Association Between Parkinson Disease and Risk of Cancer in Taiwan

Importance  Parkinson disease (PD) has been reported to be associated with a general reduced risk of cancer. These studies were mainly carried out in Western populations and little was known about associations in East Asians.

Objective  To analyze the association between PD and risk of cancer.

Design, Setting, and Participants  In this cohort study, the data were obtained from the Taiwan National Health Insurance Research Database, which contained information on approximately 24.7 million insured individuals. The cohort included individuals with newly diagnosed as having PD between 2004 and 2010. An age- and sex-matched systematic random-sampling method was used for subject selection in the reference non-PD cohort. Multivariate Cox proportional hazard regression analysis was used to determine the effects of PD on the risks of cancer, as shown by hazard ratios (HRs) with 95% CIs.

Main Outcomes and Measures  The Taiwan Population Census and National Cancer Registry Databases were used to identify patients with cancer. The last follow-up date was December 31, 2012.

Results  In 62 023 patients with PD, the HR for all subsequent cancers combined was 1.58 (95% CI, 1.50-1.65). Of the 19 types of cancer, Parkinson disease was not associated with breast, ovarian, or thyroid cancers. Increased HRs were found in the remaining 16 cancers, including malignant brain tumors (HR, 3.42; 95% CI, 1.84-6.38), gastrointestinal tract cancers (esophageal [HR, 1.81; 95% CI, 1.28-2.57], stomach [HR, 1.59; 95% CI, 1.30-1.94], colorectal [HR, 1.47; 95% CI, 1.31-1.65], liver [HR, 1.89; 95% CI, 1.67-2.14]; gallbladder [HR, 1.73; 95% CI, 1.16-2.57], and pancreas [HR, 1.48; 95% CI, 1.09-2.02]) (P < .05 for all comparisons), lung cancers (HR, 1.56; 95% CI, 1.38-1.76), some hormone-related cancers (uterine [HR, 1.83; 95% CI, 1.12-3.01], cervical [HR, 1.36; 95% CI, 1.05-1.76], and prostate [HR, 1.80; 95% CI, 1.52-2.13; P < .05 for all comparisons), urinary tract cancers (kidney and bladder cancers; HRs, 1.59 and 1.99, respectively; P < .001 for both comparisons), lymphoma and/or leukemia (HR, 1.62; 95% CI, 1.31-2.01), melanoma (HR, 2.75; 95% CI, 1.35-5.59), and other skin cancers (HR, 1.81; 95% CI, 1.46-2.23). For hepatocellular carcinoma, the highest HR resided in the 50- to 59-year-old group (HR, 2.57; 95% CI, 1.7-3.89).

Conclusions and Relevance  Our study concludes that PD is is associated with most cancers in Taiwan. Further studies are needed to clarify whether our findings can be applied to other East Asian populations. The differences between our study and most previous cohorts suggest the importance of ethnicity and environmental exposures in disease pathogenesis.

Parkinson disease (PD) is the most common neurodegenerative movement disorder and is caused by premature death of dopamine-containing neurons in the substantia nigra. It affects 1% of the population older than 60 years.¹ Although most PD cases are sporadic forms that result from a combination of polygenic inheritance and gene-environment interactions, a familial history of PD can be disclosed in approximately 10% of the cases.² Studies in families with PD led to the identification of 15 PD loci (PARK1-15).³ Several of these PARK genes, such as SNCA,^4,5PARKIN,^6-9PINK1,^10,11DJ-1,^12,13LRRK2,^14-16 and FBOX7,¹⁷ were found to play roles in both PD and cancers.

During the past 50 years, more than 25 epidemiological studies were done on the association between PD and cancers.¹⁸ On the one hand, most of these studies showed that people with PD had a decreased risk of cancer compared with those without.^19,20 The exception was a Swedish cohort²¹ including 11 786 patients with PD (hereinafter, PD patients) from 1954 to 2009 that showed a marginally increased risk of cancer in PD patients vs PD-free individuals. On the other hand, PD was variably disclosed to be associated with high rates of certain cancers in recent studies, although the results have been inconsistent to date. These cancers included melanoma, brain, breast, uterus, kidney, and prostate cancers.^16,21-24 Moreover, these studies were mostly done in Western populations, and little was known about associations in East Asians.

