Hepatitis E is a viral infection with a mostly benign course and a low case fatality rate in the general population (about 1%) with the exception of pregnant women.1 Helicobacter pylori is the main cause of peptic ulcers and stomach cancer.2 3 Both pathogens have been found in sewage.4^–11 Consequently, compensation and the prevention of hepatitis E,12 peptic ulcers13 and possibly stomach cancer14 in workers exposed to sewage must be considered.

In a previous cross-sectional study,15 the prevalence of peptic ulcers and hepatitis E and the seroprevalence of H pylori and hepatitis E were not found to be increased in workers exposed to waste water compared with non-exposed subjects. However, a cross-sectional design may provide incorrect results due to the healthy worker effect. Furthermore, incidence rates are more reliable than prevalence rates for examining the temporal relationship between exposure and disease, and a prospective design allows each subject to be used as his/her own control. Therefore, a follow-up study of a previous cross-sectional study15 in sewage workers was conducted to confirm the results of the original investigation. The endpoints were the incidence of clinical hepatitis E and peptic ulcers, and seroconversion rates.

METHODS

A detailed description of the methods has been given previously together with the results of the baseline examination,15 so only the main aspects will be briefly summarised here.

Subjects

All workers exposed to sewage in the Canton of Zurich (Switzerland) were invited to participate. Potential control subjects (garbage collectors, gardeners, workers maintaining waterways, public transport workers and forestry workers) were approached individually and invited to participate until sufficient control subjects had been recruited. The workers were included in the study only after they had given written consent. Overall, 778 subjects were included in the baseline examination (61% participation rate): 355 workers were from waste water plants and 423 from control plants (90% and 49% participation rates, respectively). Exposure to sewage during the entire working life was assessed for each job separately and whenever exposure defined by the plant differed from individual occupational history, misclassifications were corrected to ensure that no subject from the control group had any exposure to sewage and vice versa. Ultimately, 332 and 446 workers were included for follow-up as currently exposed and non-exposed to waste water, respectively.

Participation was associated with Swiss nationality (66.2% vs 47.5%) and slightly younger age (median: 44 vs 47 years) but not with gender (p>0.4; χ2 test).15 The lower participation rate in controls was associated mainly with two plants with participation rates of 15% and 21%. When these were excluded, the rate rose to 79%. Thus, some statistical analyses were re-run without data from these plants. Some immunological results are missing because blood sampling was not acceptable or possible or for organisational reasons.

Methods

This prospective cohort study conducted in the Canton of Zurich consisted of a baseline and four annual follow-up examinations. The baseline medical examination took place between June 2000 and July 2002 and the fifth examination between August 2004 and May 2006. As a general rule, examinations were planned at 1-year intervals but could be advanced or postponed because of holidays, illness or increased workload. The minimum interval between two consecutive examinations was 6 months. As the plants did not regularly inform us of workers leaving employment, we were unable to organise final examinations of such workers in time. Therefore, as soon as we knew that a worker had left, he/she was contacted and asked to complete a short checklist on the reasons for leaving and any medical reasons in particular. After three unsuccessful attempts by phone and/or mail, workers were declared missing.

Each physician examined both exposed and unexposed subjects and the coding of the answers was reviewed throughout the study by the same occupational health practitioner. Differences were resolved by checking information again and discussing codes with the examining physicians. At each follow-up examination, workers were asked whether their physician had diagnosed a liver or stomach disease since the last examination. At the final examination workers were asked if they had received a blood transfusion (to account for any HEV transmission).16 The presence of peptic ulcers and liver diseases was defined on the basis of clinical history and gastritis on the basis of biopsy data reported by the patient. Socioeconomic level, country of childhood, alcohol consumption17 and γ-glutamyltransferase (GGT) activity were assessed at baseline, smoking was assessed at baseline and at the last examination, and travel to endemic areas was assessed yearly as described previously.15 18 19 The use of personal protective equipment was assessed at baseline. Suitable gloves and adequate masks were defined as personal protective equipment worn for at least 50% of working time and adequate for work with waste water (ie, leather gloves permeable to water were not considered to be suitable). Furthermore, masks had to be maintained correctly. Exposure was assessed individually at all five examinations with the same questionnaire and defined by four indicators15 18: exposure to sewage during follow-up (yes/no), duration of exposure during follow-up (weeks), occurrence of splashes during follow-up (never exposed to sewage, never more than 20 splashes of raw sewage in any job, at least one job with more than 20 splashes), and exposure to raw sewage during follow-up (never exposed, exposure ⩽5 times monthly, exposure >5 times monthly in at least one job). Changes in occupational duties were taken into account in the exposure assessment.

