We approached 209 GPs in the State of Hesse of whom 35.4% agreed to participate in the study. Only GPs being prepared to undergo random recruitment quality controls could take part. Participating practices had to recruit consecutively every attending patient who had chest pain either as presenting complaint or on questioning. The recruitment period lasted 12 weeks for each practice. For logistical reasons recruitment was staggered in four waves between October 2005 and July 2006.

Every patient above 35 years with pain localized in the area between clavicles and lower costal margins and anterior to the posterior axillary lines was to be included. Doctors were also asked to recruit at home visits and emergency calls. Patients were eligible irrespective of the acute or chronic nature of their complaint, or of previously known conditions including Coronary Heart Disease (CHD) or related risk factors. However, patients whose chest pain had subsided for more than one month, whose current episode of chest pain had been already investigated resulting in a definite diagnosis and/or who came for follow-up investigations for chest pain (e.g. an orthopaedic doctor had already established a diagnosis and the GPs continues the treatment) were excluded. In emergency situations without sufficient time for patient information and written consent, relevant clinical items were documented on a case report form (CRF) kept by the GPs. Later, e.g. after discharge from hospital, the patient was asked to participate in the study. Only if he gave informed consent the CRF was handed over to study personnel. Like the whole study protocol, this procedure was approved by the Ethics Committee of the Faculty of Medicine, University of Marburg. The study complies with the declaration of Helsinki.

GPs took a standardized history and performed a physical examination according to a CRF that was piloted and modified accordingly. We chose determinants based on results of qualitative interviews conducted with GPs on how they diagnose chest pain. The CRF contained items about patient (e.g. patients known to the GP, patient different as usual) and pain characteristics (time, duration, location and radiation), accompanying symptoms (e.g. dyspnoea, cough, vomiting), risk factors for CHD, results of a basic physical examination and the GP's presumed diagnosis. GPs also recorded their preliminary diagnoses, investigations and management related to the patients' chest pains. Patients were contacted by phone six weeks and six months after the index consultation. Study assistants blinded to clinical data that already had been recorded asked about the course of their chest pain, treatments including hospitalisations and drugs. Discharge letters from specialists and hospitals were requested from GPs.

Participating practices were recruited from a network of research practices associated with our department. To GPs we emphasized the importance of recruiting every patient with chest pain, irrespective of the presumed likelihood of CHD. Practices were visited at four week intervals to check CRFs, recruitment logs and compliance with study procedures. Random quality controls were performed by searching routine documentation of participating practices to identify cases of chest pain not included in the study.

A reference panel consisting of one cardiologist, one GP and one research staff of the department of Family Medicine reviewed baseline and follow up data of each patient. As far as possible the definition and diagnostic recommendations of the guideline "Reflux" of the German society for digestive and metabolic diseases were taken into account. However, investigations as suggested by the guideline were often not performed by GPs. Therefore, data available to the reference panel were often limited. Apparently the diagnosis of GERD was made on the basis of the history alone. Our 'delayed-type reference standard' may provide some reassurance against alternative causes, e.g. peptic ulcer or carcinoma, of participating patients' complaints. The GP's initial diagnosis contributed to the decision made by the panel.

Our main analysis for distinctive clinical characteristics associated with GI disease and GERD is based on the sample of all patients with chest pain where diagnostic classification was possible. For univariate analyses we calculated separately for GI disease and GERD diagnostic odds ratios (OR) for all items covered by the CRF. To arrive at a small subset of criteria for clinical recommendation we selected those index test items that had a p-value < 0.05 and ORs indicating at least moderate diagnostic accuracy, i.e. OR+/- >2 or <0.5 (univariate analysis). Those were included as predictive variables in multivariate logistic regression analysis. The outcome variables were GI disease and GERD. Variable selection was conducted using the backward stepwise procedure (p < 0.05). Odds ratio and 95%-confidence intervals were calculated. All analyses were performed with SPSS software version 14.0.

The majority of participating GPs were male (67%), two thirds of practices were located in urban areas (63.5%). Mean age of GPs was 49 years. According to our estimate these 74 GPs encountered around 190.000 patients during the study period and approached 1355 patients with chest pain. 7 patients did not meet the inclusion criteria after review of patient records by the study investigators and were consequently excluded; 99 refused to participate in the study. 60 cases were lost to follow up and 11 died but provided enough information to be judged by the reference committee; 3 cases were early drop outs and were therefore not included. For 34 cases follow-up information was lacking, incomplete or ambiguous so that no final diagnosis could be made. We thus analysed 1212 patients for the aetiology of their chest pain; of these 71 were diagnosed as having GI disease including 42 patients with GERD (figure 1).

