Introduction

Patients with type 2 diabetes have a higher risk of cancer involving the breast, endometrium, stomach, colorectum, liver, pancreas, urinary bladder, and non-Hodgkin's lymphoma [1]–[8]. The mechanisms may be related to insulin resistance, hyperinsulinemia, proinflammatory status and increased oxidative stress [1], [2], [9]. With regards to thyroid cancer, there has been an increasing trend of its incidence over the past decades worldwide; and this has been ascribed to the increasing prevalence of insulin resistance [10].

Although insulin resistance may lead to an increased risk of thyroid cancer in patients with type 2 diabetes, our recent population-based study did not support a higher risk of thyroid cancer in these patients [11]. Recent studies suggested that antidiabetic drugs may play some role in the development of cancers. For example, metformin may reduce [12]–[16], while insulin and sulfonylureas may increase [1], [17]–[19] the risk of various types of cancer. On the other hand, the use of pioglitazone [20]–[22], but not rosiglitazone [21]–[23], may increase the risk of bladder cancer. Therefore, a lack of association between diabetes and thyroid cancer as observed in our previous study [11] can be a composite outcome of the interaction between the mechanisms that lead to a higher risk (e.g., insulin resistance) and the effects of the use of metformin that lead to a lower risk.

Whether metformin use can reduce the risk of thyroid cancer in patients with type 2 diabetes remains to be confirmed. Some recent in vitro studies suggested that metformin may inhibit the growth of thyroid cancer cells through inhibition of the mammalian target of rapamycin (mTOR) pathway by activating the AMP-activated protein kinase (AMPK) [24] or by diminishing the growth stimulation by insulin [25]. In humans, an earlier observational report of 4 cases showed that metformin might suppress thyroid stimulating hormone (TSH) without causing clinical hyperthyroidism [26], but this can not be confirmed in a later large sample-sized retrospective study comparing 250 patients on metformin therapy to 578 diabetic patients without metformin treatment [27]. A recent small clinical study conducted in 66 women with insulin resistance and thyroid nodular hyperplasia for 6 months suggested that metformin alone or better in combination with levothyroxine may reduce the size of the nodule [28].

All of the above in vitro and preliminary human studies suggested a potential usefulness of metformin on the prevention of thyroid cancer. However, large human observational studies are still lacking [16], [29]. Therefore, the purpose of the present study was to evaluate whether metformin use in Taiwanese patients with type 2 diabetes could affect the risk of thyroid cancer by using the National Health Insurance (NHI) database obtained from the whole nation.

Materials and Methods

A retrospective cohort analysis using the NHI databases including all patients with a diagnosis of diabetes during the period from 1996 (the earliest database available) to 2009 in Taiwan was conducted. The study using the NHI database was approved by an ethic review board of the National Health Research Institutes with registered approval number 99274. Written informed consent from the participants was not required according to local regulations because the identification information of the individuals was scrambled and de-identified prior to analysis for the protection of privacy.

Since March 1995 a compulsory and universal system of health insurance (the so-called NHI) was implemented in Taiwan. All contracted medical institutes must submit computerized and standard claim documents for reimbursement. More than 99% of citizens are enrolled in the NHI, and >98% of the hospitals nationwide are under contract with the NHI. The average number of annual physician visits in Taiwan is one of the highest around the world, at approximately 15 visits per year per capita in 2009.

The databases contain detailed records on every visit for each patient, including outpatient visits, emergency department visits and hospital admission; and include principal and secondary diagnostic codes, prescription orders, and claimed expenses. Diabetes was coded 250.00–250.93 and thyroid cancer 193, based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).

We first retrieved the databases of all patients who had been diagnosed as having diabetes and/or under treatment with either oral anti-diabetic agents or insulin during the period of 1996–2009 from the whole nation (n = 3,756,511). The selected entry date was 1 January 2006. After excluding patients who had a diagnosis of diabetes after the year 2006 (n = 773,801), patients who held a Severe Morbidity Card as having type 1 diabetes (n = 7182, in Taiwan, patients with type 1 diabetes were issued a so-called “Severe Morbidity Card” after certified diagnosis and they were waived for much of the co-payments), patients having a diagnosis of thyroid cancer before 2006 (n = 8,252), those who died (n = 169,509) or withdrew from the NHI (n = 23,784) before entry date, duplicated identification number (n = 216), unclear information on date of birth or sex (n = 14,284), and diabetic patients without any reimbursement record after the entry date (n = 1,556,791), a total of 1,414,723 patients with a diagnosis of type 2 diabetes and/or under therapy with oral anti-diabetic agents or insulin were recruited.

