Occup Environ Med 69:268-275 doi:10.1136/oemed-2011-100071
  • Workplace
  • Original article

Occupational exposures and risk of stomach cancer by histological type

  1. for the PANESOES Study Group*
  1. 1IFIMAV-Fundación Marqués de Valdecilla, Santander, Spain
  2. 2Escuela Universitaria de Enfermería, Universidad de Cantabria, Santander, Spain
  3. 3Spanish Clinical Research Network (CAIBER), Spain
  4. 4CIBER de Epidemiología y Salud Pública (CIBERESP), Spain
  5. 5Centro de Investigación en Salud y Medio Ambiente (CYSMA), Universidad de Huelva, Huelva, Spain
  6. 6Departamento de Salud Pública, Historia de la Ciencia y Ginecología, Universidad Miguel Hernandez, Elche-Alicante, Spain
  7. 7Departamento de Salud Pública, Universidad de Cantabria, Santander, Spain
  8. 8Área de Epidemiología Ambientaly Cáncer, Centro Nacional de Epidemiología, ISCIII, Madrid, Spain
  9. 9Finnish Institute of Occupational Health (FIOH), Helsinki, Finland
  1. Correspondence to Dr Miguel Santibañez, Research Support Unit, IFIMAV- Fundación Marqués de Valdecilla, 39011 Santander, Spain; ifimav.eclinicos2{at}
  1. Contributors All authors have contributed substantially to conception and design or analysis and interpretation of data, drafting the article or revising it critically for important intellectual content and final approval of the version to be published. They are also aware, and agree, that the manuscript is being re-submitted to OEM.

  • Accepted 6 October 2011
  • Published Online First 7 November 2011


Objective To explore the relationship between stomach cancer (SC), by histological type, and occupations and occupational exposures.

Methods The authors conducted a hospital-based case–control study in south-east Spain. Subjects were 399 incident histological confirmed SC cases (241 intestinal and 109 diffuse adenocarcinomas) and 455 controls frequency matched by sex, age and province of residence. Occupation was coded according to the Spanish National Classification of Occupations 1994. Occupational exposures were assessed by the FINJEM Job Exposure Matrix. ORs were estimated by unconditional logistic regression adjusting for matching variables and education, smoking, alcohol and diet.

Results In men, statistically significant increased risk of the diffuse subtype was found for ‘cooks’ (OR 8.02), ‘wood-processing-plant operators’ (OR 8.13) and ‘food and related products machine operators’ (OR 5.40); for the intestinal subtype, a borderline association was found for ‘miners and quarry workers’ (OR men 4.22, 95% CI 0.80 to 22.14). Significant increased risk was observed between the diffuse subtype of SC and the highest level of exposure to ‘pesticides’ (ORH both sexes 10.39, 95% CI 2.51 to 43.02, ptrend=0.02) and between the intestinal subtype and asbestos (ORH men 3.71, 95% CI 1.40 to 9.83, ptrend=0.07). Restricted analyses of exposures of 15 years and longer showed significant associations between the diffuse subtype and the exposure to ‘wood dust’ (OR men 3.05).

Conclusions This study supports the relationship previously suggested between SC and occupational exposure to dusty and high temperature environments. Several occupations may also increase the risk of diffuse SC but not the intestinal subtype.

What is known on this subject

  • Few studies have provided information on the role of occupational exposures in relation to stomach cancer by histological type, and results are still inconsistent.

What this study adds

  • Some occupations, such as ‘cook’, ‘wood or food-related occupations’ and ‘ship deck crews’, and occupational exposures, such as ‘pesticides’, ‘wood dust’ or ‘low-frequency magnetic fields’, may increase the risk of the diffuse subtype of stomach cancer.

  • Other occupations, such as ‘miner’, and occupational exposures, such as ‘asbestos’, may increase the risk of the intestinal subtype of stomach cancer.

  • The findings of this study in Spain support the association previously reported between occupational exposure to dusty environments and stomach cancer. Although environmental factors had been linked more often to the intestinal histological subtype, in our study, most of the significant associations were observed with the diffuse subtype.

Policy implications

  • Although further studies are necessary to confirm occupational exposures increasing the risk of stomach cancer, measures to minimise exposure to dust in the workplace should be encouraged.


