CNS infections, sepsis and risk of Parkinson’s disease

Abstract

Background Neuroinflammation may play an important role in the aetiology of Parkinson’s disease (PD); however, little is known about infections in relation to future PD risk.

Methods We conducted a register-based nested case–control study in Sweden to examine infections of the central nervous system (CNS) and sepsis in relation to PD with 18 648 patients and 93 240 matched controls. We defined the index date as the date of first recorded PD diagnosis in the Swedish Patient Register.

Results Overall, PD patients were more likely to have a previous hospitalization for CNS infections [odds ratio (OR) = 1.5, 95% confidence interval (CI): 1.2–1.9] or sepsis (OR = 1.6, 95% CI: 1.4–1.7) than controls, largely due to hospitalizations in the year before PD identification (CNS infections: OR = 3.0, 95% CI: 1.6–5.7; sepsis: OR = 3.5, 95% CI: 3.0–4.0). However, we found that subjects with multiple CNS infections at least 5 years before the index date had higher PD occurrence than those without CNS infections (OR = 3.3, 95% CI: 1.4–8.2), whereas the corresponding OR for sepsis was 1.4 (95% CI: 0.8–2.4). After the index date, PD patients were more likely to be hospitalized for CNS infections [hazard ratio (HR) =1.8, 95% CI: 1.2–2.7] or sepsis (HR = 2.2, 95% CI: 2.1–2.4) than controls.

Conclusions This study provides preliminary evidence for an association between CNS infections, but not sepsis, and a higher future risk of PD. It also shows that PD patients were more prone to CNS infections and sepsis than controls.

Introduction

Recent genome-wide association studies have identified some of the genetic links to late-onset sporadic Parkinson’s disease (PD);^1–3 however, for most PD cases, the causes as well as the underlying mechanisms remain poorly understood. On the other hand, accumulating evidence in the past two decades suggests that microglia-mediated neuroinflammation may play an important role in PD aetiology.⁴ Once initiated, this inflammatory process may form a vicious cycle that eventually leads to progressive neuron loss. Furthermore, murine experiments⁵ showed that a single peripheral injection of the endotoxin lipopolysaccharide (LPS) induces lasting microgliosis and death of dopaminergic neurons, raising the possibility that even a single severe peripheral infection may initiate processes ultimately leading to PD.

Few epidemiological studies have examined the link between infectious diseases and PD^6–8 and results are inconsistent. Infections affecting the central nervous system (CNS) and sepsis are good candidates for investigation because the former directly involves the target organ and the latter is the human analogue of LPS intoxication. However, to the best of our knowledge, neither CNS infections nor sepsis has been examined in relation to PD in epidemiological studies. We therefore explored the hypothesis that CNS infections or sepsis increased future PD risk, using data from nationwide population and health registers in Sweden.

Methods

Swedish Patient Register

Since 1964–65, the Swedish National Board of Health and Welfare has compiled data on hospital discharge diagnoses coded according to the International Classification of Diseases (ICD) (ICD-7 before 1969, ICD-8 for 1969–86, ICD-9 for 1987–96 and ICD-10 since 1997). This Inpatient Register started in a few counties and became nationwide in 1987. The Outpatient Register was established in 2001, covering outpatient specialist visits in Sweden. Both the Inpatient and Outpatient Registers (hereby referred as the Patient Register) record one primary diagnosis and up to eight secondary diagnoses for each clinical visit, along with information on the dates of contact and discharge. Since most PD patients are unlikely to be hospitalized directly after diagnosis, we limited our case identification to the period January 1, 2001 to December 31, 2007, in order to include cases from the Outpatient Register.

