A case-control study of presentations in general practice before diagnosis of coeliac disease

Participants

Eligible cases had biopsy-proven coeliac disease and fulfiled the updated European Society of Paediatric Gastroenterology and Nutrition (ESPGAN) criteria,20 were diagnosed after 1 January 1998, aged over 18 years at time of diagnosis, and registered with a general practice in south-east Wales. The 1 January 1998 was used as a cut-off date for diagnosis to increase the chances of completeness of GP clinical records studied during the 5 years prior to this date. Cases were identified at the coeliac clinic at Llandough Hospital, Cardiff. This is the main secondary care centre for coeliac disease for Cardiff and the Vale of Glamorgan and serves a total adult population of approximately 340 000. Referrals to this clinic arise from both primary and secondary care.

Five potential controls were matched to each case by sex, general practice, and date of birth (within 1-year bands) with the aim of recruiting two controls per case. Exclusion criteria for both cases and controls were unobtainable GP clinical records.

Cases were recruited following a formal diagnosis of coeliac disease and prior to a routine consultation at the hospital coeliac clinic. Once consent was obtained, the patients' GPs were contacted to obtain access to their general practice clinical records, and to enlist the GPs' help in recruiting controls. Potential controls were contacted via a letter from the GP, with an attached information sheet, and a consent form which would allow the research team to extract relevant data from their clinical records. A form indicating non-consent was also included to estimate the proportion that declined to participate. After 2 weeks if less than two control patients per case had consented, the practice was approached once more to identify five additional potential controls.

Obtaining and extracting clinical data

Photocopies were made of the general practice clinical records for the consenting case and corresponding controls for the 5 years prior to the date in which coeliac disease was first mentioned in the clinical records of cases. Computer and paper clinical records were copied along with the summary sheet. A research nurse anonymised the clinical records and recorded the number of referrals to secondary care consultants and clinical investigations in primary care for each included year. The clinical records were anonymised using a coding system that was known only to the researcher concerned. Missing clinical records for any period during the 5 years were documented.

Two GPs not involved in obtaining copies of the clinical records and blinded to case-control status extracted and recorded information on to a pro-forma, which listed possible common features of the presentation and investigation of coeliac disease. These features were derived from searching the literature and from expert opinion. The overall number of consultations per annum, as well as the number of times that the specified symptoms and diagnoses presented, were recorded. Comorbidities that could account for a higher than average consultation rate were also noted. Data extraction was protocol driven, with explicit and detailed inclusion and exclusion criteria for the consultations and presenting features.

Data extraction was piloted by the two GPs who each extracted data on the same set of 20 patient clinical records. As a result of this pilot phase, minor changes were made to the data extraction protocol, for example clarification of the inclusion/exclusion criteria and inclusion of extra categories of presenting features.

Statistical analysis

Recruitment

A flow chart of the recruitment process is shown in Figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flow chart of recruitment.

Cases and controls

Two hundred and twenty-eight clinical records were analysed; 68 cases and 160 controls, an average of two controls per case. Cases and controls were well matched (Table 1). There was no notable or significant difference between cases and controls with respect to comorbidity.

The average consent rate among those invited to be a control was 43.4% (standard deviation = 23.1 %), ranging from 0 to 100% for each case. There was a higher consent rate from older controls (P = 0.001) and from practices in less deprived areas (P = 0.016).

Comparison of consultations

There was strong evidence of a higher rate of consultations in cases compared with controls during the 5 years prior to diagnosis (P = 0.001) (Table 2). These differences were also seen for the 1 year prior to diagnosis and excluding the 1 year prior to diagnosis (P<0.001 and P = 0.008 respectively).

Identification of predictors of coeliac disease

In total, 26 variables were considered for univariate conditional logistic regression. Eight patient symptoms and diagnoses, referral by GP and five types of investigations were significantly associated at the 10% level, with a subsequent diagnosis of coeliac disease during the 5 years prior to diagnosis (Table 3). An OR of greater than one implies a higher probability in cases and a 95% CI including the value 1 indicates no statistically significant difference between cases and controls. Thus, all associated variables were more common in those with a subsequent diagnosis of coeliac disease. However, the same proportion of patients with coeliac disease were issued with sick certificates as controls (26 versus 24% respectively), and had a similar number of barium radiological investigations (7 versus 8% respectively).

These statistically significantly associated factors were subsequently entered in two separate multivariate models. The multivariate conditional logistic regression showed three patient symptoms and diagnoses, referral by GP, and three kinds of investigations independently associated with a subsequent diagnosis of coeliac disease (Table 4). Increased risk of being diagnosed with coeliac disease was associated with patients presenting with depression and/or anxiety, diarrhoea, or being diagnosed with anaemia during the preceding 5 years. Coeliac disease was also associated with being referred at least once during the 5-year period, not necessarily to a gastroenterologist. Haematology or immunology blood tests were also associated with having coeliac disease, as was referral for any kind of endoscopy.