Those who had ever been prescribed metformin before entry date were defined as ever-users; and never-users were defined as those who had never been prescribed metformin before entry date. To evaluate whether a dose-response relationship could be seen between metformin and thyroid cancer, the tertile cutoffs for the following parameters were applied: 1) cumulative duration of therapy in months; and 2) cumulative dose in mg.

Baseline characteristics of ever- and never-users of metformin at entry were compared by Chi square test. Age and all comorbidities and covariates were determined as a status/diagnosis before the entry date. The ICD-9-CM codes for the comorbidities were [11]–[14], [30]: nephropathy 580–589, hypertension 401–405, chronic obstructive pulmonary disease (a surrogate for smoking) 490–496, stroke 430–438, ischemic heart disease 410–414, peripheral arterial disease 250.7, 785.4, 443.81 and 440–448, eye disease 250.5, 362.0, 369, 366.41 and 365.44, obesity 278, dyslipidemia 272.0–272.4, benign thyroid disease 240–246, and cancer other than thyroid cancer 140–208 (excluding 193). Medications included sulfonylurea, insulin, acarbose, pioglitazone, rosiglitazone, statin, fibrate, angiotensin-converting enzyme inhibitor and/or angiotensin receptor blocker, calcium channel blocker, aspirin, ticlopidine, clopidogrel, dipyridamole and non-steroidal anti-inflammatory drugs (excluding aspirin).

The incidence density of thyroid cancer was calculated for ever-users and never-users and for different subgroups of exposure. The numerator for the incidence was the number of patients with incident thyroid cancer during the 4-year follow-up from January 1, 2006 to December 31, 2009, and the denominator was the person-years of follow-up. For ever-users, the follow-up duration was either censored at the date of thyroid cancer diagnosis or at the date of the last record of the available reimbursement databases in individuals without incident thyroid cancer. In the lack of information on the vital status or migration of the patients, the last reimbursement record may serve as a surrogate because patients who die or migrate out of the country should be withdrawn from the NHI in Taiwan. For never-users, the follow-up was censored at the date of metformin initiation or thyroid cancer diagnosis or the last reimbursement record, depending on whichever occurring first. This ensured no exposure to metformin throughout the whole follow-up period until censor in the referent group of never-users.

Cox proportional hazards regression was performed to estimate the hazard ratios for thyroid cancer for ever-users versus never-users, and for the various subgroups of dose-response parameters. The following fully adjusted models were created with consideration of all covariates mentioned above: 1) Models I conducted in the whole original cohort; 2) Models II conducted after excluding users of both metformin and sulfonylurea, in order to avoid the potential bias resulted from multi-collinearity because a very high proportion of metformin users were also users of sulfonylurea; and 3) Models III conducted after excluding users of antidiabetic drugs other than metformin, in order to create a “cleaner” comparison for metformin only versus diabetic patients without use of antidiabetic drugs (i.e., on diet control only).

Additional sensitivity analyses were conducted by creating the following Cox regression models to estimate the fully adjusted hazard ratios for metformin users versus non-users: 1) In subgroups of men, women, age <50 years and age ≥50 years, respectively; 2) Excluding patients with a diagnosis of cancer other than thyroid cancer; 3) Excluding users of insulin, sulfonylurea, and insulin and/or sulfonylurea, respectively; 4) Additional adjustment for potential detection examinations including thyroid sonography, thyroid aspiration and thyroid function test (described previously [30]) that might have led to differential detection of thyroid cancer between ever-users and never-users of metformin.

Analyses were conducted using SAS statistical software, version 9.3 (SAS Institute, Cary, NC). P<0.05 was considered statistically significant.