It is becoming clear that genetic backgrounds play important roles in a diverse array of disease pathogenesis.^3,25-27 Parkinson disease and lung cancer are 2 examples that demonstrate the importance of ethnicity in disease pathogenesis. LRRK2 G2019S mutation causes typical PD. The mutation occurs in approximately 40% of Arab Berbers with PD,²⁸ 20% of Ashkenazi Jewish patients,²⁹ but only 1% to 7% of PD patients who are of European origin.^30,31EGFR tyrosine kinase domain–activating mutations convert it into a driver oncogene in lung adenocarcinoma.^32-34 The percentage of EGFR-activating mutations approaches 50% in East Asian patients with lung adenocarcinoma but only 10% in white patients.^35,36 We therefore aimed to investigate whether the association between PD and cancers differs between East Asians and those of European origin. Because East Asians share similar genetic backgrounds,³⁷ we conducted the study using the Taiwan National Health Insurance (NHI) Research Database (NHIRD). This nationwide population-based follow-up cohort study examined whether PD is a risk or a protective factor among different cancer types.

Box Section Ref ID

The NHI is a single-payer universal health insurance program that has been operating since 1995, and more than 99% of the population are enrolled. The National Health Research Institute compiles all medical claims in the NHI program and releases the database for research purpose. Data for our cohort study were obtained from the NHIRD, which comprises comprehensive information of clinical visits for each insurant, such as demographic data, date of visits, diagnostic codes according to the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM), and prescriptions. The details of the NHIRD were described in the previous studies.³⁸ This study was approved by the ethical review board of the Taichung Veterans General Hospital.

The cohort included 133 322 individuals with PD newly diagnosed (ICD-9-CM code 332) between 2004 and 2010. The date of initial diagnosis was set as the index date for each person. A total of 66 385 patients with secondary parkinsonism (ICD-9-CM code 332.1), 350 with missing information on sex and age, and 4564 with malignant neoplasms before the index date were excluded. For the reference non-PD cohort, 2 subjects were selected using a systematic random-sampling method for each corresponding PD patient. The individuals in the PD and non-PD cohorts were matched for age, sex, and index year under the same exclusion criteria. Finally, a total of 62 023 PD patients and 124 046 without PD were included in this study (Figure 1).

We used the Taiwan Population Census and National Cancer Registry Databases³⁹ to identify patients with cancer in this study. The National Cancer Registry Database recorded information of cancer diagnostic date and histological types. Each study patient was followed until a diagnosis of malignant disease (ICD-9-CM code 140-208) was made, or until the patient was censored for death, lost to follow-up, withdrew from the database, or until the date of December 31, 2012, whichever came first.

The follow-up in person-years for an event was used for estimating the incidence density of cancer. Multivariate Cox proportional hazard regression models were used to determine the effects of PD on the risk of cancer, as shown by hazard ratio (HR) with 95% CIs. Furthermore, this study also estimated the effects of PD on the risk of specific cancer type and histological type. All analyses were performed using SAS statistical software (version 9.3; SAS Institute Inc), and results were considered statistically significant for 2-tailed P < .05.

Figure 2 shows the comparison of incidence density of cancers between the PD cohort and the non-PD reference cohort. The HR comparing the PD and the non-PD cohorts in association with all cancer outcomes combined was 1.58 (95% CI, 1.50-1.65; P < .001). The data showed that PD was not significantly associated with 3 of 19 cancers studied, including breast cancer (HR, 1.11; 95% CI, 0.90-1.37; P = .33), ovarian cancer (HR, 0.73; 95% CI, 0.37-1.44; P = .36), and thyroid cancer (HR, 1.10; 95% CI, 0.63-1.91; P = .75). For the other 16 cancers studied, PD was associated with an increased cancer incidence density. These included malignant brain tumors (HR, 3.42; 95% CI, 1.84-6.38; P < .001), gastrointestinal tract cancers (esophageal [HR, 1.81; 95% CI, 1.28-2.57], stomach [HR, 1.59; 95% CI, 1.30-1.94], colorectal [HR, 1.47; 95% CI, 1.31-1.65], liver [HR, 1.89; 95% CI, 1.67-2.14]; gallbladder [HR, 1.73; 95% CI, 1.16-2.57], and pancreas [HR, 1.48; 95% CI, 1.09-2.02]); P < .05 for all comparisons), lung cancers (HR, 1.56; 95% CI, 1.38-1.76; P < .001), hormone-related cancers (uterine [HR, 1.83; 95% CI, 1.12-3.01], cervical [HR, 1.36; 95% CI, 1.05-1.76], and prostate [HR, 1.80; 95% CI, 1.52-2.13]); P < .05 for all comparisons), urinary tract (kidney and bladder cancers; HR, 1.59 and 1.99, respectively; P < .001 for both comparisons), lymphoma and/or leukemia (HR, 1.62; 95% CI, 1.31-2.01; P < .001), melanoma (HR, 2.75; 95% CI, 1.35-5.59; P = .005) and other skin cancers (HR, 1.81; 95% CI, 1.46-2.23; P < .001).