Study design

We conducted a nested case–control study based on all Swedish residents who were born in Sweden between 1901 and 1970 and participated in the Swedish Population and Housing Survey in 1990. We restricted the study to subjects born before 1970 in order to exclude patients with young-onset parkinsonism, which is more likely to have a monogenic cause. We used the unique Swedish national registration number to link each study subject to the Patient, Causes of Death and Emigration Registers. For those who had a clinical visit for PD, died or emigrated out of Sweden, the follow-up ended at the time of the event. PD cases were defined as individuals with one or more hospital contacts for PD (ICD-10 code: G20) as either the primary or a secondary diagnosis. Prevalent PD patients who had a record in the Inpatient Register before January 1, 2001, were not eligible. For each case, we used the date of first contact for PD as the index date. From the Patient Register, we identified a total of 18 648 subjects with a first-ever PD-related clinical contact between 2001 and 2007. When PD was first identified as a secondary diagnosis and later as a primary diagnosis, we used the date from the first secondary diagnosis as the index date. The validity of inpatient PD diagnoses has been evaluated against clinical examinations among participants of a Swedish twin study; this validation showed a sensitivity of 72.7% and a positive predictive value of 70.8%.⁹ PD diagnoses from the Outpatient Register have not yet been validated. Using incidence density sampling, we randomly selected 93 240 controls (5 controls per case) who were individually matched to the PD cases on year of birth and gender. Persons eligible to be controls were alive in Sweden and free of PD on the index date. The study protocol was approved by the Regional Ethical Vetting Board in Stockholm, Sweden.

Exposure assessment

Our exposure of interest was hospitalization due to CNS infections or sepsis identified from the Inpatient Register. To ensure a systematic identification of such hospitalizations, we restricted consideration to those occurring between January 1, 1987, when the Inpatient Register obtained nationwide coverage, and the index date. We used ICD codes for CNS infections and sepsis according to a previously published study.¹⁰ Briefly, CNS infections were defined as ICD-9 codes: 006F, 013, 036A, 036B, 045–049, 052B, 053A, 053B, 054D, 054H, 055A, 056A, 062–064, 072B, 072C, 094, 136C and 320–325 and ICD-10 codes: A06.6, A17, A39, A80–A89, B00.3, B00.4, B01.0, B01.1, B02.0, B02.1, B05.0, B05.1, B06.0, B22.0, B26.1, B26.2, B37.5, B38.4, B43.1, B50.0, B58.2, B60.2, G00–G08 and R29.1. For sepsis, we used the narrow criterion of sepsis as described in the same paper¹⁰ with ICD-9 codes: 036C–036E, 036X, 038, 084, 112F, 117D, 286G and 999D and ICD-10 codes: A02.1, A04.0–A04.3, A39–A41, A42.7, A48, A90–A99, B37.7, B38.7, B39.3, B40.7, B41.7, B42.7, B44.7, B45.7, B46.4, B95–B99, D65 and T80.2. The validity of using these codes to identify cases hospitalized with CNS infections or sepsis in Sweden had been previously evaluated against records from the intensive care units.¹⁰ The results of this validation showed high accuracy for CNS infections with a sensitivity of 84.0% and specificity of 99.7%. For sepsis, the sensitivity was lower (∼43%), but the specificity was still high (∼97%).

Statistical analysis

We evaluated the associations between previous hospitalizations for CNS infections or sepsis and a later PD occurrence using conditional logistic regression models. In the case of multiple hospitalizations for CNS infections or sepsis, the first event was used in the primary analysis. Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were controlled for year of birth and gender. The exposure was first defined as ever hospitalization for CNS infections or sepsis from January 1, 1987, to the index date. We further categorized such hospitalizations according to the lag time between the event and PD ascertainment (<1, 1–4, 5–9 and ≥10 years) in order to explore the temporal relationship between infections and PD occurrence and to evaluate the possibility of infections occurring as a result of PD (the so-called ‘reverse causality’) rather than as its cause. Finally, we conducted additional analyses examining risk of hospitalization due to CNS infections or sepsis after PD identification, using Cox proportional hazards models.

We examined the importance of multiple hospitalizations (once vs more than once) and cumulative days of hospital stays for respective infections (<1 week vs ≥1 week) in relation to later PD occurrence. In these latter analyses, we only counted hospitalizations at least 5 years before the index date to reduce the potential impact of reverse causality. If consecutive hospital admissions were identified (e.g. transfers among different clinics), the admission date of the first hospitalization and the discharge date of the last hospitalization were used to define the length of hospitalization.

