The inflammatory reflex and risk for rheumatoid arthritis: a case–control study of human vagotomy

Abstract

Recent data suggest remarkable effects of vagus stimulation (reduction) and vagotomy (exacerbation) on acute inflammation in rats, the so‐called “inflammatory reflex”. Its role in humans remains unknown. Therefore, the aim was to explore whether surgical vagotomy in humans would affect the risk of a prototype inflammatory disease, rheumatoid arthritis. This was a case–control study. Assessment of the relative risk (RR) of developing rheumatoid arthritis after surgical vagotomy during 1964–2001 in 63 092 prevalent rheumatoid arthritis cases versus 125 404 matched controls from the general population and in 2548 incident rheumatoid arthritis cases versus 24 357 matched controls from the general population, respectively, was done. For comparison, we assessed RRs for hospitalisation for gastric disorders not including vagotomy. Data on exposures and rheumatoid arthritis were retrieved from population‐based and prospectively recorded Swedish registers. A pre‐rheumatoid arthritis vagotomy was not significantly associated with an increased risk for rheumatoid arthritis (RR = 1.17, 95% CI 0.97 to 1.40). RRs in the same range were observed for several other pre‐rheumatoid arthritis gastric conditions that do not include vagotomy (eg, gastric ulcer RR = 1.21, 95% 1.11 to 1.33). Vagotomy has no specific effect on the risk of developing rheumatoid arthritis in humans. Gastroduodenal ulcers occur more often than expected even before the occurrence of rheumatoid arthritis.

Blockade of tumour necrosis factor (TNF) has proven to be a highly successful treatment for rheumatoid arthritis. Accordingly, factors that govern the natural regulation of TNF in humans may be of pivotal importance for the occurrence of chronic inflammation; interference with vagus function may be one such factor. In a series of recent experimental animal studies, Tracey and colleagues have shown the existence, function and importance of a cholinergic anti‐inflammatory pathway, also known as “the inflammatory reflex”, which is a fast and efficient vagus‐mediated mechanism for neural inhibition of systemic inflammation.¹,²,³ In experimentally induced endotoxaemia, electrical stimulation of the vagus nerve considerably attenuates serum levels of TNF and prevents the development of shock.¹,²,³ In addition, vagus nerve stimulation has been shown to reduce localised peripheral inflammation² and also joint inflammation in arthritis models (J Lampa, unpublished results). Conversely, vagotomy in rats has been associated with exacerbated TNF response and shortened survival during endotoxaemia.¹,²,³

Although appropriate function of this inflammatory vagus reflex is critical for acute inflammatory responses, little is known about its role for chronic inflammatory responses, let alone its role in humans. Before the introduction of H2 blockers and proton pump inhibitors, vagotomy—isolated or as part of ventricular resection—was a common surgical procedure to treat gastric ulcer diseases. Such vagotomised individuals therefore offer the possibility of exploring whether the remarkable effects of interference with vagus nerve function on inflammation in rats also apply to chronic inflammation in humans. We therefore performed a case–control study to assess the risk of rheumatoid arthritis after vagotomy, taking advantage of Swedish longitudinal, population‐based and prospectively recorded health register data.

Subjects and methods

Data sources

1. The Swedish Inpatient Register is a population‐based register that contains information on individual discharges from Swedish inpatient care (which is public and population based) since 1964. For each hospitalisation, the National Registration Number (a unique number assigned to all Swedish residents alive in 1947 or born thereafter), medical diagnoses, codes for surgical procedures, dates of admission and discharge, department and hospitals are recorded. Validations against the American College of Rheumatology criteria suggest that 90% of the registered rheumatoid arthritis diagnoses are correct.⁴ Estimations based on the expected prevalence of rheumatoid arthritis suggest that over half of all prevalent Swedish patients with rheumatoid arthritis today are covered by this register.

2. The Early Arthritis Register is a semi‐national register established in 1997 and includes incident (<1 year of symptoms) cases of rheumatoid arthritis fulfilling the American College of Rheumatology criteria.⁵

3. The Register of Population and Population Changes is the Swedish census register since 1969.

4. The Cause of Death Register contains information on dates and causes of death for all Swedish residents deceased since 1952.