We further analyzed liver cancers and malignant brain tumors by sex and age stratifications (Table 1). In both men and women, PD was a significant risk factor for liver cancers (HR, 1.97 and 1.85, respectively; P < .001 for both comparisons). When stratified by age, the highest HR resided in the group aged 50 to 59 years (HR, 2.57; 95% CI, 1.7-3.89; P < .001). Hepatocellular carcinoma (HCC) comprised most liver cancers in our study population, and the association between PD and intrahepatic cholangiocarcinoma was not significant (HR, 1.34; 95% CI, 0.82-2.18; P = .24). As to the malignant brain tumors, we showed that PD was a risk factor in both women and men (HR, 3.75 and 3.32, respectively; P = .02 and .002) and in both glioblastoma (HR, 3.07; 95% CI, 1.36-6.93; P = .007) and nonglioblastoma brain malignant neoplasms (HR, 4.34; 95% CI, 1.62-11.6; P = .003). When stratified by age, PD was associated with an increased risk only in patients 70 years or older (HR, 4.71; 95% CI, 1.93-11.5; P < .001).

To investigate the association between PD and lung cancers of different cell types, we further stratified lung cancers by sex, age, and histologic characteristics (Table 2). We showed that PD was a significant risk factor for lung cancers in both sexes (HR, 1.56 and P < .001 for both comparisons). The risk could not be observed in patients younger than 60 years, and the HRs increased with aging. The HR was 1.02 (95% CI, 0.24-4.29; P = .98) in the group younger than 50 years, 1.09 (95% CI, 0.63-1.91; P = .75) in the group aged 50 to 59 years, 1.44 (95% CI, 1.08-1.92; P = .01) in the group aged 60 to 69 years, and 1.47 (95% CI, 1.28-1.68; P < .001) in the group that was 70 years or older. When analyzing different histological types, PD was a risk factor for adenocarcinoma (HR, 1.61; 95% CI, 1.35-1.92; P < .001) and small cell carcinoma (HR, 1.79; 95% CI, 1.17-2.74; P = .007), but not squamous cell carcinoma (HR, 1.30; 95% CI, 0.99-1.72; P = .06) or other types of non–small cell lung cancers (HR, 1.10; 95% CI, 0.68-1.77; P = .71).

To determine if the factor of age contributed to the nonsignificant results between PD and breast cancers, PD and ovarian cancers, and PD and thyroid cancers, we further examined the hormone-related cancers by age stratification. As shown in Table 3 and Table 4, PD remained an insignificant factor with regard to thyroid cancers in both sexes and in each individual age group. Similarly, PD was not associated with the risk of breast or ovarian cancers in each individual age group. For uterine cancers, the risk effect of PD (overall HR, 1.83; 95% CI, 1.12-3.01; P = .02) was not observed when patients were stratified by age. For cervical cancers, the risk effect of PD (overall HR, 1.36; 95% CI, 1.05-1.76; P = .002) remained only in the patients who were 70 years or older (HR, 1.50; 95% CI, 1.03-2.20; P = .04). For prostate cancers, PD was associated with an increased cancer risk in all patients (HR, 1.80; 95% CI, 1.52-2.13; P < .001) and in patients 70 years or older (HR, 1.70; 95% CI, 1.42-2.05; P < .001).