In spite of the decreasing incidence and mortality observed in many parts of the world, including Spain, stomach cancer (SC) is the second leading cause of oncologic death worldwide.1–6 Adenocarcinoma is by far the most common type of SC, and the two histological subtypes proposed by Lauren in 1965—intestinal and diffuse—are accepted and commonly used in epidemiological studies. Two distinct carcinogenesis pathways have been suggested in relation to these histologic subtypes of SC, with differing influences of environmental factors and a varying presence and predominance of genetic factors and molecular changes.7

Regarding risk factors, chronic infection by Helicobacter pylori is considered a causal factor for SC, although other factors such as a diet poor in fruits and vegetables and lacking in antioxidant mechanisms may also play a protective role.1 5 Tobacco smoking is a firmly established risk factor for SC, although that is not the case for alcohol consumption.1 Occupational factors have also been related to an increased risk for SC, and significant associations have been found for several occupational groups such as miners and quarrymen, farmers, fishermen, masonry and concrete workers, machine operators, food industry workers, cooks, ceramic and textile industry workers, launderers, and dry cleaners.8 9 Occupational exposures such as to specific dusts, asbestos and other insulating material, organic solvents, nitrogen oxides, N-nitroso compounds, ionising radiation and pesticides (herbicides, insecticides and fungicides) have been also related to SC.5 10 11 However, information on the role of occupational exposures by the intestinal and diffuse subtype is very scarce. Thus, associations did not differ substantially when analyses were done for intestinal and diffuse types in a Polish case–control study.12 Another case–control study in Sweden found an association of similar magnitude for both subtypes in relation to herbicide exposure.11

We evaluated the association between occupational history and specific exposures in relation to SC risk by histological subtypes in a case–control study carried out in south-east Spain.


Study population

This research was part of the PANESOES Study, a hospital-based case–control study designed to explore the influence of major lifestyles and diet on the risk of three gastrointestinal cancers: pancreas, oesophagus and stomach. Details of this study have been published elsewhere.13 14 The PANESOES Study planned to recruit approximately 200 cases for oesophagus cancer, 200 for pancreatic cancer, 400 cases for SC and 450 controls frequency matched to the expected distribution of case subjects of the overall PANESOES Study by sex, age and province (Alicante and Valencia). Eligible subjects were Spanish-speaking men and women, 30–80 years old, and hospitalised between January 1995 and March 1999 in any of the nine participant hospitals in the provinces of Alicante and Valencia. All subjects were informed of the study objectives and gave their informed consent prior to their inclusion in the study. Research protocols were approved by the local ethics and/or research committees of the participating hospitals and the university.

A total of 434 cases were identified (283 men and 151 women). Nine cases were not interviewed, which leaves 425 cases. Twenty-four diagnoses were not histologically confirmed. From the available 401 cases, two cases did not answer the occupational exposure section of the questionnaire, which left 399 cases (265 men and 134 women).

Cases were classified into intestinal type adenocarcinoma (n=241), diffuse adenocarcinoma (n=109), mixed-type adenocarcinoma (N=11), lymphoma (n=20) and other (n=18). Cases and controls were recruited concurrently. A wide inclusion criterion was used to select controls from diseases not related a priori to the main exposures of interest (tobacco, alcohol and diet). The overall participation rate of the 457 eligible controls was 99.6%, leaving 455 control (285 men and 170 women) subjects with completed interviews for analysis. The distribution of the main diagnostic groups for control subjects was: hernias (28.3%), fractures or injuries (35.2%), appendicitis (6.4%), eventrations (5.5%), acute cholecystitis (2.6%) and other diagnoses (21.9%). Main characteristics of cases and controls are shown in table 1. SC cases were on average 1.7 years older (p=0.03), presented a similar distribution by sex and educational level and reported a higher prevalence of cigarette smoking (p=0.01), alcohol consumption (p=0.03), a lower intake of fruits and vegetables (p<0.01) and a higher daily energy intake than controls (p<0.01).

Table 1

Distribution of socio-demographics characteristics and lifestyle variables among cases and controls

Exposure assessment

Face-to-face interviews were conducted in hospital for all participants by trained interviewers, using a structured questionnaire. Eighty-nine per cent of the interviews in cases were performed with the patient (61% with the patient alone and 28% with the patient receiving help from a relative), whereas in controls, 70% were performed alone and 26% with help. Proxies provided information for the other cases and controls.

While interviewers could not be blinded to the case/control status, they were unaware of the main study hypothesis and were trained to administer strictly the structured questionnaires in an equal manner to case and controls alike.

