Introduction

Parkinson’s disease (PD) is a common neurodegenerative disease with primary clinical features of resting tremor, rigidity, bradykinesia, and postural instability. In addition to the motor symptoms of PD, there are also non-motor symptoms such as dysfunctions of smell, sleep, somatic sense, the autonomic nerve system, and cognition. Although Lewy bodies (LBs)/Lewy neurites (LNs) deposits and dopaminergic cell loss in the substantia nigra are key components of the pathology of PD, various etiologies have also been proposed as pathophysiological mechanisms underlying the development of PD, including genetics, endogenous factors, and environmental risk factors [1,2,3,4]. Recently, evidence of an association between PD and the gut has been reported, such as the appearance of LB in the gut, gut dysbiosis, and the presence of inflammation in the enteric nervous system (ENS).

The well-known pathological findings of PD include intracellular deposits of LB/LN that are composed primarily of alpha-synuclein. The medulla oblongata and olfactory bulb are the initial regions affected by inclusion bodies during the progression of PD [5, 6]. It has been proposed that unmyelinated vagal preganglionic neurons might play an important role in PD by providing a route for the pathology to affect the peripheral nervous system (PNS) as well as central nervous system (CNS) [7]. Furthermore, unidentified pathogens might enter the CNS via the PNS along the so-called “gut-to-brain” axis [8, 9]. Many studies have demonstrated that PD-related lesions, such as deposits of LBs/LNs, occur in the CNS and autonomic nerve system in the gastrointestinal tract, heart, and bladder [1, 10,11,12,13,14,15]. Gastrointestinal dysfunction is a prominent non-motor symptom of PD. For example, weight loss, oropharyngeal dysfunction, dysphagia, gastric dysfunction, colonic dysmotility, and anorectal dysfunction are common gastrointestinal complications of PD [16]. Alpha-synuclein deposits in the ENS can be detected before the onset of PD, even in normal people [11, 17, 18].

There are over 100 trillion microbes in the human gut. The normal microbiome in the gut is believed to play a role in the barrier function of the intestine without acting as pathogens. Recent studies have suggested that there is a bidirectional pathway between the gut and the brain, the so called “brain–gut–microbiome (BGM) axis”, which could explicitly affect brain activity [19]. For example, changes in bidirectional interactions within the BGM axis may result in gastrointestinal and neurological disorders, such as PD [20, 21]. Moreover, inflammatory bowel disease (IBD), such as ulcerative colitis and Crohn’s disease, is also linked to PD in terms of involving cytokines, related animal models, and LRRK2 gene studies [22,23,24].

Based on the results of previous studies of the gut and PD, we speculated that colectomy may be associated with the development of PD by potentially altering the BMG axis. To assess this hypothesis, we investigated the risk of PD associated with colectomy, using data from the Korea National Health Insurance Service (NHIS) program collected from 2002 to 2015, and the risk of PD.

Materials and methods

Data acquisition

The NHIS is a universal health insurance program that offers medical care coverage to all residents of South Korea (http://nhiss.nhis.or.kr); almost all Korean residents are registered. Likewise, all medical care institutions in South Korea must be registered in this program for insurance claims. The NHIS–National Sample Cohort (NHIS-NSC) involves approximately 1 million NHIS and Medical Aid program subscribers, approximately 2% of all Koreans, who were extracted by a stratified sampling method in 2002. This 14-year cohort (2002–2015) of NSC subscribers was tracked in terms of socioeconomic variables (residence, year and month of death, cause of death, and income level) and medical treatment details (health examinations, medical care history, and medical care institutions), allowing for long-term observations and investigations of causal relationships among the variables. Within the NSC, diseases are registered using the Korean Classification of Disease, sixth edition (KCD-6), which was modified from the International Classification of Disease, 10th revision (ICD-10), for use in the NHIS and medical care institutions in South Korea [25]. For medical institutions to make a claim to the NHIS for a specific medical practice, the corresponding distinct treatment and procedure codes for the claim, which correspond to individual KCD-6 codes, must be used.