To our knowledge, this is the first nationwide large-scale study that has investigated the association between PD and cancers in East Asians. Our study concludes that PD is overall a risk factor for cancer in Taiwan. There were only 3 types of cancers that were not associated with PD in our cohort: breast, ovarian, and thyroid cancers. Increased melanoma risk in PD patients was consistently reported in previous trials.^40-42 In line with these findings, our study also revealed an increase of melanoma risk in the PD cohort despite the fact that melanoma incidence was relatively low among Asians.⁴³ The excess melanoma risk might be the result of shared biosynthetic pathways between tyrosine and L-dopa.⁴⁴ Family members of PD patients were more likely to develop melanoma, and patients with melanoma and their family had an increased risk of PD.⁴⁵ Increased incidences of malignant brain tumors, prostate, uterine, and renal cancers in PD patients were variably reported in the previous studies, and these results were validated in our cohort.^23,24

Previous studies in Western populations described a reduced risk of smoking-related cancers in the PD patients.^21,22,24 In contrast, our study disclosed that PD was associated with increased cancer risks in 16 out of 19 cancers studied, including those categorized as smoking-related: esophageal, lung, and bladder cancers. In addition, PD was not associated with breast cancer risk in our cohort, whereas it was reported to be associated with increased breast cancer risks in a Danish cohort,²² an English cohort,²⁴ and a pooled analysis¹⁶ done in LRRK2-mutated PD patients (patients’ ethnicities included Ashkenazi and Sephardic Jews and whites). These discrepancies could be attributed to a combination of different genetic backgrounds and habitual and/or environmental exposures.

Smoking was the most discussed habitual exposure related to both PD and cancers. In Taiwan, the smoking percentage was 16.4% in the overall adult population, 29.2% in men, and 3.5% in women.⁴⁶ Despite the differences in smoking percentage between men and women, PD remained a risk factor for most cancers examined in our study with or without sex stratification.

Pesticide exposure was reported to be one of the etiologies of PD. The diverse associations between pesticide exposures and PD onset in different geographic areas were attributed to gene-environment interactions in several studies.^47,48 Although precise data on the differences of pesticide exposures between Taiwan and Western countries were unavailable, it was reasonable to speculate that this discrepancy contributed partly to the nearly opposite results between our study and the previous cohorts in view of PD and risk of cancer.

Esophageal cancers demonstrate how different habitual and/or environmental exposures lead to diverse processes of carcinogenesis. Adenocarcinoma is currently the most predominant cell type of esophageal cancers in several Western countries,⁴⁹ whereas most esophageal cancers in our cohort remain squamous cell carcinoma. Use of the combination of tobacco, alcohol, and areca nuts leads to a multiplicative carcinogenic effects for esophageal squamous cell carcinoma.⁵⁰ A high percentage of patients in Taiwan are users of this combination.⁵⁰ This may partly explain the disparate cancer types observed and the distinct associations between PD and risk of esophageal cancers found in different geographic areas.

Lung adenocarcinoma is a well-understood model that illustrates the importance of ethnicity in disease pathogenesis. The percentage of EGFR-activating mutations is unexpectedly high among East Asians with lung adenocarcinoma.^35,36 Similar stories were surprisingly found in triple-negative breast cancers (TNBCs). It was reported that EGFR-activating mutations were found in 11.4% of TNBCs in a Chinese cohort, whereas none were found in white cohorts.^51-53 No direct evidence to date links EGFR mutations to cancer risk in PD patients. Nevertheless, the opposite associations between lung and/or breast cancers and PD in our cohort vs white cohorts and the strikingly different EGFR-activating mutation rates between East Asians and whites provide a hint of possible linkage. Indeed, EGFR was reported to interact with several PARK genes, such as PARK2, LRRK2, and DJ-1.^54-56PARK2 mutation has been shown to be associated with familial lung cancer,⁹ and DJ-1 has been reported to play roles in breast cancer.^12,13 Further studies are needed to elucidate the role of EGFR in respect of cancer development among PD patients.

We also showed in our study an increased risk of HCC in the PD cohort, with the highest HR resided in the group aged 50 to 59 years. It is not clear why the highest risk occurs in the fifth decade, but the possible contribution by genetic alteration is worth considering. The association between PARK2 mutation and early-onset PD has been well established.³ Furthermore, somatic mutations of PARK2 were associated with multiple types of cancers.⁶ It has been revealed that HCC developed spontaneously in a Park2 knockout mouse model, and PARK2 has been suggested to be a tumor suppressor for HCC.⁷ There is a diverse heterogeneity in the rate of progression from a steady hepatitis carrier state to HCC development. The HCC risk distribution pattern among our PD cohort suggests a possible association between PARK genes and HCC. The results provide a new way of thinking in regard to the variable HCC onset age.

