Increased risk of laryngeal and pharyngeal cancer after gastrectomy for ulcer disease in a population-based cohort study

Background: Gastrectomy has been indicated as a risk factor for laryngeal cancer, and possibly also for pharyngeal cancer, but few studies are available. The postulated mechanism is increased bile reflux following gastrectomy. Methods: This was a population-based cohort study of patients who underwent gastrectomy for peptic ulcer disease between 1964 and 2008 in Sweden. Follow-up data for cancer was obtained from the Swedish Cancer Register. Relative risk was calculated as standardised incidence ratios (SIRs) with 95% confidence intervals (CIs). Results: The gastrectomy cohort comprises 19 767 patients, contributing 348 231 person-years at risk. The observed number of patients with laryngeal (n=56) and pharyngeal cancer (n=28) was two-fold higher than the expected (SIR: 2.0, 95% CI: 1.5–2.6 and SIR: 2.4, 95% CI: 1.6–3.5, respectively). After exclusion of 5536 cohort members with tobacco- or alcohol-related disease, the point SIRs remained increased (SIR: 1.6, 95% CI: 1.1–2.2 and SIR: 1.7, 95% CI: 0.9–2.8, respectively). The SIRs of laryngeal and pharyngeal cancer increased with time after gastrectomy (P for trend <0.0001), and were particularly increased ⩾30 years after gastrectomy (SIR: 4.8, 95% CI: 2.1–9.5 and SIR: 10.2, 95% CI: 3.7–22.3, respectively). Conclusion: Gastrectomy for peptic ulcer disease might entail a long-term increased risk of laryngeal and pharyngeal cancer.

The only established risk factors for laryngeal and pharyngeal squamous cell carcinoma are high consumption of tobacco and alcohol (Zeka et al, 2003). An association between oesophageal biliary regurgitation and these tumours has also been indicated, particularly laryngeal cancer, but this relation is debated (Galli et al, 2006a, b). The lack of protective mechanisms against proximal reflux has been proposed to contribute to the harmful effect, but any influence of bile regurgitation on the development of laryngeal and pharyngeal cancer is difficult to assess in humans. Problems arise from difficulties with assessing exposure to bile reflux, and the long expected latency period before invasive tumours will develop. The specific situation occurring after partial or total gastrectomy, however, resembles a human experimental model of bile reflux, since the anatomical rearrangement after such surgery facilitates the flow of duodenal contents to reach the oesophagus. A large proportion of patients who have undergone gastrectomy suffer from oesophageal bile regurgitation (Stoker and Williams, 1991;Matei et al, 2010). Thus, if bile regurgitation is a carcinogenic exposure for the larynx or pharynx, gastrectomy could be followed by an increased risk of tumours at these sites. This hypothesis has gained support from a limited number of investigations, including a large case -control study showing a four-fold increased risk of laryngeal cancer among patients who reported having undergone a gastrectomy (Cammarota et al, 2004). However, no cohort studies are available. To further evaluate the relation between bile regurgitation and risk of laryngeal and pharyngeal cancer, we conducted a large cohort study assessing the association between gastrectomy for peptic ulcer disease and risk of developing these tumours.

Design
This was a Swedish population-based cohort study addressing the risk of developing laryngeal or pharyngeal cancer after gastrectomy for peptic ulcer disease, using the entire Swedish population during the period of 1964 -2008 as a database. The nationwide Swedish Patient Register was used to identify a cohort of all patients who had undergone gastrectomy for peptic ulcer disease during the study period in Sweden. New tumours in the gastrectomy cohort were identified through linkage to the nationwide Swedish Cancer Register. The incidence of laryngeal or pharyngeal cancer among the gastrectomy patients was compared with the incidence in the Swedish background population of the corresponding age, sex, and calendar year. The cancers occurring in the comparison population were assessed through the data on the cancer incidence and prevalence in the Swedish Cancer Register. Both in the study cohort and in the comparison population, only first tumours were included. In addition, the first year after gastrectomy was excluded to reduce the risk of earlier detection of any cancer because of the gastrectomy, and to allow a minimum time of exposure. To obtain the correct censoring of person-time, the dates of all deaths were collected through linkage to the Swedish Cause of Death Register. The personal identity number, a 10-digit number assigned to every resident in Sweden since 1947, made the linkage of individuals across registers possible (Ludvigsson et al, 2009). The study was approved by the Regional Ethics Committee in Stockholm.