We collected information on demographic characteristics, tobacco and alcohol use, medical history, diet and other lifestyle factors, and occupational history. The interview elicited details on usual tobacco use including tobacco type and the age at which the habit started and stopped. A ‘never-smoker’ was defined as someone who had smoked fewer than 100 cigarettes or less than one cigarette per day for 1 year. A ‘former smoker’ was defined as someone having stopped smoking ≥1 year before the interview. Alcohol consumption patterns were assessed through enquiries into the usual intake of alcoholic beverages. A ‘never-drinker’ was someone having consumed less than one drink per month. A ‘former drinker’ was defined as having stopped drinking at least 1 year before the interview.

The fruit and vegetable intake was assessed by food frequency questionnaire. For this study, we adapted and validated a food frequency questionnaire of 93 food items similar to the Harvard questionnaire in order to assess the diet 5 years before the interview in the hospital.14 15 Participants were asked to report their average consumption of 12 vegetables and 10 fruit items. The average daily intakes for each fruit and vegetable were summed to compute the total fruit and vegetable intake in grams, and they were adjusted for energy intake using the residual method.14 16

We collected information for the two main occupations for each person: job title, number of years worked and the products and/or substances that were used in each occupation. Main occupations were defined as those occupations in which a subject had worked longest. All reported occupations were coded according to the Spanish National Classification of Occupations of 1994 (CNO94), which is based on the European Union version of the International Standard Classification of Occupations, ISCO 88 (COM). The coding process was carried out by an occupational physician (MS) who was blinded to the case–control status of the study subjects. The proportions of cases and controls reporting one, two or three occupations were very similar (p=0.29). The mean of the total number of worked years was 39.9 (SD=15.8) in cases and 38.3 (SD=14.8) in controls (p=0.14). With respect to the variability of the occupations, the 399 cases reported 114 different occupation codes, while the 455 controls reported 119.

In addition, we used the FINJEM job exposure matrix17 to explore occupational exposure to 21 chemical agents, four physical exposures and two ergonomic exposures. FINJEM has three dimensions: occupations, agents and exposure periods. Exposure to an agent in an occupation during a period of time is characterised with respect to the prevalence of exposure (probability) and the average level of exposure (intensity), mostly in milligrams per cubic metre or in parts per minute. FINJEM was designed by a team of scientists experienced in assessing physical, chemical, ergonomic and psychosocial exposures, making use of Finnish industrial hygiene measurements, interview surveys and workforce surveys. For this study, the CNO94 job titles were converted into Finnish Occupational codes according to the Combined Occupational Classification of Finnish Censuses in 1970–1985 and then into probabilities (estimated proportions of exposed, range 0.06–1) and into mean levels of exposure to the 27 selected agents during the period 1960 through 1984. We used their product as our exposure metric to classify exposure to each agent in three categories: substantial (high), low and unexposed. The cut-off points between low and substantial were set as close as possible to the 75th percentile of the distribution of the product of the probability and the intensity of exposure in controls.

Statistical analysis

We estimated the risk of SC for those occupations held for at least 1 year in both sexes and in men and women separately. In addition, we only estimated the effects of those job titles if at least 10 exposed men or women (either cases or controls) were observed (five subjects for a priori high-risk occupational groups), rising from the more general one-digit classification of major occupational groups to the more specific four-digit classification of occupations (subcategories). The reference group comprised subjects who had never held the specific occupation. Agents assessed by FINJEM were incorporated as categorical variables on a 3-point scale. We performed analyses for all types of SC combined and also for the intestinal, diffuse adenocarcinoma and lymphomas. We also performed restricted analyses for exposures of at least 15 years.

Adjusted ORs and 95% CIs were calculated using unconditional logistic regression. The following potential confounders were included in the models: sex, age (<60, 60–70 and >70 years), reference hospital (Valencia/Alicante), educational level (lower than primary, primary completed and higher than or equal to secondary school), smoking (non-smokers and four categories for ever smokers based on pack-years), alcohol use (never-drinker and three categories according to tertiles of total grams of alcohol consumption in former drinkers and current drinkers, respectively), the energy adjusted daily intake of fruits and vegetables in grams per day (in tertiles) and the total energy intake as continuous variable as suggested by Willett.16

Tests for trend in the ORs across exposure strata were calculated for occupational exposures by using logistic models that included categorical terms as continuous variables in a model with all the potential confounders. For trend tests, we used the likelihood ratio test statistic with 1 degree of freedom. The level of statistical significance was set at 0.05, and all tests were two tailed. All analyses were performed with SPSS V.15.0.


