Randomised controlled trial of Helicobacter pylori testing and endoscopy for dyspepsia in primary care

Participants

All patients aged 18-49 years who consulted their general practitioner with dyspepsia of more than four weeks duration were eligible for the trial. We excluded patients who had had endoscopy or a positive barium meal examination in the past three years, who were unable to give informed consent, or who were unfit for endoscopy. Dyspepsia was defined as epigastric pain or heartburn with or without nausea and bloating.11

Randomisation and concealment of allocation

We randomised patients individually using sealed, opaque, sequentially numbered envelopes. The randomisation schedule was done on a 60:40 basis (study: control) and used a computerised random number sequence without blocking or stratification. We kept a log of numbers issued to practices.

Interventions

Patients were randomised to “test and endoscopy” or to usual management. The Helisal test (Cortecs Diagnostics, Deeside) was done by the general practitioner or practice nurse. Endoscopies on patients with positive results were carried out according to usual practice at open access services at six local hospitals. Patients with negative results were not referred for endoscopy but received empirical acid suppressing drugs chosen by their general practitioner.

Patients randomised to the control arm of the trial were managed according to the practitioner's usual management strategy. This allowed outpatient referral to a specialist gastroenterologist but excluded initial referral to open access endoscopy. Patients in whom initial management failed could be referred for endoscopy after six weeks.

Outcomes

The main outcomes were effectiveness (assessed by symptoms) and costs of managing dyspepsia. We measured symptoms at recruitment and 15-18 months using the Birmingham dyspepsia symptom score, a postal measure previously validated in the local population.12 We calculated the costs of dyspepsia from a health service perspective. We assessed use of resources in primary and secondary care for 12 months after randomisation by abstracting data from primary care case records. All data were double entered, and we verified inconsistencies by referring to the original case records.

We used a questionnaire derived from a validated measure for patients with peptic ulcer disease to measure quality of life in terms of pain, emotion, and social function.13 Patient satisfaction was assessed by a validated measure of satisfaction with the primary care consultation14 supplemented with additional questions relating to secondary care and endoscopy.

Analysis

We analysed data by intention to treat. A sample of 430 patients would detect differences of 2 units (SD=4) in the dyspepsia score, 9 units (SD=22) in the pain dimension of the quality of life, and 8 units (SD=20) in the emotion and social dimensions, and would detect a reduction in general practice consultation rates from 3 to 1 a year (SD=3). These estimates were based on a power of 90% at the 5% significance level and assumed 25% loss to follow up. Ethical approval was obtained from all local research ethics committees.

We recorded numbers of endoscopies, barium meal examinations, and primary care consultations from each patient's notes. Drugs prescribed, including those for eradication of H pylori, were recorded as defined daily doses of drug per patient. Table 1 shows the unit costs for management, which we obtained from national reference sources for 1998.

Secondary analyses included comparison of procedure, rates of diagnosis, and use of services in the first year after randomisation; changes in quality of life; and patient satisfaction. We compared use of resources per patient using t tests18 and changes in symptom and quality of life scores from baseline to 18 months using two sample ttests. We identified variables relating to response rates using multiple logistic regression analysis and included them in analysis of covariance models to assess their effect on symptom and quality of life scores. Data were analysed with SAS (version 6.12).

Interventions and diagnostic findings

The Helisal test gave positive results in 40% (112/278) of patients (fig 2). The expected prevalence of H pylori in this population was 30%.12 We used the test performance and the observed numbers of positive and negative results to calculate underlying prevalence and predictive values. With a sensitivity of 89% and a specificity of 84%, a positive rate of 40% reflected an underlying prevalence of 33%.9 On this basis, 27% of positive results and 6% of negative results would be false.

FIG 2

Numbers of participants having open and direct access endoscopy in study and control groups

• Download figure
• Open in new tab
• Download powerpoint

Overall, 127 (45%) of the study group had endoscopy compared with 48 (25%) of the control group. Figure 2 shows the numbers of patients who had open access endoscopy, consultant booked endoscopy, and no endoscopy. Fifteen patients who were positive for H pylori did not have endoscopy, of whom one refused, nine did not attend, and five were not referred. Of the 166 patients who were negative for H pyloriin the test group, 30 (18%) had endoscopy during the 12 month follow up. Only one of these patients had open access endoscopy within six weeks. No control patients were inappropriately managed by test and endoscopy.

Among the patients who had endoscopy, significantly more peptic ulcers were detected by the test and endoscopy strategy than by standard management (21 (7%) v 4 (2.1%), χ²= 6.4, df=1, P=0.011). Compared with the control patients, fewer patients in the study group had oesophagitis (17% v 31%, χ²= 4.1, df=1, P=0.04) and more had duodenitis (19% v6%, χ²= 4.3, df=1, P=0.04; table 3).

Outcomes and costs

Symptoms and quality of life scores in the test and control groups significantly improved by 18 months. There was no evidence of a difference in the size of improvement between the groups (table 4). Non-respondents were more likely to smoke and were younger than respondents (smoking odds ratio=1.63, 95% confidence interval 1 to 2.65; age 0.96, 0.93 to 0.99), but no difference in sex or baseline symptoms was observed. Analysis of covariance found that age and smoking had no significant effect on symptoms or quality of life. No significant differences were observed in the satisfaction questionnaire.

There were no significant differences in the use of drugs for dyspepsia between the groups (table 5). The numbers of outpatient attendances, general practice consultations, or regimens for eradication of H pylori did not differ significantly between the groups (table 6). Mean total costs were £367.85 for test and endoscopy and £253.16 for usual management. This increased cost of £114.69 per patient was not associated with any significant difference in effects. The test and endoscopy strategy was thus less cost effective than usual management.