Detection of laryngeal and pharyngeal cancer
All cancers diagnosed during follow-up of the cohort (1965 -2008) were identified through linkage to the Swedish Cancer Register, a nationwide register initiated in 1958 and administered by the National Board of Health and Welfare, which is at least 98% complete (Barlow et al, 2009). The register contains information about the location, histological type, and date of diagnosis of all malignant tumours in Sweden. The codes for all tumour diagnoses are converted into the ICD codes version 7, and the study outcomes were defined by the codes 161 for laryngeal cancer and 146 -148 for pharyngeal cancer, combined with the histological code 146 for squamous cell carcinoma.

Statistical analyses
Person-time at risk was accumulated from 1 year after the surgery until the first occurrence of any cancer, death, or the end of observation (31 December 2008), whichever came first. The relative risk was estimated as the standardised incidence ratio (SIR), that is, the observed number of laryngeal or pharyngeal tumours in the gastrectomy cohort divided by the number of these tumours in the comparison population (expected number). The expected number of cancers was calculated by multiplying the observed person-time by cancer incidence rates specific for age, sex, and calendar year. The expected rates were derived from the Swedish Cancer Register data through the Swedish population, and aggregated into 5-year intervals. The SIRs were inherently adjusted for the potential confounding factors age, sex, and calendar year, as the incidence in the observed cohort was compared with the corresponding incidence in the age-, sex-, and calendar yearmatched general population. Confidence intervals (CIs) of SIRs were calculated on the assumption that the observed number of events followed a Poisson distribution (Breslow and Day, 1987).

Patients
The study cohort included 19 767 patients who had undergone gastrectomy. Some characteristics of these patients are presented in Table 1. The follow-up time was a median of 17 years, providing 348 231 person-years at risk. There was a male predominance (62.3%). Most operations were partial gastrectomies (96.4%), while the remaining proportion represented total gastrectomies.

Risk of laryngeal cancer
A total of 56 patients developed laryngeal cancer at least 1 year after the gastrectomy, rendering a two-fold higher risk than expected (SIR: 2.0, 95% CI: 1.5 -2.6) ( Table 2). The risk was further increased in periods representing longer latency intervals after gastrectomy, that is, longer time of exposure (P for trend o0.0001). Among those who were operated on at least 20 years earlier, the SIR was 2.7 (95% CI: 1.6 -4.2), and among those who were operated on at least 30 years earlier, the SIR was nearly a fivefold increase (SIR: 4.8, 95% CI: 2.1 -9.5). There were no substantial differences between sexes, age groups, or calendar periods regarding risk of laryngeal cancer after the gastrectomy (Table 2). Analyses restricted to patients with a histologically confirmed squamous cell carcinoma of the larynx revealed similar results (data not shown). After exclusion of 5536 cohort members with any disease linked with tobacco smoking or alcohol abuse, the overall SIR remained increased (SIR: 1.6, 95% CI: 1.1 -2.2). The SIR was 2.2 (95% CI: 1.1 -4.0) at least 20 years after the gastrectomy, and the SIR was 4.0 (95% CI: 1.3 -9.3) among patients who were followed up for at least 30 years (Table 3).