In the analyses, we first included all cases and matched controls and then separately according to the diagnosis type (primary or secondary only) of cases. Due to the lack of accurate information on the date of PD diagnosis or onset, our analyses likely included prevalent cases. To reduce the potential impact from this, we further conducted a sensitivity analysis by restricting to cases and controls with index date after 2003, assuming that cases identified in the first 2 years of the Outpatient Register were more likely to be prevalent cases. All statistical analyses were performed with SAS 9.1 (SAS Institute, NC, USA) and the statistical tests were two tailed with α = 0.05.

Results

Table 1 shows the characteristics of PD cases and controls. Most cases (73.8%) had PD as the primary diagnosis. Compared with these cases, those with PD only as a secondary diagnosis were substantially older. Eighty-six PD patients (0.5%) had a history of hospitalization for CNS infections, most (n = 53, 61.6%) occurring ≥5 years before the index date (Table 1). In comparison, more PD patients (n = 799, 4.3%) had been hospitalized for sepsis; however, most of the sepsis hospitalizations (n = 628, 78.6%) occurred within 5 years before the index date.

Table 1.

Characteristics of subjects of a population-based nested case–control study of PD in Sweden, 2001–07

	Cases
Variables	All cases	Primary diagnosis	Secondary diagnosis only	Controls
Number of subjects	18 648	13 762	4886	93 240
Age (years) at the index date, mean (SD)	73.8 (10.0)	71.7 (9.9)	79.9 (7.7)	73.8 (10.0)
Male, n (%)	10 799 (57.9)	8045 (58.5)	2754 (56.4)	53 995 (57.9)
CNS infections before the index date
No	18 562	13 710	4852	92 953
Yes	86	52	34	287
Years before the index date
<1	15	7	8	25
1–4	18	10	8	53
5–9	24	19	5	91
≥10	29	16	13	118
Number of infections at least 5 years before the index date
Once	45	32	13	197
≥2 times	8	3	5	12
Duration of hospital stay at least 5 years before the index date (weeks)
<1	22	16	6	84
≥1	31	19	12	125
Sepsis before the index date
No	17 849	13 409	4440	90 662
Yes	799	353	446	2578
Years before the index date
<1	350	112	238	518
1–4	278	138	140	1178
5–9	120	69	51	611
≥10	51	34	17	271
Number of infections at least 5 years before the index date
Once	155	95	60	824
≥2 times	16	8	8	58
Duration of hospital stay at least 5 years before the index date (weeks)
<1	71	47	24	382
≥1	100	56	44	500

SD: standard deviation.

Compared with controls, PD cases were 50% more likely to have a previous hospitalization for CNS infections (OR = 1.5, 95% CI: 1.2–1.9) (Table 2). When examined by the lag time between hospitalization for CNS infections and the index date, the OR was 3.0 for a lag time of <1 year, 1.7 for 1–4 years, 1.3 for 5–9 years and 1.2 for ≥10 years. A higher OR was also observed for previous hospitalization for sepsis (OR = 1.6, 95% CI: 1.4–1.7); however, higher PD occurrence was only found for sepsis experienced during the 1-year period before the index date. The OR was 3.5 for a lag time of <1 year, 1.2 for 1–4 years, 1.0 for 5–9 years and 1.0 for ≥10 years.

Table 2.

ORs and 95% CIs of CNS infections or sepsis in relation to PD: a nested case–control study in Sweden, 2001–07^a