Results

Among the patients 317,508 (22.4%) were on diet control without any pharmacological intervention, 907,599 (64.2%) were using oral antidiabetic drugs only, 179,847 (12.7%) were using oral antidiabetic drugs plus insulin, and 9,769 (0.7%) were using insulin only.

Table 1 compares the baseline characteristics between ever- and never-users of metformin. All variables differed significantly between the two groups. Ever-users are characterized by older age distribution, higher proportion with a diabetes duration >5 years, higher proportions of all comorbidities except with a lower proportion of a diagnosis with other cancer, and higher proportions of using other medications.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics between never-users and ever-users of metformin.

https://doi.org/10.1371/journal.pone.0109852.t001

Table 2 shows the incidences of thyroid cancer between ever-users and never-users of metformin, and among the different categories of the dose-response parameters for metformin exposure. During follow-up, a total of 2,297 (0.16%) patients developed thyroid cancer. Among them, 1,614 were never-users and 683 were ever-users of metformin at entry. The incidence rate in never- and ever-users of metformin was 87.33 and 24.09 per 100,000 person-years, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Incidence of thyroid cancer by metformin exposure at entry.

https://doi.org/10.1371/journal.pone.0109852.t002

Table 3 shows the fully adjusted hazard ratios with regards to different categories of metformin exposure in Models I, II and III. In the models evaluating the overall hazard ratios for ever-users versus never-users, all of the hazard ratios showed a significantly lower risk associated with metformin use. In the models evaluating the dose-response exposure to metfomrin, a lower risk could be seen in all tertiles of the dose-response parameters, with significant P-values for the trend. The dose-response effect in Models I was not as remarkable as in Models II, suggesting potentially biased estimates resulting from the high correlation between use of metformin and sulfonylurea in Models I.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Metformin exposure at entry and hazard ratios for thyroid cancer.

https://doi.org/10.1371/journal.pone.0109852.t003

Table 4 shows the fully adjusted hazard ratios for thyroid cancer in the additional sensitivity analyses. The results suggested that the protective effect of metformin on thyroid cancer was not restricted to a certain group of age or sex; and was not affected by excluding users of insulin, sulfonylurea, and insulin/and or sulfonylurea, respectively, by previous diagnosis of other cancers or by potential detection examinations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Additional sensitivity analyses for metformin and thyroid cancer.

https://doi.org/10.1371/journal.pone.0109852.t004

Discussion

Observational studies suggested that metformin may be associated with a reduced risk of some cancer involving the bladder, breast, liver, and colorectum [14]–[16]. However, whether metformin use can be associated with a reduced risk of thyroid cancer has not been investigated previously [16], [29]. This is the first observational study that clearly demonstrated a protective effect of metformin on thyroid cancer risk in patients with type 2 diabetes (Tables 2, 3 and 4).

Several randomized clinical trials are being conducted to evaluate the potential usefulness of metformin as an adjuvant to other chemotherapeutic agents on some solid cancers like the breast, endometrial, prostate, and lung cancer [31]. However, clinical trials focusing specifically on the effect of metformin on thyroid cancer are still lacking. This study provided an important rationale for future in-depth investigation on metformin for the prevention and treatment of thyroid cancer.

A large number (n = 1,556,791) of the patients with a diagnosis of diabetes did not have reimbursement data after study entry and had been excluded from the analyses. This could be a main source of bias in the study. It is assumable that this population of excluded diabetic patients might have mainly composed of patients being on diet control without any pharmacological intervention. The consistent finding in the comparison of metformin-treated patients to patients on diet control alone conducted in Models III of Table 3 would obviate the concern of this potential bias.

The present study suggested that a significantly lower risk of thyroid cancer could be observed with a cumulative duration of 9 months or a cumulative dose of 263,000 mg (Table 3). This time window for a protective effect of metformin may be supported by a small clinical trial showing that metformin use for 6 months may significantly reduce the size of thyroid nodules [28], and by another study showing that the adjusted hazard ratio for prostate cancer-specific mortality was 0.76 (95% confidence interval: 0.64–0.89) for each additional 6 months of metformin use [32].