This population-based matched cohort study conducted using the NHIS-NSC dataset (NHIS-2018-2-197) was approved by the Institutional Review Board of Kangwon National University Hospital (KNUH-2018-06-004).

Analysis strategy

Colectomy was identified using the procedure codes for claims that included total colectomy (Q2672 and QA672), subtotal colectomy (Q1261 and Q1262), hemi-colectomy (Q2671 and QA671), colectomy-segmental resection (Q2673 and QA673), and/or colectomy with colostomy (Q2679 and QA679) from 2002 to 2009. The colectomy group was subdivided into five groups according to the surgical methods described above. The index date for these subjects in the colectomy group was the date the colectomy was performed.

The possible causes of colectomy were inferred manually by two investigators (SH Lee and CM Lee) based on the disease codes registered at the time point adjacent (± 50 days) to the colectomy. Based on the causes of surgery, the colectomy group was also subdivided into six groups: malignant neoplasm, benign tumor, disease of appendix, non-infective inflammation, injury, and other disease of the intestines.

To ensure an observation period of at least 6 years, the colectomy group included subjects aged 40 years or older who underwent colectomy between 2003 and 2009. Patients who underwent colectomy in 2002 or 2010–2015 (n = 105), died within 1 year (n = 4), or had a previous PD diagnosis prior to the index date or within 1 year after colectomy (n = 23) were excluded from the analyses.

The NHIS-NSC database contains information on 1,108,369 subscribers from 2002 to 2015. Since PD is a progressive neurodegenerative disorder influenced by age, we selected subscribers who were aged 40 years or older in 2003 (n = 588,235). The occurrence of PD was defined by the code G20 based on the ICD-10 protocol; subscribers with secondary parkinsonism (G21) or atypical parkinsonism (G22 and G23) were excluded.

In addition to information about the diagnosis and surgical procedures of interest, data regarding the sex, age, body mass index (BMI), and diabetes and hypertension statuses of each subject were extracted [26, 27]. When multiple BMI measurements were available for a single subscriber, the BMI closest to the time of registration was used for the present analyses. The BMI measurements were subcategorized as underweight (< 18.5 kg/m2), normal weight (18.5–2.9 kg/m2), or overweight/obese (≥ 23 kg/m2) [28]. Of the subjects in the exposed group, 185 did not have BMI information and were excluded. Thus, the present study analyzed the data of 511 subjects who underwent colectomy (Fig. 1).

Fig. 1
figure 1

Strategy for the selection of the colectomy and control groups

Full size image

For each subject in the colectomy group, 10 control subjects with no history of colectomy were selected randomly and matched in terms of age, sex, and index date. Subjects in the control group were given the same index date as the subjects with a history of colectomy. Subjects in both groups were followed from the index date until the occurrence of PD, death, or December 31, 2015.

Statistical analyses

The present study estimated the prevalence rates of colectomy and PD in 2003 and the incidence rates of colectomy per 100,000 person-years from 2003 to 2009.

Pearson’s Chi square test was conducted to compare differences in PD occurrence between the colectomy and control groups. Fisher’s exact tests were performed to compare the difference in PD occurrence among the subgroups according to the causes and surgical methods of colectomy. A Cox proportional hazards model was used to calculate adjusted hazard ratios (HR) and 95% confidence intervals (CI) to determine whether colectomy is an independent risk factor for PD, after adjusting for sex, age, diabetes, hypertension, and BMI. Survival was defined as the time from the index date until the diagnosis of PD, censoring due to death, or the end of the study (December 31, 2015); subjects who were never diagnosed with PD by the end of the study were treated as censored. Cumulative incidence curves with 95% confidential intervals for the colectomy and control groups were estimated.

All statistical analyses were performed using the statistical package SAS for Windows, ver. 9.4 (SAS; Cary, NC).