		Diagnosis type
Variables	All cases (n = 18 648)	Primary diagnosis (n = 13 762)	Secondary diagnosis only (n = 4886)
CNS infections
No	1.0 (ref.)	1.0 (ref.)	1.0 (ref.)
Yes	1.5 (1.2–1.9)	1.3 (1.0–1.8)	1.9 (1.3–2.8)
Years before the index date (years)
<1	3.0 (1.6–5.7)	2.3 (1.0–5.7)	4.0 (1.6–10)
1–4	1.7 (1.0–2.9)	1.5 (0.7–3.1)	2.0 (0.9–4.6)
5–9	1.3 (0.8–2.1)	1.4 (0.9–2.4)	1.0 (0.4–2.6)
≥10	1.2 (0.8–1.8)	1.0 (0.6–1.7)	1.8 (1.0–3.4)
Number of infections at least 5 years before the index date
Once	1.1 (0.8–1.6)	1.1 (0.8–1.7)	1.1 (0.6–2.1)
≥2 times	3.3 (1.4–8.2)	1.9 (0.5–7.1)	6.2 (1.7–23)
Duration of hospital stay at least 5 years before the index date (weeks)
<1	1.3 (0.8–2.1)	1.2 (0.7–2.1)	1.8 (0.7–4.5)
≥1	1.2 (0.8–1.8)	1.2 (0.7–1.9)	1.4 (0.7–2.6)
Sepsis
No	1.0 (ref.)	1.0 (ref.)	1.0 (ref.)
Yes	1.6 (1.4–1.7)	1.1 (1.0–1.2)	2.6 (2.3–2.9)
Years before the index date (years)
<1	3.5 (3.0–4.0)	1.7 (1.3–2.0)	7.2 (5.9–8.8)
1–4	1.2 (1.0–1.4)	0.9 (0.8–1.1)	1.8 (1.4–2.2)
5–9	1.0 (0.8–1.2)	0.8 (0.7–1.1)	1.3 (1.0–1.8)
≥10	1.0 (0.7–1.3)	1.1 (0.7–1.6)	0.8 (0.5–1.4)
Number of infections at least 5 years before the index date
Once	1.0 (0.8–1.1)	0.9 (0.7–1.1)	1.1 (0.8–1.4)
≥2 times	1.4 (0.8–2.4)	1.3 (0.6–2.9)	1.5 (0.7–3.2)
Duration of hospital stay at least 5 years before the index date (weeks)
<1	1.0 (0.7–1.2)	0.9 (0.7–1.3)	1.0 (0.7–1.6)
≥1	1.0 (0.8–1.3)	0.9 (0.7–1.2)	1.2 (0.9–1.7)

^aAdjusted for year of birth and gender.

Subjects with more than one hospitalization for CNS infections ≥5 years before the index date had higher occurrence of PD (OR = 3.3, 95% CI: 1.4–8.2) (Table 2) than subjects without such a hospitalization. The risk elevation was more apparent for cases with PD as the secondary diagnosis (OR = 6.2, 95% CI: 1.7–23, five exposed cases), compared with cases with PD as the primary diagnosis (OR = 1.9, 95% CI: 0.5–7.1, three exposed cases). On the other hand, multiple hospitalizations for sepsis were not associated with a higher risk of PD (Table 2).

In the analysis among subjects with index date after 2003 (Table 3), we found that CNS infections 5–9 years before the index date was associated with higher PD occurrence (OR = 2.0, 95% CI: 1.2–3.4), but no risk elevation was found for infections ≥10 years before the index date. Furthermore, PD patients were three times (OR = 3.1, 95% CI: 1.0–9.6) more likely than controls to have multiple hospitalizations for CNS infections ≥5 years before the index date. In the analysis for sepsis, the corresponding ORs were 1.1 (95% CI: 0.9–1.4) for sepsis 5–9 years before the index date and 1.8 (95% CI: 1.0–3.2) for multiple septic events ≥5 years before the index date.

Table 3.

ORs and 95% CIs of CNS infections or sepsis in relation to PD among subjects with index date after 2003^a

	CNS infections		Sepsis
Variables	Cases/controls	OR (95% CI)	Cases/controls	OR (95% CI)
No	11 375/56 979	1.0 (ref.)	10 838/55 392	1.0 (ref.)
Yes	59/191	1.5 (1.2–2.1)	596/1778	1.7 (1.6–1.9)
Years before the index date
<1	8/17	2.4 (1.0–5.4)	254/332	4.0 (3.4–4.7)
1–4	9/30	1.5 (0.7–3.2)	205/806	1.3 (1.1–1.5)
5–9	21/52	2.0 (1.2–3.4)	93/436	1.1 (0.9–1.4)
≥10	21/92	1.1 (0.7–1.8)	44/204	1.1 (0.8–1.5)
Number of infections at least 5 years before the index date
Once	37/136	1.4 (0.9–2.0)	122/597	1.0 (0.9–1.3)
≥2 times	5/8	3.1 (1.0–9.6)	15/43	1.8 (1.0–3.2)
Duration of hospital stay at least 5 years before the index date (weeks)
<1	19/58	1.6 (1.0–2.8)	60/285	1.1 (0.8–1.4)
≥1	23/86	1.3 (0.8–2.1)	77/355	1.1 (0.9–1.4)

^aAdjusted for year of birth and gender.