If insulin resistance is the underlying cause of thyroid cancer as advocated by Gursoy [10] and metformin treatment in patients with type 2 diabetes would reduce the incidence of thyroid cancer (Tables 2, 3 and 4), it is expected that the incidence of thyroid cancer should be increased gradually from obesity, impaired glucose tolerance through the early stage of type 2 diabetes when they were put on diet control without pharmacological intervention, but the risk would decrease when metformin was initiated. However, such a temporal change in thyroid cancer incidence becomes more complicated when the disease progresses to the development of diabetes complications, when other clinical comorbidities set in, or when multiple drugs are used. Because most patients with type 2 diabetes require multiple antidiabetic drugs for controlling their blood glucose [33], the joint effects and interactions among the various antidiabetic drugs are interesting and important. However, these are out of the scope of the present study and await future investigations.

The anti-cancer effects of metformin may probably involve the activation of the AMPK, inhibition of the mTOR pathway, and inhibition of insulin-like growth factors [34]. All of these have been demonstrated in in vitro studies conducted in thyroid cancer cell lines [24], [25], in women with insulin resistance and benign thyroid nodules [28] and in patients with type 2 diabetes [35]. The suppression on TSH by metformin observed in humans [26], [36] may also contribute to the lower risk of thyroid cancer in metformin users, although this beneficial effect can not be confirmed by another recent study [27].

Leptin, a hormone produced by adipocytes and regulates energy balance in the brain, signals the expression of thryotropin-releasing hormone, leading to increased secretion of TSH [37]. Therefore, one of the possible mechanisms explaining the inhibitory effect of metformin on TSH secretion may be through its leptin inhibition [37]. Recent basic researches also suggested that metformin may sensitize thyroid cancer cells to oxidative stress [38], inhibit thyroid cancer stem cells [25], [39], cause cell cycle arrest at G0/G1 phase [25], induce apoptosis [25], inhibit colony formation [25], inhibit stimulation of mitogen-activated protein kinase by insulin [25], and inhibit growth of doxorubicin-resistant thyroid carcinoma cells [25]. The growth inhibitory effect of metformin is not only demonstrated in thyroid carcinoma cells, a recent study also demonstrates such a growth inhibitory effect on medullary thyroid cancer cells [24]. Metformin may have an advantage of inducing growth arrest specifically in cancer stem cells through the inhibition of Akt, which is not affected in differentiated cells [40], [41].

There are several strengths in the present study. First, we made a special request to include all patients diagnosed as having diabetes from the NHI database covering the whole period since its availability in 1996. Such an approach avoided the possibility of selection bias and insufficient cases of thyroid cancer for subgroup analyses. In the absence of a special request, the usually provided NHI database is a 1 million random sample of the general population, and the number of diabetic patients derived from such a 1 million sample is surely too small for the analyses of the association between the use of anti-diabetic drugs and some cancers. Second, the consistency of a lower risk of thyroid cancer in metformin users in the sensitivity analyses strengthened the preventive role of metformin on thyroid cancer. Third, the database included outpatients and inpatients, and we included diagnoses from both sources. The use of medical records also reduced the bias related to self-reporting.

The limitations of the study may include a lack of the histological types of thyroid cancer for analysis. However, because thyroid papillary cancer represents 80.5% and 87.1% of thyroid cancer in men and women, respectively, in Taiwan [42], the findings should better be applied to thyroid papillary cancer. Second, although some cases of thyroid cancer may have been misclassified, such an occurrence was probably low because labeled diagnoses should be printed on all prescriptions handed out to patients in Taiwan. Mislabeling of a cancer diagnosis would not be acceptable to the patients when they saw the diagnosis. Third, ionizing radiation has been implicated as a risk factor for thyroid cancer [43]. A confounding effect due to the lack of such data collection could not be excluded. Fourth, we did not have biochemical data for evaluating their impacts.

In summary, this study supports a protective effect of metformin on the development of thyroid cancer in Taiwanese patients with type 2 diabetes. The uses of metformin for reversing the increasing trend of thyroid cancer incidence in the general population and for the treatment of thyroid cancer are worthy of further investigation.