The prevalence of chest pain during the study period was 0.7% (CI: 0.66-0.73) of all patient contacts. Major diagnostic groups were: GI disease 5.8% (95% CI: 4.5% - 7.1%), Chest Wall Syndrome 46.6% (95% CI: 43.8% - 49.4%), Ischemic Heart Disease (stable) 11.1% (95% CI: 9.4% - 12.8%), psychogenic disorders 9.5% (95% CI: 7.9% - 11.3%) and upper respiratory infections 8.1% (95% CI: 6.6% - 9.8%).

Gastroesophageal reflux disease (GERD) alone contributed with 3.5% (95% CI: 2.6% - 4.7%) (n = 42), benign stomach problems with 2.1% (95% CI: 1.5% - 3.1%) (n = 26) and diseases of the biliary tract with 0.3% (CI: 0.1% - 0.7%) (n = 3) to the aetiology of chest pain.

Compared with CHD patients, patients with GI disease and GERD tend to have a lower mean age and a higher proportion of females. While around half of the patients with GI disease had chest pain at the time of consultation, this was only the case with one third of CHD and GERD patients. In comparison to the other 2 groups, patients with GERD were presenting in a lower proportion with acute, i.e. duration <48 hours, chest pain (table 1).

26.8% of patients with GI disease and 19.0% with GERD complained of vomiting. In 21.1% (GI disease) and 21.4% (GERD) pain was worse with food intake. 49.3% of patients with GI disease and 51.2% with GERD thought that their chest pain would be of cardiac origin.

For GI disease pain was localised retrosternal in 71.8% (60.5%-80.9%), epigastric in 29.6% (20.2%-41.0%), left in 32.45 (22.7%-43.9%) and right in 23.9% (15.55-35.0%). For GERD retrosternal location was 83.3% (69.4%-91.7%) and epigastric location was 19.2% (10.0%-33.3%).

The majority of patients reported a pain frequency of more than one episode per day (58.2% (46.3%-69.3%) for GI disease, 48.8% (34.3%-63.5%) for GERD). Pain character was perceived by most patients as either pressing (40.8% (30.2%-52.5%) for GI disease, 35.7% (23.0%-50.8%) for GERD) or burning (29.6% (20.2%-41.0%) for GI disease, 30.9% (19.1%-46.0%) for GERD). Pain characteristics are listed in figure 2.

71.8% of patients with GI disease and 71.4% of patients with GERD still complained of chest pain when contacted after 6 weeks by telephone. After a follow up period of 6 months these numbers were reduced to 47.9% and 52.4% respectively. 74.6% of patients with GI disease and 85.7% of patients with GERD continued to consult their GP during the follow up period related to the initial problem of chest pain.

15 items (listed in the footnote of table 2) fulfilled our univariate selection criteria for GI disease and GERD and were selected for multivariable analysis the results of which are reported in table 2.

Pain worse with food intake and retrosternal pain radiation were associated positively with both GI disease and GERD. Retrosternal pain localisation, vomiting, burning pain, epigastric pain and an average pain episode < 1 hour were associated positively with GI disease. Negative associations were found for localized muscle tension (GI disease and GERD) and pain depending on exercise, pain worse with breathing or movement, pain on the left side (GI disease).

In table 3 we present for clinical use diagnostic items that were significant in the multivariable analysis with corresponding likelihood ratios (LR). These can help to rule in or to rule out GI disease and GERD 9.

In conclusion, GI disease including GERD is the fifth most frequent cause of chest pain in primary care. This study provides data on the diagnostic accuracy of selected signs and symptoms which will help primary care practitioners to narrow the likelihood of GI disease.

JRS acts as scientific advisor for MSD and ESSEX. All other authors do not declare any competing interests.

NDB formulated the research question, designed the study and supervised its conduct together with ACS. NDB, EB, JH, AB, ACS, MAH, HK, JRS, KK and SB were involved in acquisition, analysis and interpretation of data (MAH, HK in data acquisition, NDB, JRS, ACS, KK, SB in interpretation of data of the reference panel, NDB, EB, AB, JH, SB in data analysis). SB drafted the article. All authors read and approved the final manuscript.