After the index date, 32 (0.2%) PD cases and 97 (0.1%) controls were hospitalized at least once for CNS infections with a hazard ratio (HR) of 1.8 (95% CI: 1.2–2.7); the corresponding numbers for sepsis were 1057 (5.7%), 2942 (3.2%) and 2.2 (95% CI: 2.1–2.4), respectively.

Discussion

To the best of our knowledge, this is the first epidemiological study on CNS infections or sepsis in relation to PD risk. Using nationwide registers in Sweden, we did not find evidence that sepsis increased future risk of PD. Higher ORs were found only for cases with sepsis occurring around and after PD identification and for cases with PD as a secondary diagnosis; therefore, the association was more consistent with the hypothesis that PD patients are prone to sepsis rather than with a true casual relationship between sepsis and PD. The results for CNS infections and PD are intriguing: although the temporal relationship suggests that PD patients are more likely to have CNS infections than controls, a potential aetiological association between CNS infections and PD cannot be excluded.

More than two decades ago, McGeer et al.¹¹ first reported active microgliosis in post-mortem PD brains, and this observation was subsequently confirmed by other pathological studies and also in animal models of parkinsonism. These findings have established neuroinflammation as an important feature of PD pathology,⁴ although its role in PD pathogenesis remains to be determined. On the other hand, animal experiments showed that LPS treatment activated microglia and elevated proinflammatory cytokines, inducing dopaminergic neuron death in rats.^5,12 Even a single prenatal injection of LPS induced persistent dopaminergic neuron degeneration lasting into late adulthood.¹² In humans, there are rare case reports of parkinsonism after sepsis¹³ or accidental LPS intoxication.¹⁴ In support of the hypothesis that neuroinflammation plays an important role in PD aetiology, we recently reported that higher plasma level of IL-6, measured years before PD diagnosis, was associated with a higher risk of PD.¹⁵ Furthermore, regular users of ibuprofen had lower PD risk than non-users.^16,17

Although it is not yet known whether infections can lead to PD in later life, there have been historical and case reports of CNS infections and subsequent parkinsonian signs following influenza outbreaks or viral infections.^18,19 More recently, an experimental study showed that the highly pathogenic H5N1 virus could enter the CNS and induce long-term microglia-mediated neurodegeneration.²⁰ Several epidemiological studies showed that teachers and health-care providers were at higher risk of PD.^21–24 Together these findings provide convergent evidence that infection may contribute to PD aetiology and progression. On the other hand, direct epidemiological analyses of infectious diseases and PD are sparse and inconsistent. An early study in China found similar antibody titres to several viruses in PD cases and controls.⁶ Two other studies explored the link between childhood infections and later life PD risk. One reported lower PD risk among individuals with certain childhood viral infections,⁷ whereas the other linked higher PD occurrence to croup and rheumatic fever, but not to other early-life infectious diseases.⁸ A recent study in India found 9 Japanese B encephalitis patients among 345 PD patients but only 1 among 370 controls.²⁵ On the other hand, analysis of the UK-based General Practice Research Database found no association of seasonal influenza with future higher PD risk.²⁶ These previous studies therefore provide conflicting evidence regarding the role of infectious diseases in PD.

The focus of our analysis on severe systemic and CNS infections is well justified. The brain is protected by the blood–brain barrier, so one would hypothesize that CNS infections would be more likely to have long-term neurological consequences than peripheral infections. Sepsis, however, is a systemic inflammatory response to infections. A recent study found that severe sepsis was associated with long-term cognitive impairment,²⁷ providing novel evidence for a link between acute systemic inflammation and chronic neurodegeneration in humans. Furthermore, from a medical practice perspective, it is important to understand the possible long-term consequences of severe infections.