Cases, controls and exposures

We identified, as a first set of cases, all individuals >16 years of age who were discharged from inpatient care with rheumatoid arthritis at least once during 1964–2001 and who had not at any point in time also been discharged with psoriatic arthritis, ankylosing spondylitis or systemic lupus erythematosus (n = 63 092). As a second set of cases, we identified all 2458 individuals registered in the Early Arthritis Register as of 2003. As controls, we randomly selected individuals from the Register of Total Population individually matched (1:2 for cases from the Inpatient Register and 1:10 for cases from the Early Arthritis register) to their cases by year of birth, sex, civil status and county of residence. Table 1 displays the characteristics of cases and controls.

Table 1 Characteristics of the two sets of Swedish cases with rheumatoid arthritis and their matched general population controls.

	Inpatient Register		Early Arthritis Register
	Cases n = 63 092	Controls* n = 125 404	Cases n = 2458	Controls* n = 24 357
Sex
Males	18 370	36 619	733	7210
Females	44 722	88 785	1725	17 147
Age at entry (years)
0–49	11 619	23 205	709	6909
50–74	38 115	75 721	1419	13 831
⩾75	13 358	26 478	330	3617
Period of entry
1964–1979	24 500	48 845	–	–
1980–1989	21 517	42 694	–	–
1990–2001†	17 075	33 865	2458	24 357

*Controls sampled from the general population.

†1995–2001 for cases–controls identified in the Early Arthritis Register.

For cases and controls, we identified all surgical procedures in the Inpatient Register from 1964 until the date of diagnosis (early arthritis cases)/first discharge with rheumatoid arthritis (inpatient register cases) listing isolated vagotomy, or abdominal ventricular resection (a truncal vagotomy was often performed in conjunction with ventricular resection). Isolated vagotomies were originally truncal (performed where vagus enters through the diaphragm), but selective (only denervating the stomach) vagotomies became increasingly popular. To assess the specificity of any observed association between vagotomy and rheumatoid arthritis, and to detect associations due to reversed causality (eg, non‐steroidal anti‐inflammatory drug‐related ulcers in patients with incipient or not yet hospitalised rheumatoid arthritis) or confounding (eg, smoke‐related gastric ulcers⁶), we identified related gastric surgical procedures or medical diagnoses that do not include vagotomy (gastrorraphy, any gastric or duodenal ulcer). To further reduce the potential for confounding by indication in the analyses based on cases identified in the Inpatient Register and their controls, we only included surgical procedures occurring before the first hospitalisation listing any musculoskeletal diagnosis (eg, unspecified mono‐arthritis and poly‐arthritis, arthralgias) rather than first hospitalisation listing rheumatoid arthritis. Similarly, for cases identified in the Early Arthritis Register and their controls, we only included exposures occurring at least one year before the diagnosis of rheumatoid arthritis. To detect operability‐related biases, we also assessed the relative risk (RR) associated with a pre‐rheumatoid arthritis cholecystectomy, for which there is no known association with rheumatoid arthritis.⁷ RRs were estimated as odds ratios (ORs) including 95% CI using conditional logistic regression in SAS.

Results

In all, 179 (0.28%) of the 63 092 cases with rheumatoid arthritis identified in the Inpatient Register and 304 (0.24%) of their matched 125 404 controls had a history of vagotomy or abdominal ventricular resection (OR = 1.17, 95% CI 0.97 to 1.40; table 2).

Table 2 OR with 95% CI associated with a history of vagotomy and other gastrointestinal medical conditions or surgical procedures among 63 029 Swedish cases with rheumatoid arthritis identified in the Inpatient Register 1964–2001 and their population controls and among 2458 Swedish cases with rheumatoid arthritis identified in the Early Arthritis Register 1995–2001 and their population controls.

	Inpatient Register (cases = 63 029/controls = 125 404)			Early Arthritis Register (cases = 2458/controls = 24 357)
	Cases	Controls	OR (95% CI)	Cases	Controls	OR (95% CI)
Vagotomy or ventricular resection	179	304	1.17 (0.97 to 1.40)	10	64	1.56 (0.65 to 3.76)
Gastrorraphy	61	66	1.83 (1.29 to 2.60)	3	14	2.14 (0.48 to 9.55)
Ulcer disease	753	1233	1.21 (1.11 to 1.33)	40	312	1.29 (0.92 to 1.80)
Cholecystectomy	1458	2868	1.01 (0.95 to 1.08)	83	859	0.96 (0.77 to 1.21)

There were similar ORs for isolated vagotomy and for abdominal ventricular resection. The RR for vagotomy/ventricular resection more than a decade before rheumatoid arthritis was 1.0, and there was no clear pattern in RR for vagotomy when stratified by decade of surgery or by type (data not shown). Overall, RRs in the same range were observed for gastrorraphy (OR = 1.83, 95% CI 1.29 to 2.60) and for gastroduodenal ulcer disease (OR = 1.21, 95% CI 1.11 to 1.33) with little difference between gastric, duodenal or unspecified location. In contrast, a history of cholecystectomy was not associated with rheumatoid arthritis (OR = 1.01, 95% CI 0.95 to 1.08).