Our study revealed PD as a risk factor for most cancers examined. In breast, ovarian, and thyroid cancers, no association with PD was found. We could not identify PD’s protective role in any single cancer in our study. The finding was nearly opposite of those in most cohort studies or meta-analyses derived from Western populations.⁵⁷ There were, however, nationwide Swedish²¹ and English²⁴ cohort studies that partly supported our findings. In other words, the associations between PD and cancers were still nonconclusive, and the results derived from Western populations could not be applied to East Asians. Our findings suggested that different genetic backgrounds and habitual and/or environmental exposures combined resulted in the disparate outcomes. It is, therefore, reasonable to speculate that different alterations of PARK genes may contribute to distinct interactions with specific tumor-associated genes that end up in a diverse outcome of cancer development in different ethnic groups. The study highlights the fact that ethnicity matters in disease pathogenesis.

The study has a number of strengths. First, this cohort study was conducted from a nationwide population-based database containing more than 24 million subjects that allowed an accurate evaluation of the associations between specific diseases (eg, PD and cancers), age, and sex. Second, a cohort study would be the method of choice in the exploration of the association between PD and cancers. Randomized clinical trials were not feasible in this regard, and survival effects were considerable in case-control studies in that if PD-related mortalities developed before the onset of cancer, exposures to cancer would be underestimated in the case group. Finally, the health insurance program in Taiwan has a coverage rate exceeding 99% of the entire population, including patients with cancers, and all the medical records could be accurately traced in the NHIRD. Therefore, PD patients with or without cancers could hardly be lost to follow-up under our NHIRD program, and vice versa. In addition, the reference cohort was matched by the most important confounding factors—age and sex.

However, there were still limitations in this study. First of all, PD was defined according to the medical records in our study, and underascertainment of PD cases in the reference cohort was very likely. Either service-based studies or studies by record linkage system underestimated the incidence of PD because they exclude patients who do not seek medical advice.⁵⁸ This would make the HRs underestimated, indicating that PD patients would have an even higher risk for cancer than the current measurement. Second, smoking status was not available in the NHIRD record and was thus not included in our analysis. Third, although we selected the reference cohort by matching the most important confounding factors, age and sex, and the robust estimation of HRs was obtained by multivariate analyses, there were still others not considered. Finally, the hypothesis of this study was generated from a genetic point of view. Nevertheless, we have not yet investigated the PARK gene alterations of the PD patients who developed subsequent cancers in our cohort. Further studies are needed to elucidate the possible genetic correlations between these 2 disease entities.

Based on this nationwide study on the association between PD and cancer risk, we conclude that PD is a risk factor for most cancers in Taiwan. In our cohort, only breast, ovarian, and thyroid cancers show no association with PD. Further studies are needed to clarify whether our findings can be applied to other East Asian populations. The striking differences between our study and the previous studies in Western cohorts suggest the importance of ethnicity and environmental exposures in disease pathogenesis.

Corresponding Author: Pan-Chyr Yang, MD, PhD, Department of Internal Medicine, National Taiwan University College of Medicine, No. 1, Sec. 4, Roosevelt Road, Taipei 1061, Taiwan (pcyang@ntu.edu.tw).

Accepted for Publication: May 1, 2015.

Published Online: June 18, 2015. doi:10.1001/jamaoncol.2015.1752.

Author Contributions: Drs Yang and C.-H. Lin had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs P.-C. Yang and C.-H. Lin contributed equally to this work.

Study concept and design: P.-Y. Lin, C.-H. Lin, Yang.

Acquisition, analysis, or interpretation of data: P.-Y. Lin, Chang, Hsiao, Huang, C.-H. Lin.

Drafting of the manuscript: P.-Y. Lin.

Critical revision of the manuscript for important intellectual content: P.-Y. Lin, Chang, Hsiao, Huang, C.-H. Lin, Yang.

Statistical analysis: Chang, C.-H. Lin.

Obtained funding: C.-H. Lin, Yang.

Administrative, technical, or material support: P.-Y. Lin, Hsiao, C.-H. Lin, Yang.

Study supervision: C.-H. Lin, Yang.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by grants from the National Science Council (NSC103-2923-B-002-003) and from Taichung Veterans General Hospital, Taiwan (TCVGH-NHRI10405).

Role of the Funder/Sponsor: The National Science Council and Taichung Veterans General Hospital had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank The Collaboration Center of Health Information Application, Ministry of Health and Welfare and the Healthcare Service Research Center (HSRC) of Taichung Veterans General Hospital for statistical support.