All immunological determinations were carried out in the Clinical Immunology Unit, University Hospital, Zurich according to the manufacturer’s instructions and quality controlled. Analyses were done in batches and the laboratory was blinded to exposure status. The same methods were used throughout the study. The limit values recommended by the manufacturer and followed by the University Hospital were used. Immunoglobulin G antibodies to H pylori (H pylori IgG) and immunoglobulin A antibodies to H pylori (H pylori IgA) were determined with enzyme linked immunosorbent assays (ELISA) (Synelisa H pylori (IgG) Abs, Pharmacia and Upjohn, Germany; Quanta Lite H pylori IgA ELISA, Inova Diagnostics, California, USA). Test results below 10 and 20 U/ml for H pylori IgG and IgA, respectively, were considered negative. A questionable seroconversion was defined by antibody titres in a range including the cut-off plus 10% (the inter-assay coefficient of variation given by the manufacturer). Antibodies (IgG/IgM) to HEV were determined with an ELISA using rDNA as an antigen (ORF3 and part of ORF2) (Abbott Laboratories, Abbott Park, IL 60064, USA). Each positive result for antibodies to HEV was based on a duplicate determination. HEV borderline results were classified as positive for statistical analyses. As seropositive subjects were not followed up, seroreversion rates could not be calculated.

Data analyses

The normality of the distribution was tested and logarithmic transformations carried out if necessary or non-parametric tests used. Survival was defined as the time between the baseline examination and the occurrence of seroconversion, the end of exposure to waste water, or the end of follow-up, whichever came first. Three survival times (antibodies to HEV, H pylori IgG, H pylori IgA) were calculated for each worker and examined with log-rank tests and Cox regressions, which included the variables from the final model presented by Jeggli et al.15 Cox regression used the option “exact” for ties handling and age was included as a time-dependent covariate. Sensitivity analyses were used for informative censoring by considering either all censored cases as having seroconverted at the time of censoring or all censored cases as having remained seronegative until the longest follow-up duration (270 weeks). All calculations were performed with SAS v 8.2 statistical software (SAS Institute, Cary, North Carolina, USA).

RESULTS

The median intervals were 60 (n = 700), 46 (n = 676), 52 (n = 642) and 52 (n = 603) weeks between the first and second, second and third, third and fourth, and fourth and fifth examinations. The shortest and longest follow-ups were 8 and 270 weeks, respectively (median: 196 weeks).

Figure 1 summarises the reasons for non-participation and loss to follow-up. Table 1 summarises the characteristics of the subjects. At baseline, there were several statistically significant differences between groups including age, gender, educational level, nationality, country of childhood, smoking status (but not pack-years), duration of follow-up, and GGT activity (but not alcohol consumption) (0.001<p<0.02). All possible risk factors were considered in the Cox regression, although the differences were mostly too small to be viewed as clinically relevant for H pylori or hepatitis E. Importantly, participation until the end of the study was associated with “job change because of any health problem” (12 (21%) vs 44 (79%) in participants and non-participants, respectively; p<0.001). However, job changes during follow-up were not associated with seroconversion for either HEV or H pylori (p⩾0.3; Fischer’s exact test) and no job changes occurred because of hepatitis or liver disease. One sewage worker left follow-up because of early retirement due to a perforated peptic ulcer, but this was not an incident case (a first bleeding episode because of peptic disease had already occurred 10 years before the beginning of follow-up). Women left the study during follow-up slightly more often than men (p = 0.06). There were no further differences between those participating and not participating until the end of follow-up with respect to the other variables listed in table 1 (p>0.4) and the lifelong duration of exposure to sewage was similar in those exposed at baseline whether or not they participated until the end of follow-up (p = 0.1). Four workers changed location once during follow-up, while a single worker was in the same plant at the beginning and end of follow-up but worked between times for 2 years in another sewage plant.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Flow chart showing reasons for non-participation.

According to occupational history, the number of sewage-exposed jobs the workers had been exposed to at the beginning of follow-up ranged from one to five (the number of subjects and median duration in years were 332 and 8, 92 and 6, 25 and 5, 5 and 4, and 1 and 0.5 for the current and the first, second, third and fourth prior sewage-exposed jobs, respectively).