Results

In 2003, the prevalence of patients with PD per 100,000 individuals aged 40 years or older was 155.75. The prevalence rate of PD according to age and sex gradually increased up to 80 years of age (Fig. 2). The crude incidence rates of colectomy and PD per 100,000 person-years were 18.39 and 57.88, respectively, in 2003. Table 1 summarizes the incidence rates of colectomy and PD according to age from 2003 to 2015.

Fig. 2
figure 2

Prevalence of Parkinson’s disease according to sex and age in 2003

Full size image
Table 1 Incidence rates of colectomy and Parkinson’s disease according to age from 2003 to 2015
Full size table

According to NSC, 511 people underwent colectomy from 2003 to 2009. Of these, 359 underwent hemi-colectomy, 104 colectomy-segmental resection, 30 colectomy with colostomy, 16 total colectomy, and two subtotal colectomy. The causes of colectomy were 395 malignant neoplasms, followed by 35 benign tumors, 16 disease of the appendix, 8 non-infective inflammation, 3 injuries, and 54 other diseases of the intestines. However, there was no significant difference in the frequency of PD according to the methods or causes of colectomy (p > 0.05, Table 2).

Table 2 Comparison of PD frequency among groups according to surgical methods (A) and causes (B) of colectomy
Full size table

The frequency of colectomy was higher in subjects with diabetes than in those without diabetes (p < 0.05). There were no other significant differences in age, sex, hypertension, or BMI between the colectomy and control groups (p > 0.05). Throughout the entire observational period, there were 10 (2.0%) total occurrences of PD in the colectomy group and 61 (1.2%) in the control group (p = 0.14). Although there was no significant difference between the two groups in terms of PD occurrence, there tended to be more cases of PD in the colectomy group compared with the control (Table 3, Fig. 3). A higher risk of PD was found in subjects with colectomy than in those without colectomy, after adjusting for sex, age, and BMI with a multivariate Cox regression model (adjusted HR 1.987; 95% CI 1.015–3.891). Further adjustment for sex, age, BMI, diabetes, and hypertension also showed a higher risk associated with the adjusted HR (1.962; 95% CI 1.002–3.840) (Table 4).

Table 3 Baseline characteristics of subjects with and without colectomy
Full size table
Fig. 3
figure 3

Cumulative incidence curves for the colectomy and control groups

Full size image
Table 4 Multivariate Cox proportional hazards analysis
Full size table

Discussion

This study examined a population-based matched cohort of approximately 1 million subscribers to medical insurance and aid in Korea. Without exception, all medical service providers must register with the NHIS program for reimbursement, and thus, the frequencies of colectomy and PD identified during the observational period in the present cohort are likely reliable because medical care institutions are required to submit a claim to the NHIS for colectomy. In 2003, the prevalence of patients with PD per 100,000 individuals aged 40 years or older was 155.75. The prevalence of PD in the present study was somewhat different than those reported previously in South Korea and other Asian countries [29,30,31,32]. These discrepancies may be due to differences in diagnostic criteria and/or statistical approaches. In comparing subjects with and without colectomy, the occurrence of PD tended to be more frequent in the colectomy group. In addition, after adjusting for confounding factors such as hypertension and diabetes, this study found that a history of colectomy was significantly associated with a higher risk of PD.

This study had a large sample size and a 6–12-year observation period, which allowed the investigation of causal relationships between colectomy and PD occurrence. It is possible that the observation period was insufficient for the development of clinical PD, seeing that constipation could precede PD diagnosis by 10 years with wide range of the interval in previous studies of constipation of PD [33,34,35,36]. However, cohort studies with relatively short follow-up durations (about 6 years) reported a positive association between constipation and PD [37,38,39]. In addition, a neuropathology study predicted a 4.6-year preclinical state [40], and a neuroimaging study predicted that the mean preclinical period in PD is unlikely to exceed 7 years, although this report was mainly about CNS disease [41]. Therefore, we believe that our follow-up period was sufficient for assessing the risk of developing PD after colectomy.

