Epidemiology Group, 30 Hyde Terrace, University of Leeds, UK; 6Centre for Digestive Diseases, The General Infirmary at. Leeds, Great ...... cDepartment of General Practice, Wellington School of Medicine, PO ... Marsh Street Surgeryo, High Field Surgeryo, Lingwell ... Robin Lane Medical Centreo, Manor Park Surgeryo,.
Ó Springer 2005
European Journal of Epidemiology (2005) 20: 455–465 DOI 10.1007/s10654-004-6634-0
INFECTIOUS DISEASES EPIDEMIOLOGY
The association between Helicobacter pylori infection and adult height Paul Moayyedi1, David Forman2, Sara Duffett3, Su Mason4, Julia Brown4, Will Crocombe4, Richard Feltbower5 & Anthony Axon6 on behalf of the Leeds HELP study group 1
AstraZeneca-Richard Hunt Chair of Gastroenterology, Department of Medicine, McMaster University, 1200 Main St. W., HSC 3N51D Hamilton, Ontario, Canada; 2Centre for Cancer Research, University of Leeds, Arthington House, Cookridge Hospital, Leeds; 3Institute of Epidemiology and Health Services Research, 30 Hyde Terrace, University of Leeds, Leeds; 4 Northern and Yorkshire Clinical Trials and Research Unit, University of Leeds, Springfield Mount, Leeds; 5Paediatric Epidemiology Group, 30 Hyde Terrace, University of Leeds, UK; 6Centre for Digestive Diseases, The General Infirmary at Leeds, Great George Street, Leeds, UK Accepted in revised form 23 November 2004
Abstract. Objectives: A cross-sectional survey was performed to evaluate the association between H. pylori and adult height. Methods: H. pylori infection was assessed using a 13C-urea breath test and height measured by a research nurse using a stadiometer in participants between the ages of 40– 49 years. Results: Height was measured in 2932/3682 participants that attended and were evaluable. H. pylori infected women were 1.4 cm shorter than uninfected women (95% confidence interval,
CI ¼ 0.7–2.1 cm) and this statistically significant difference persisted after adjusting for age, ethnicity, childhood and present socio-economic status (H. pylori positives 0.79 cm shorter; 95%CI: 0.05– 1.52 cm). H. pylori positive men were 0.7 cm shorter than uninfected men but this did not reach statistical significance (95% CI: )0.1–1.5 cm). Conclusion: Although H. pylori infection is associated with reduced adult height in women, this maybe due to residual confounding.
Key words: Cross-sectional survey, Height, Helicobacter pylori, 13C-urea breath test
Introduction The World Health Organisation has stated that short stature is one of the most useful indices of childhood well being [1]. The assessment of growth and subsequent adult height is important to public health workers and for researchers concerned with the development of children and the health of adults [2]. Diminished adult height is associated with increased risk of ischaemic heart disease [3, 4], stroke [5] and general ill health [6]. A prospective cohort study suggested that the risk of dying from any cause during the 20 year follow-up was increased with a relative risk of 1.13 (95% CI: 1.07–1.19) for every 10 cm reduction in height after adjusting for confounding factors [7]. The reason for the association between adult height and mortality is uncertain but it is likely that factors that determine childhood growth are also linked with diseases in later life. Body height is determined by both genetic and environmental factors. Poor nutrition and serious chronic childhood infections reduce growth in the developing world [8, 9]. These have largely been overcome in the developed world but is likely that the environment continues to influence growth as there has been a
secular trend towards increasing height even in recent times [10] and childhood morbidity is still associated with small stature in the UK [11]. The infections that play a part in this are likely to be much subtler than those seen in developing countries as common selflimiting illnesses are unlikely to have a lasting impact on childhood growth. H. pylori infection is a plausible candidate for causing growth retardation, as it is common even in developed countries. The organism is often acquired in infancy or early childhood and infection usually persists throughout life. H. pylori infects 30–50% of the normal population in the United Kingdom and is the cause of 80–95% of gastric and duodenal ulcers [12]. This infection is also strongly associated with gastric cancer [13] and it has been implicated in diseases outside the gastrointestinal tract [14].The possibility that H. pylori infection may cause disease outside the gastrointestinal tract is biologically plausible as the organism causes a life-long gastric chronic inflammatory response that may have systemic consequences [15]. These systemic inflammatory effects might be expected to impact on growth and final adult height. Patel et al. [16] assessed children at the age of seven measuring their height at baseline and 4 years later.
