptors; thyroid hormone T₃ increased estrogen receptor-α levels, resulting in increased progesterone receptor levels, prolactin production, and tumor growth (35). It has also been shown that chronic estradiol treatment reduces total T₄ levels in ovariectomized rats (36). Use of HRT and levothyroxine by women with hypothyroidism was shown to change the levels of thyroid-stimulating hormone and free T₄ (37) and to require additional levothyroxine intervention (38). This change in thyroid-stimulating hormone and free T₄ levels is probably not mediated through changes in sex hormone–binding globulin protein levels (39) and is noted with oral but not with transdermal HRT (40). Slattery et al. (41) demonstrated that the presence of the R allele of the estrogen receptor-β gene was associated with reduced risk of CRC among HRT users but increased risk among nonusers. The same association may account for our reported differences in risk of CRC magnitude associated with levothyroxine use in HRT users and nonusers.

One of the limitations of our study is possible recall bias because exposure data were collected retrospectively. However, because a levothyroxine–CRC risk association was not known, participants were unlikely to differentially report on this. Furthermore, we were able to validate self-reported levothyroxine use through a prescription database, although we did not have data regarding dose or duration of use. Other limitations were that the slightly imbalanced participation rates of case patients and control subjects could result in selection bias and that the study was underpowered to study associations in premenopausal women.

In conclusion, we found that long-term use of levothyroxine is associated with a reduction in risk of CRC after adjustment for other known risk factors. Together with aspirin and statins, these agents provide powerful tools for combating CRC. However, the disappearance of the risk reduction effect of aspirin and levothyroxine in HRT users calls for further study and caution in considering HRT for CRC prevention.

Funding

National Cancer Institute (1R01CA81488 to S.B.G. and G.R.).