Sewage plant workers must receive training, most waste water workers undergo 5 years of training (“Klärwerkermeister”), and courses are held to remind them of the importance of hand washing and personal protective equipment. Each plant has a shower unit which is much used. Work clothes are always worn; details on the wearing of gloves and masks are given in table 1. Goggles, if any, are used primarily in plant laboratories.

A total of 667 workers were seronegative for hepatitis E at the beginning of follow-up and had at least one follow-up determination. During follow-up no clinical hepatitis E was diagnosed and seroconversion was found in 26 subjects, with identical incidence rates in both exposed and non-exposed workers (table 2). Twenty five subjects had definite positive titres; in only one case was seroconversion based on a borderline finding (as the subject left employment after seroconversion, no further determination was available). In 24 subjects seroconversion was not accompanied by clinical symptoms. In two further patients, liver disorders were noted but no diagnosis of hepatitis E was made (one subject had hepatitis B and one subject had pathological liver enzymes). Thirteen of the 26 workers with seroconversion had never been to an endemic area and no association was found between seroconversion and travel to endemic areas (p>0.6; χ2 test). Survival did not differ between sewage and control workers (fig 2) for any exposure indicator (p⩾0.4). Neither of the two subjects who had received a blood transfusion during follow-up seroconverted. Wearing personal protective equipment had no statistically significant effect (p>0.6).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Helicobacter pylori IgG survival distribution function. AHP, anti-Helicobacter pylori.

There were no differences with respect to clinical endpoints as regards H pylori (table 2). Concerning H pylori IgG and IgA, valid follow-up data were available for 395 and 534 subjects, respectively, who were seronegative at baseline and seroconversion occurred in 125 and 242 of these subjects, respectively. Incidence rates were very similar in control and exposed subjects (table 2). Survival curves and hazard ratios from Cox regression (figs 2 and 3) did not reveal any statistically significant effect of exposure to waste water (p>0.2 for all four exposure indicators and both antibody classes). After excluding subjects with questionable seroconversion, the number of seroconverters decreased markedly to 93 and 162 for anti-H pylori IgG and IgA, respectively, but this exclusion did not alter the results of survival analyses. In the Cox regression no independent variable was associated with seroconversion in a significant and consistent way for either IgG or IgA antibodies. Specifically, shift work and personal protective equipment had no statistically significant effect (p>0.2) and no dose–response relationship was found. Smoking habits at the end of follow-up (instead of at baseline) did not alter the results.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Helicobacter pylori IgA survival distribution function. AHP, anti-Helicobacter pylori.

None of the six sensitivity analyses examining informative censoring showed an increased seroconversion rate in sewage workers for any of the three serological outcomes. Excluding the workers from both plants with a very low participation rate did not alter the results.

Most recent European studies on H pylori seroconversion have assessed seroconversion rates in “%/year” in populations consisting only of participants with valid results at baseline and at a follow-up examination (no survival curves were calculated). Thus, a subgroup similarly selected was drawn from the whole study population to calculate seroconversion rates in “%/year”. This subgroup comprised all subjects who had participated in both the first and the last examinations with valid results both times, were seronegative at baseline, and had a follow-up period longer than 4 years. This subpopulation was followed for a median time of 4.5 years (range: 4.01–5.10 years; the first and last examinations were usually in 2000 and 2005, respectively). The seroconversion rates were 6.8%/year and 10.8%/year for H pylori IgG and IgA, respectively, and decreased to 5.8%/year and 8.2%/year after excluding subjects with questionable seroconversion. In all four comparisons exposure to waste water had no effect on survival (p⩾0.3).

DISCUSSION

This is the first prospective cohort study to examine the incidence of hepatitis E and peptic ulcers as well as the incidence of antibodies to HEV and H pylori in a large group of sewage workers. Our purpose was to examine the hypothesis that sewage workers are at increased risk of HEV and H pylori infection as both agents have been found in sewage.4^–11

These results from the Canton of Zurich, Switzerland, a non-endemic country, are in line with recent findings from England.20 21 22 A total of 26 workers seroconverted but none had a clinical diagnosis of hepatitis E. Thirteen of the 26 workers had never visited endemic areas, which suggests that the hepatitis E virus circulates in Switzerland. However, contrary to the hypothesis, none of the four indicators of exposure was associated with seroconversion. At first glance, this finding seems at variance with the results of Vaidya et al23 and El-Esnawy.24 However, Vaidya et al23 examined sewage workers in India, a hyperendemic area, where the possibly more virulent genotype 1 prevails in humans. Furthermore, the virus titres in sewage may be much higher in India8 allowing the infectious dose to be reached more easily. Thus, the different findings may represent two different clinico-epidemiological forms of hepatitis E.1 4 As Egypt is also an endemic area, the increased prevalence found by El-Esnawy24 may be similarly explained.