There have been many conflicting reports regarding the relationships between the risk of PD and vascular risk factors, including hypertension and diabetes [42,43,44]. Diabetes, hypertension, and obesity are also reported to be associated with colorectal cancer [26, 27]. Regarding the association between PD and colorectal cancer risk, some reported that PD had a negative association with the risk of colorectal cancer [45,46,47], while others reported the opposite [48] or no association [49]. Therefore, it is possible that colon cancer or risk factors for colon cancer, rather than colectomy itself, are related to the development of PD: colectomy is likely to be an interim outcome that reflects several related prior diseases. However, we could not find a difference in the subgroup analysis among the groups according to the causes of colectomy. Moreover, the association between colectomy and PD was confirmed after adjusting for diabetes, hypertension, and BMI.

The highlight of this study is that we examined a hypothesis regarding the brain–gut and BGM axes with real-world data [19,20,21]. This analysis revealed that colectomy was associated with the development of PD for the first time.

Although the pathogenic mechanisms of PD are not fully understood, recent studies suggest that the protective effect of vagotomy on PD based on hypotheses of the transmission of causative agents from the PNS along the vagus nerve [50, 51]. Similarly, it seems reasonable to assume that a colectomy would block the spread of causative agents to the CNS and thereby lower the risk of developing PD. Interestingly, however, the results were the opposite of what was expected by the hypothesis mentioned above.

Two important inferences can be made based on the present results. First, we did not find a significant difference in the occurrence of PD according the method, depending on the extent of gut removal. This result did not support the possibility that the extent of gut removal may influence the spread of the causative agent of PD. Based on this result, we thought that there might be other mechanisms in the gut besides the direct propagation of a causative agent. Interestingly, some studies have shown an increased risk of PD associated with appendectomy, suggesting direct invasion by a neurotrophic enteric pathogen or trigger processes for retrograde propagation of alpha-synuclein to the brain [52], while there was one report with the opposite result [53]. It is unclear whether there is a shared mechanism between appendectomy and colectomy in the development of PD. However, it may suggest that changes in the anatomical structures of the colon due to surgical procedure play an important role in the brain–gut interaction.

Second, colectomy may result in changes in the environment of the colon, which could affect the composition and habitat of the microbiome. Several current studies have investigated the CNS, behaviors, neurotransmitters, and neurotrophic factors based on hypotheses regarding the BGM axis [19,20,21, 54,55,56,57]. One study analyzed fecal microbiomes and found that the abundance of Prevotellaceae in feces was significantly lower in PD patients than in controls, whereas the abundance of Enterobacteriaceae was associated with the severity of gait impairments [58]. Another study reported that PD patients exhibit alterations in colonic microbiota and dysbiosis [59]. Furthermore, another study revealed the presence of small intestinal bacterial overgrowth (SIBO) in patients with PD [60]. Alterations in intestinal anatomy or motility could result in SIBO [61], and some studies have reported an association between SIBO and motor fluctuations in PD, including poor motor function, prolonged off latencies, and delayed on latencies [62, 63]. A higher prevalence of SIBO was observed in patients with colectomy responding to antibiotics [64] and in an experimental rat model [65]. Based on the abovementioned studies, we speculate that alterations in the microbiome resulting from colectomy might be associated with the development of PD.

This study was limited in terms of a lack of clinical PD information related to medical history, social history, and clinical manifestations such as severity and clinical stage in the NHIS-NSC data. According to previous studies, the diagnostic accuracy of PD is approximately 75–90%, which was also a potential bias of selection of PD [66,67,68,69,70]. Additionally, depending on the date of surgery, the duration of observation in the case group was 6–12 years, which may necessitate caution in interpreting our results because of long duration until the development of PD.

In conclusion, we demonstrated that colectomy is associated with the development of PD. This suggests that the colon plays an important role in the pathophysiological mechanisms of PD.