Risk of pharyngeal cancer
A total of 28 patients developed pharyngeal cancer during the follow-up of the gastrectomy cohort. The overall risk of pharyngeal cancer was over two-fold higher than expected (SIR: 2.4, 95% CI: 1.6 -3.5). The SIR was particularly increased in periods representing longer exposure time after gastrectomy. Among patients who were operated on at least 30 years earlier, there was a 10-fold increased risk (SIR: 10.2, 95% CI: 3.7 -22.3). There were no notable differences between sexes or age groups after the gastrectomy, while the SIR was higher in the later calendar period ( Table 2). The results based only on patients with a histologically confirmed squamous cell carcinoma of the pharynx were similar (data not shown). After exclusion of cohort members with any disease linked with tobacco smoking or alcohol abuse, the SIRs decreased (Table 3). The overall SIR was 1.7 (95% CI: 0.9 -2.8), and after at least 30 years of follow-up the SIR was 2.3 (95% CI: 0.5 -16.3).

DISCUSSION
This study supports the hypothesis that gastrectomy increases the risk of laryngeal and pharyngeal cancer.  Methodological advantages of the study include the populationbased cohort design, the complete assessment of the exposure and the outcomes, the long and complete follow-up, and the large sample size. The use of a gastrectomy cohort mimics an experimental model that allows the assessment of the effects of the exposure to bile reflux in humans. Adjustments for age, sex, and calendar year were achieved through the design, but a drawback was the lack of direct data on tobacco smoking and alcohol abuse, the established risk factors for these tumours (Zeka et al, 2003). Nevertheless, it was possible to address confounding by these factors by excluding all patients who had ever been recorded with a diagnosis linked with tobacco smoking or high alcohol consumption in the Patient Register.
The results of the present study are in line with those reported in a limited number of studies that have addressed the relation between gastrectomy and risk of laryngeal and pharyngeal cancer. The hypothesis of an association between gastrectomy and laryngeal or pharyngeal cancer has been put forward in a small case series that demonstrated abnormal bile reflux in three patients with laryngeal cancer after gastric resection (Cianci et al, 2000). This finding was followed up by a case -control study of 40 gastrectomy patients with biliary reflux due to gastrectomy and 40 non-gastrectomy dyspeptic patients, which provided some further support for the hypothesis that gastrectomy is a risk factor for precancerous and squamous cell carcinoma of the larynx or pharynx (Galli et al, 2002). Another case -control study compared the presence of laryngeal lesions in 93 patients who had undergone gastric resection with 93 matched dyspeptic controls, and found that seven patients (7.5%) had malignant or premalignant laryngeal lesions in the gastrectomy group, while one control patient had a lesion. Finally, a large case -control study found that previous gastrectomy was reported by 8.1% of the 828 cases of laryngeal cancer and by 1.8% of the 825 matched controls with myocardial infarction, and the relative risk was 3.8 (95% CI: 2.1 -7.0) after adjustment for tobacco smoking and alcohol consumption (Cammarota et al, 2004). The association was strongly dependent on latency time; the odds ratio was 14.8 (95% CI: 3.4 -64.6) X20 years after the gastrectomy. The finding of a stronger association with longer follow-up is in agreement with the findings of the present study. Taken together, although based on a scarce number of studies, there are data to support an association between gastrectomy and risk of laryngeal and pharyngeal cancer.
The biological mechanism behind a possible association between gastrectomy and these tumours is uncertain. There have been a number of studies performed and some have provided evidence that acid reflux may be a risk factor for laryngeal cancer (Qadeer et al, 2005). The situation that occurs after partial or total gastrectomy mimics a human experimental model of bile reflux, since such surgery is often followed by a substantially increased risk of oesophageal exposure to bile (Toye and Williams, 1965;Yumiba et al, 2002;Matei et al, 2010). The anatomical rearrangement after gastrectomy means that duodenal contents easily flows back and can reach the oesophagus, the pharynx, and the larynx. The pharynx and larynx have no defence mechanisms against such proximal reflux, for example, no peristalsis, and such reflux might cause chronic injury and inflammation on the epithelium (Glanz and Kleinsasser, 1976). Tobacco smoking and alcohol consumption, the main risk factors for laryngeal and pharyngeal cancer, might increase the vulnerability of the mucosa for such reflux exposure.
In conclusion, this large and population-based cohort study with long and complete follow-up provides further support for the hypothesis that gastrectomy increases the risk of squamous cell carcinoma of the larynx and pharynx.