456 H. pylori infection significantly reduced growth in girls after adjusting for confounding factors but no effect was seen in boys. It was not clear whether H. pylori had simply delayed the onset of the pubertal growth spurt or whether the infection may have an influence on final adult height. A study of European adults suggested H. pylori infection was associated with reduced height but this was lost after adjusting for education and lifestyle factors [17]. Two further studies have reported an association between H. pylori and reduced adult height after controlling for present socio-economic status [18, 19]. These studies did not control for childhood socio-economic status and it is possible that H. pylori infection is acting a marker for deprivation in early life. We have evaluated the association between H. pylori infection and adult height in a cross-sectional survey controlling for indicators of childhood deprivation.
Methods Design This was a cross-sectional survey of 40–49 year old subjects recruited as part of a trial evaluating the medical benefits and health economics of population H. pylori screening and treatment in primary care. The main results of this trial have been published elsewhere [20]. Subjects on the list of 36 primary care centres were selected using computer generated random numbers. The age range of 40–49 years was selected to maximise the probability of H. pylori infection whilst minimising the possibility of suffering from serious illness over the follow-up period. Subjects were invited to attend their local practice to complete a questionnaire and have a 13C-urea breath test (13C-UBT). The design for the main trial involved giving subjects antibiotic therapy and therefore pregnant or breast feeding women, subjects taking concomitant anti-convulsant therapy, warfarin, theophyllines, terfenadine or astemizole, allergy to proton pump inhibitors, macrolides or 5-nitroimidazoles were excluded. Subjects taking proton pump inhibitor within one week or antibiotics within two weeks of the 13C-UBT were also excluded. The Leeds and Bradford Research Ethics Committees approved the trial. Assessment of exposure A non-fasting 13C-UBT was performed according to standard protocols. Participants were given 4 g citric acid in 200 ml of water and two teaspoons of refined sugar to delay gastric emptying and provided a breath sample by exhaling through a straw into four exetainers. They were then given 100 mg of 13 C-labelled urea (99% pure: Boston isotopes) and a further breath
sample was taken at 30 min. All samples were analysed using a mass spectrometer (ABCA-NT, Europa Scientific) at the Leeds General Infirmary and an excess d13 CO2 excretion of greater than 5 per ml was considered positive for H. pylori infection as defined by the European standard [21]. This protocol has been validated in our local population [22] and was found to have 98% sensitivity and 96% specificity for detecting H. pylori compared with a gold standard. A researcher also interviewed the participant on childhood and adult socio-economic factors that could influence H. pylori status and adult height using a standard questionnaire. Assessment of outcome The majority of subjects were H. pylori negative. To reduce the cost of the study, the height of only a proportion of uninfected subjects was assessed. It was important that the investigator and participant were unaware of their H. pylori status so height was measured at the follow-up interview. Subjects defined as H. pylori positive according to 13C-UBT were randomised to receive eradication therapy (omeprazole 20 mg bd, clarithromycin 250 mg bd and tinidazole 500 mg bd for 1 week) or identical placebo using computer generated random number tables. In addition a proportion of H. pylori negative subjects were randomised to receive placebo so that a numerically comparable group of uninfected cases could be followed-up. The remaining H. pylori negative subjects were informed of their H. pylori status and withdrawn from the study. The subjects and researchers were unaware of the H. pylori status as both infected and uninfected cases were followed-up at 6 months and 2 years. A trained nurse assessed height at the 6-month interview. Height was measured with participants in stocking feet using a wall-mounted stadiometer. Statistical analysis Power calculations and sample size The sample size calculations for this study were based on dyspepsia and health service dyspepsia costs, as these were the main outcomes of the study. The sample size these outcomes dictated was then used to calculate the difference in height the present study had the power to detect. A sample size of 500 uninfected and 1000 infected men would detect a 1.0 cm difference in height between the two groups at the 80% power and 95% confidence level assuming a standard deviation of 6.8 cm [7]. Similarly, 500 uninfected and 1000 infected women would detect a 0.9 cm difference in height between the two groups at the 80% power and 95% confidence level assuming a standard deviation of 6.1 cm [7]. These power calculations also assumed a 10% drop-out rate.
457 Analysis of results Height in the H. pylori positive and negative groups and other dichotomous variables was compared using the Student’s independent t-test without assuming equal variance. Height was compared in variables with more than two groups using one way analysis of variance and simple linear regression if appropriate. Factors associated with H. pylori infection were evaluated by calculating the odds ratio and 95% confidence interval for infection. Data were also entered into a multiple linear regression model to determine independent predictors of height. The variance in the data explained by the model was calculated by r2 using the formula r2 ¼
modelv2 2xðdegrees of freedomÞ : 2 log likelihood
Outliers were identified using Cook’s distance and influential points were assessed using leverage values. These data were omitted from analysis to exclude the possibility that the conclusions of the logistic regression model were influenced by a small number of atypical results. A p value of