Table 2 shows the association between a selection of job titles and SC by histological subtype in men. Significant associations were found in men between diffuse adenocarcinoma and several occupations such as ‘cooks (code 5010 of the CNO94)’ (OR 8.02, 95% CI 1.14 to 56.34), ‘employed skilled workers in agricultural activities (code 602)’ (OR 6.16, 95% CI 1.10 to 34.60), ‘wood-processing-plant operators (code 8141)’ (OR 8.13, 95% CI 1.52 to 43.55) and ‘food and related products machine operators (code 837)’ (OR 5.40, 95% CI 1.21 to 24.08).

Table 2

Adjusted OR for stomach cancer, intestinal adenocarcinoma, diffuse adenocarcinoma and lymphoma subtype, according to occupations in Spanish men working at least 1 year in the same occupation

Although statistically non-significant, direct associations were also found between diffuse adenocarcinoma subtype in men and wood-related occupations such as ‘carpenters and joiners (code 7130)’ (OR 4.26) and ‘cabinetmakers and related workers (code 7920)’ (OR 1.87). Evidence of association was also observed in men between the diffuse subtype and ‘ship deck crews and related workers (code 8550)’ (OR 10.90, 95% CI 0.87 to 136.85).

In women, non-significant associations were found between diffuse adenocarcinoma subtype and ‘cooks (code 5010)’ (OR 2.01, 95% CI 0.35 to 11.49, data not shown in table) and ‘self-employed skilled workers in agricultural activities (code 601)’ (OR 1.97, 95% CI 0.31 to 12.62, data not shown in table).

For the intestinal type, evidence of a borderline association was found in men for ‘miners and quarry workers (code 7421)’ (OR 4.22) and in ‘waiters, waitresses and bartenders (code 5020)’ (OR 2.01).

For the lymphoma subtype, among occupations with a higher number of exposed cases (N=2 exposed cases), a statistically significant association was found for SC in ‘shoe-makers and related workers (code 7942 of the CNO94)’ (OR both sexes 10.38, 95% CI 1.20 to 89.66, data not shown in table).

Tables 3 and 4 show the associations between a selection of occupational exposures and SC by histological subtype in men and women, respectively. We found a significant association between diffuse adenocarcinoma and a high exposure to pesticides in both sexes (OR at the high level of exposure (ORH) 10.39, 95% CI 2.51 to 43.02, ptrend=0.02, data not shown in tables). Statistical significance remained when the analysis was restricted only to men (see table 3).

Table 3

Selected occupational exposures by FINJEM in Spanish men for which borderline or significant associations were found, and risk of stomach cancer (for all histological types and for intestinal adenocarcinoma, diffuse adenocarcinoma and lymphoma subtype separately) in terms of duration of exposure

Table 4

Selected occupational exposures by FINJEM in Spanish women and risk of stomach cancer (for all histological types and for intestinal adenocarcinoma, diffuse adenocarcinoma and lymphoma subtype separately) in terms of duration of exposure

Regarding dusty exposures, a significant association was found between diffuse adenocarcinoma and ‘wood dust’ when the low- and high-exposure categories were grouped and the analyses were restricted to exposures longer than 15 years (OR men 3.05, 95% CI 1.11 to 8.32).

A significant association was also found for diffuse adenocarcinoma in men and the exposure to low-frequency magnetic fields (OR 2.06, 95% CI 1.10 to 3.85). The association was also observed when the low- and high-exposure categories were grouped, and the analysis was restricted to exposures longer than 15 years (OR 1.92, 95% CI 1.00 to 3.68).

For intestinal adenocarcinoma, a significant association was found in men for a high-level exposure to asbestos (ORH 3.71, 95% CI 1.40 to 9.83, ptrend=0.07), which remained when the analysis was restricted to long duration exposure (ORH 3.21, 95% CI 1.07 to 9.63).

The associations for the rest of occupational exposures by FINJEM are presented for men and women in the additional online supplementary tables 1 and 2, respectively.


We have found increased risk of SC among cooks, wood and food-related occupations for the diffuse adenocarcinoma subtype of SC. Increased risks were found among ship deck crews for the diffuse adenocarcinoma subtype and in miners for the intestinal adenocarcinoma subtype.