The present study showed that previous hospitalization for sepsis was unlikely to be associated with higher future risk of PD. We observed higher occurrence of PD only within 1 year after such hospitalization, and the association was especially pronounced for cases with PD as a secondary diagnosis. Neither multiple infections nor longer hospital stay ≥5 years prior to the index date was associated with higher PD occurrence. These results suggest that PD patients are more prone than controls to sepsis rather than that sepsis increases PD risk. However, we could not exclude the possibility that timely antibiotic treatment for sepsis may help prevent future higher risk for PD.

Interpretation of our results on CNS infections and future PD risk is less straightforward and we should consider several possible explanations. First, as with sepsis, the association of PD with CNS infections could be due to the possibility that PD patients are more likely to contract CNS infections than controls. In support of this, we found stronger associations for infections within 5 years before the index date, and higher ORs with PD ascertained as a secondary diagnosis. Secondly, CNS infections may indeed increase future risk of PD or perpetuate the process of neurodegeneration. Multiple CNS infections ≥5 years prior to the index date were more strongly associated with a higher PD risk compared with single infection. Furthermore, in the sensitivity analyses focused on cases more likely to be incident (i.e. cases identified after 2003), CNS infections 5–9 years prior to the index date were associated with higher PD occurrence. Thirdly, one could also speculate that the association was actually due to parkinsonism secondary to infection, misdiagnosed as PD. Another possibility is ascertainment bias: individuals with a previous history of CNS infections may be more likely to be diagnosed with PD than individuals without it because of increased medical vigilance. It is possible that more than one of these explanations contributed to the observed associations. Due to the lack of accurate clinical data on PD and the small number of participants exposed to CNS infections, we were unable to disentangle these various possibilities. Furthermore, the small number also made it impossible to identify specific types of CNS infections (e.g. viral vs bacterial) that were responsible for the observed association. Therefore, the interpretation of results on CNS infections and PD needs caution, and this finding should be further evaluated in future clinical studies.

Previous studies have shown that pneumonia and urinary tract infections are among the common reasons for hospitalization of PD patients.²⁸ Although the primary aim of our study was to evaluate the role of severe infections in PD aetiology, we found that PD patients were more likely to be hospitalized for CNS infections or sepsis than controls.

Register-based research on PD has been previously successfully carried out by others.^29,30 Use of historical hospitalization records excludes recall bias, which may be a concern in questionnaire-based case–control studies. The present study has several limitations. As the first analysis of CNS infections and sepsis in relation to PD, our study is exploratory in nature. Furthermore, because both the exposures and outcome were identified from hospitalization and outpatient registers, misdiagnosis, underdiagnosis and coding errors were possible. For the same reason, we did not have information on the date of disease onset or diagnosis and the analysis may therefore have included some prevalent cases. Specifically, exposures were defined using the Inpatient Register and some less severe patients might not have been identified. Previous validation showed patients with CNS infections were likely to be accurately identified, but many sepsis patients might have been missed due to the low sensitivity (∼43%) of using register data to identify sepsis cases.¹⁰ Randomly missing information on sepsis might have attenuated the association towards null; otherwise, the association might have been biased in a less predictable way. Finally, subjects of our study were fairly old, and the exposure assessment only went back to 1987 when the Inpatient Register became nationwide. Our study therefore was unable to investigate whether early-life infections increase PD risk in later life.

In summary, our study provides preliminary evidence for an association between CNS infections, but not sepsis, and a higher future risk of PD. The possibility that CNS infections increase PD risk should be further investigated.

Funding

The Swedish Research Council (SIMSAM Grant No. 80748301); intramural research program of the National Institutes of Health, the National Institute of Environmental Health Sciences (Z01-ES-101986 and unmet need grant); the Swedish Society for Medical Research; the Swedish Medical Society and the Parkinson Foundation in Sweden; Hjärnfonden (postdoctoral fellowship to F.F.).