Ten (0.41%) of the 2458 cases with incident rheumatoid arthritis identified in the Early Arthritis Register and 64 (0.26%) of their 24 357 matched controls had a history of vagotomy or abdominal ventricular resection (OR = 1.56, 95% CI 0.65 to 3.76; table 2), but gastrorraphy (OR = 2.14, 95% CI 0.48 to 9.55) and gastroduodenal ulcer disease (OR = 1.29, 95% CI, 0.92 to 1.80) were associated with increased risks of the same order of magnitude. In contrast, a history of cholecystectomy was not associated with rheumatoid arthritis (OR = 0.96, 95% CI 0.77 to 1.21).

Discussion

In this study, a history of vagotomy was associated with a modestly but non‐significantly increased (+17%) risk of future rheumatoid arthritis. This association was not specific for vagotomy, but observed also for gastric ulcer disease and surgical treatment of gastric ulcers not including vagotomy.

The inflammatory vagus‐mediated reflex acts through afferent vagal signals triggered by inflammatory products (eg, interleukin 1 release). Such stimulation results in release of acetylcholine from efferent vagus motor branches, which stimulates nicotinic receptors expressed on macrophages, which in turn respond by suppressing the release of TNF.⁸ In addition, sensory (afferent) vagus stimuli also increase central nervous system production of adrenocorticotropic hormone that activates the humoral anti‐inflammatory response. Although quite impressive results have come out of experimental studies of vagotomy and vagus stimulation in animal models of acute inflammation, the effects on chronic inflammation remain unstudied. Because of the obvious difficulties in designing experimental studies including human vagotomy, or large‐scale vagus stimulation, little is also known about its role in the clinical setting, in humans. In our study the vagotomy was truncal or selective—performed where the vagus nerve enters through the diaphragm or further distal—isolated or as part of a surgical resection of the abdominal ventricle including its vagal branches. Since the relative contributions of the vagus nerve branches from the brainstem—and distal thereof—on the inflammatory reflex are unknown, we cannot exclude that a more proximal vagotomy than that covered by our investigation might have yielded a different result. Suppression of inflammation through the vagus nerve is fast acting. Therefore, it may be that these immune‐regulatory mechanisms are effective on a different timescale (minutes) than that taken to induce rheumatoid arthritis (years). Similarly, although our results indicate a limited role, if any, of vagotomy/the inflammatory reflex in the aetiology of human rheumatoid arthritis, they do not preclude that the vagus may exert other effects on rheumatoid arthritis.

We observed equally strong or stronger associations with rheumatoid arthritis for gastric surgery or gastric diseases that do not include vagotomy than for vagotomy per se. Smoking is a risk factor for peptic ulcer⁶ and seropositive rheumatoid arthritis.⁹,¹⁰ The slightly increased RRs for all gastric conditions assessed in our study might thus be due to confounding by smoking. The effects of smoking on inflammation are, however, complex, best exemplified by the increased risk of ulcerative colitis in ex‐smokers, and the (although limited) evidence for a beneficial effect of nicotine patches in the treatment of the same disease.

The variation in RRs between our study populations was within sampling error. The matched design and analyses ensured adjustment for age, sex, civil status and county of residence, but we lacked information on other potential common risk factors for rheumatoid arthritis and the exposures under study, which is a limitation. Although we took precautions to avoid reversed causality, we cannot exclude that such bias may partly explain the observed 17% increase in risk.

Our results indicate that the impressive effects of vagotomy on acute inflammation in rats are not mirrored in chronic arthritis in humans. Inadvertently, our results also highlight that gastric ulcer disease occurs more often than expected before the diagnosis of rheumatoid arthritis. Hence, ulcers in rheumatoid arthritis should not uncritically be attributed to non‐selective non‐steroidal anti‐inflammatory drugs and other rheumatoid arthritis treatments.

Abbreviations

TNF - tumour necrosis factor