Regarding specific occupational exposures, we observed significant associations between diffuse adenocarcinoma subtype and exposure to ‘pesticides’, ‘wood dust’ and ‘low-frequency magnetic fields’ and between the intestinal subtype and ‘asbestos’ exposure.

Most of the published studies in SC have been focused on dust-exposure occupations8 10 with several studies showing positive results, mainly related to minerals (particularly coal dust),9 18–23 ceramic, textile, metal and wood dust.20 24–27 After inhalation, dust may be cleared by the lung and then swallowed, allowing potential carcinogenic compounds to act in the precursor lesions of SC.10 In a case–control study on occupational exposures and SC in Spain,20 an increased non-significant risk was observed for wood and furniture workers (OR 1.76) and a borderline association for occupations related to silica and mineral dust exposures (OR 1.80, 95% CI 0.90 to 3.59) after adjusting for socio-professional status and dietary habits.

As geological differences exist between Spain and Finland, we decided not to use variables on mineral dust exposure, including quartz dust, therefore exposure to mineral dust was not specifically assessed through FINJEM in our study.28 However, a strong effect of borderline significance was found between the intestinal subtype and miners and quarrymen. Although no coal mines exist in south-east Spain, there are other industries related to ceramic and other building materials probably generating other mineral dusts of various kinds. Regarding ‘wood dust’, we found some statistically significant associations between the diffuse adenocarcinoma subtype and wood-related occupations. In addition, a significant association for ‘wood dust’ exposure assessed by FINJEM was found but no evidence for a dose–response. Overall, our results support the suggested association between SC and exposure to occupations with dust environments.

Working under conditions of heat stress or in occupations that involve heavy work has been related to SC in follow-up and case–control studies, ecological studies and experiments with animals. It has been suggested that a high salt intake to recover the amount of salt lost by sweating could be responsible of such an association.29 We did not find such associations with heat or perceived physical workload assessed through FINJEM, although our association found in cooks might support this hypothesis.

The association between pesticide exposure and SC found in some studies has not been confirmed in others.10 11 30–32 Although the association we found for the diffuse adenocarcinoma might support in part such association, this finding would require further confirmation.

We also found a significantly increased risk for intestinal subtype in men in the highest category of asbestos exposure. Since the job titles of these cases were especially ‘carpenters and joiners’ (n=9) and ‘miners and quarry workers’ (n=8), and there are no mines with asbestos exposure in Spanish, the association could be due to some over-assignment of exposure to asbestos from FINJEM in Spanish workers. Other studies have also shown mixed results on this exposure.10–12 33–35 The observed association in relation to ‘low-frequency magnetic fields’ is only supported among Finnish women,36 and thus, we cannot disregard a false-positive result.

Several studies have found an excess risk of SC in food industry workers.23 37 38 We found a significantly increased risk of SC for ‘food and related products machine operators’ (code 837 of the CNO94) but not so in ‘food processing and related trades workers’ (code 78 of the CNO94). The evidence of an increased risk of diffuse adenocarcinoma in our study for deck crews is in agreement with results from previous published studies in sailors10 39 and fishermen.2 9 39–41 Dietary factors (high smoked food and low vegetable intake) were thought responsible for this excess risk.8

The use of case–control studies to explore the effect of occupational exposures may present serious limitations, particularly if exposures are uncommon. Thus, we mainly focused on those exposures with at least 10 exposed subjects or five subjects if an association had previously been reported. Nevertheless, we should be cautious about those associations based on a low number of observations even if they were previously reported. Regarding statistical power, our study had a larger sample size than a previously published case–control study in Spain20 and detected several significant associations, some with a significant dose–response. Another limitation relates to the high number of comparisons made in our study; thus, we cannot rule out that some of the associations were spurious or due to chance (ie, false-positive associations).

In the PANESOES Study, we did not use the lifetime history for every job held for at least 1 year but the two most important occupations for each person (a third occupation was reported by only a few cases and controls because of some doubts about their relevance). Seventy per cent of cases reported only one occupation in their life with an average working time of 39.2 years, 29% of cases reported two occupations with an average working time of 41.6 years in both occupations and only 1% of cases (n=4) and 1.8% of controls (n=5) reported a third occupation with an average working time of 48.0 years in cases and 31.2 years in controls.