Peptic ulcers in sewage workers have been previously examined in two cross-sectional studies13 15 with inconclusive results. In the present cohort, the incidence of peptic ulcers did not differ between the exposed and the control group. This finding agrees well with the odds ratio (OR) of 1.4 (95% CI 0.31 to 6.1) reported by Friis et al13 in Swedish sewage workers and with the OR of 0.72 (95% CI 0.32 to 1.61) from the baseline examination of this cohort15 and does not support an increased risk of clinical peptic ulcers in this population.

At baseline H pylori IgG seroprevalence was lower in the exposed than in the non-exposed group.15 It was suggested that unknown bias or residual confounding most likely explained this decreased prevalence. The results of the cohort study support this hypothesis and are in accordance with findings on H pylori IgG from Sweden25 and from a preliminary Belgian study.26 Overall, the available evidence, although still limited, does not confirm an increased risk of H pylori infection in workers exposed to sewage, even though the bacterium can be identified in waste water. Possible explanations are the occurrence in water of the coccoid form which is unable to colonise the human stomach5 and too low an infectious dose.4

The lack of association is unlikely to be explained by a selection bias at baseline, even though 488 eligible workers did not enter the study. However, non-participants were more often foreigners, a subgroup with a fairly high H pylori seroprevalence in this study.15 As the workers not entering the study mainly came from two control plants, their inclusion should have rather increased the seroprevalence of H pylori in the control group and, therefore, their inclusion would not have increased risk in sewage workers. Nor can the lack of association be explained by ill workers leaving the plant. Indeed, job changes during follow-up were not associated with seroconversion for HEV or for H pylori, hepatitis or liver disease or incident peptic ulcer. Furthermore, subjects participating and non-participating until the end of the study were comparable regarding important characteristics associated with H pylori seroprevalence (age, education, nationality, country of childhood) and lifelong duration of exposure to sewage was similar in both groups. Finally, as few workers changed location during follow-up, bias due to high turnover causing misclassifications in exposure is unlikely.

This study was planned as a comparison of survival curves, whereas most other European studies based on serological results have determined seroconversion rates in %/year for the subpopulation having valid results at two time points, that is, without considering cases lost to follow-up. In a comparable subpopulation, the seroprevalence rates were consistent with those of these studies. However, the range is very wide (0.08–22%/year), although possible differences associated with time period or areas were limited by study selection (studies eligible for comparison were reported in fully published original papers, were conducted in Europe and were published between 2000 and 200727^–32). Besides possible genuine differences in transmission of H pylori between areas, different test kits33 and large differences in population selection procedures and in the proportion of cases lost to follow-up probably play a role. Indeed, population selection influences test performance33 and workers lost to follow-up may not be comparable to participants, which modifies the results.32 Moreover, including or excluding borderline cases may have a non-negligible effect on the seroconversion rates as found in this study and reported by Rosenstock et al32 and Kuipers et al.34 However, as serology was not repeated after seroconversion, it was not possible to examine survival curves with different cut-offs. The difference in seroconversion rates between H pylori IgG and H pylori IgA is likely due to the inferior diagnostic performance of the IgA-based tests.33

This study has some limitations. With respect to H pylori seroprevalence, the study was designed to detect a small relative risk, but the power necessary to detect the clinical endpoints could not be assessed meaningfully at the beginning of the study. Indeed, the rate of infection with H pylori in industrialised countries has decreased substantially in the last decades and eradication has greatly reduced the recurrence and duration of peptic ulcers. Therefore, cases of gastritis and eradication were also recorded in order to examine a possible bias due to treatment. The data show no trend towards more cases of clinical endpoints in the exposed group (table 2), which is reassuring. However, a lack of power cannot be definitely excluded and results might have been different if there had been more subjects or if the study had lasted longer.

The use of personal protective equipment is difficult to assess accurately and was based on the baseline interview only. Repeated objective assessments would have been better but may cause a surveillance bias and are very labour intensive. The main limitation is that the Canton of Zurich is not an endemic area and has good sanitation. Therefore, the results of this study may not be applicable to endemic areas or regions with poor sanitation.