We should also note that exposure estimates of FINJEM were not adapted to specific Spanish exposure circumstances and other factors such as H Pylori infection could cause confounding. However, it has been suggested that associations found for some occupational exposures, such as phenoxy herbicides, seem to be independent of the H Pylori infection.11

On the other hand, since we compared the effect for every job title and occupational exposure in relation to the remaining ones (reference category), it may be possible that the effects were in fact stronger because of the inclusion among the non-exposed groups of other occupations and substances also causing an increase in risk of SC. The presence of other potential biases was minimised by selecting cases with pathological confirmation and controls from the same hospital as the cases, by the use of the same interviews and procedures in both cases and controls to avoid any differential misclassification and by the blinding of the case–control status when coding job titles.

Overall, we consider that our findings based on job titles and on specific exposures as evaluated by FINJEM show a good correspondence. Thus, most of the specific exposures for which significant associations were found were included in the occupations showing increased risks of SC. Cooks, carpenters, cabinetmakers and food machine operators were deemed to be exposed to low-frequency magnetic fields by FINJEM. Selfemployed skilled workers in agriculture and wood-processing-plant operators were deemed to be exposed to pesticides. Finally, all the wood-related occupations at higher risk of SC, such as wood-processing-plant operators, carpenters, cabinetmakers and related workers, were deemed by FINJEM to be exposed to wood dust.

In conclusion, our study, although explorative, gives support to several relationships previously reported between SC and the occupational exposure to dusty environments and working in high-temperature environments. Although previous studies have suggested a major role of environmental factors on the intestinal adenocarcinoma subtype, in our study, occupational factors were more related to the diffuse adenocarcinoma subtype.


We would like to appreciate the writing assistance provided by Mr. Jonathan Whitehead.


  • * Members of PANESOES Study Group: Jesus Vioque (principal investigator and coordinator of the study), Esperanza Ponce, María Guillén, Miguel Santibáñez, Xavier Barber, Manuela García de la Hera, Departamento de Salud Pública, Universidad Miguel Hernández, Elche-Alicante, Spain; Miguel Bixquert, Jorge Alonso, Vicente Cervera, Remedios Giner, Juan Ruiz, Carlos Sanchos-Aldás, Javier Arenas, Hospital Arnau Vilanova de Valencia; Joaquin Berenguer, Teresa Sala, Sonia Pascual, Liria Argüello, Marco Bustamante, Salvador Sancho, Constantino Herranz, Jorge Aparicio, Dr. Baixauli, Jorge Mir, Pedro Sendrá, Hospital La Fe de Valencia; Enrique Medina, Alicia Tomé, Luis Ferrer, Ramón Truyenque, Luis Olabarrieta, Ricardo Fabra, Carlos Camps, Jose Maria Vicent, Hospital General de Valencia; Eduardo Moreno-Osset, Ramón Añón, José Ballester, Vicente Alfonso, Dr. Martínez-Abad, Francisco Blanes, Carmen Molins, Daniel Almenar, Santiago Olmos, Dr. Fenollosa, Hospital Doctor Peset de Valencia; Adolfo Benages-Martinez, Andrés Peña-Aldea, Dra. I. Pascual, Dr. García-Conde, Andrés Cervantes, Pilar Azagra, Dr. Lledó, Blas Flor, Vicente Martí, Hospital Clínico de Valencia; Miguel Pérez-Mateo, Juan Antonio Casellas, Eva Girona, Jose Ramón Aparicio, Mar López, Antonio Arroyo, Fernando Camuñas, Jesus de Anta, Hospital General de Alicante; Juan Custardoy, Concepción Martínez, Enrique Gaspar, Eduardo Muñoz, Hospital Comarcal de la Vega Baja; Alfredo Carrato, Maria Luisa Gozálvez, Rafael Calpena, Dr. Gassent, Dr. Pérez, Carlos Sillero C, Hospital General de Elche; Justo Medrano, Francisco Mauri, Marta Corona, Jorge Minguel, Hospital Universitario Sant Joan de Alicante.

  • Funding This work received financial support from the Spanish Ministry of Health (Instituto Salud Carlos III FIS 91/0435), the Generalitat Valenciana (CTGCA/2002/06; G03/136, EVES 030/2005), Fundación Bienvenida Navarro Luciano-Trípodi and Fundación Hospital Universitario de Elche (code 01/07).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval As noticed in methods, the study was approved by the ethics committee of the participating hospitals, and patients gave their informed consent prior to their inclusion in the study.

  • Provenance and peer review Not commissioned; externally peer reviewed.