Of the 68 cases, 54 (79%), had presented with either diarrhoea, depression and/or anxiety, or had been diagnosed with anaemia, compared with 71 (44%), of the controls. Similarly, 67 (99%) of cases had at least one of the investigations listed in Table 4 or had been referred, compared with 128 (80%) of the controls.

Thirty-five (51%) cases had presented with diarrhoea, 28 (41%) depression and/or anxiety, and 23 (34%) had been diagnosed with anaemia at least once during the 5 years prior to diagnosis. This compares with 31 (19%), 51 (32%) and seven (4%), respectively of the controls. Similarly, 14 (21 %) cases had presented with diarrhoea and depression and/or anxiety compared with 14 (9%) controls, eight (12%) cases presented with depression and/or anxiety and had been diagnosed with anaemia compared with four (3%) controls, 13 (19%) cases had presented with diarrhoea and been diagnosed with anaemia compared with three (2%) controls. Three (4%) cases had all three features compared to three (2%) controls.

Analysis excluding 1 year prior to date of diagnosis

Multivariate analysis using data excluding the 1 year prior to diagnosis was performed to identify early features predictive of a subsequent diagnosis of coeliac disease. Only diarrhoea, anaemia, and microbiology tests were independently associated (Table 4). Supplementary Figure 1 shows the timing of consultations with diarrhoea and anaemia, and the cumulative consultations for cases and controls.

Summary of main findings

This study has shown that patients subsequently diagnosed with coeliac disease have an increased number of consultations in general practice during the 5 years prior to diagnosis.

Despite exploring a large number of symptoms and diagnoses previously identified as potential predictors of a diagnosis of coeliac disease, it was found that only diarrhoea, depression and/or anxiety, or a diagnosis of anaemia were significantly independently associated with the subsequent diagnosis of coeliac disease. Diarrhoea and anaemia remained associated even after excluding data for the 1 year prior to diagnosis, indicating that depression and/or anxiety was more prevalent in cases during the year prior to diagnosis.

Patients subsequently diagnosed with coeliac disease were also referred more frequently to secondary care specialists and underwent more investigations in primary care during the 5 years prior to diagnosis compared with controls, but this difference was particularly marked for the 1 year prior to diagnosis. A higher proportion of microbiology investigations were found in cases when data for the year preceding diagnosis was excluded from the analysis, which may be related to the testing of stool specimens.

Strengths and the limitations of the study

Sixty-eight cases were successfully recruited and each case was matched with at least two controls, thus achieving the sample size calculation target. Restricting researcher access to the clinical records of only those patients who have provided written consent may lead to greater bias in contemporary case-control studies compared to studies conducted before this became standard practice. In this study, a high consent rate for sequentially approached cases in the secondary care centre resulted in a low risk of selection bias for case. The sample of cases in this study have comparable demographic characteristics to patients with coeliac disease in other studies.9^,22 However, unsurprisingly, a much lower proportion of the potential controls approached posted back signed consent forms, making selection bias among this group more possible.

There is some evidence of a higher rate of consent among older controls. People may also be less likely to consent to researchers accessing their clinical records if they frequently consult, and especially if they have consulted with sensitive problems.23 The average consultation rate per annum for the control group was 4.5, slightly less than the consultation rate for the adult Welsh population of 4.95.21 Given the sensitive nature of mental health conditions, the finding that presentations of depression could be predictive of a subsequent diagnosis of coeliac disease may also be due to selection bias in the control group, as controls with recent presentations with depression may have been less likely to consent.

Comparison with existing literature

The finding that diarrhoea and anaemia are the most important predictors of a subsequent diagnosis of coeliac disease are congruent with studies that examined current symptoms in patients newly diagnosed with coeliac disease through screening. Sanders et al identified 12 new cases among 1200 people they screened for coeliac disease, and found that the diagnosis was more prevalent among those who had iron deficiency anaemia.24 Another screening study found that anaemia was most commonly associated, while gastrointestinal symptoms, especially irritable bowel syndrome (IBS), were not associated with a diagnosis of coeliac disease.8 Neither of these studies found depression commonly present at the time of diagnosis through screening, despite depression being generally more common in coeliac disease.18

Dickey and McMillan found that diarrhoea was the most common symptom present in patients diagnosed in primary care with coeliac disease (34%), followed by iron deficiency without gastrointestinal symptoms (30%). IBS symptoms other than diarrhoea were present in only 6%.25

Implications for future research and clinical practice

Previous studies have suggested that a heightened index of suspicion is important for making the diagnosis, and stress the crucial role of GPs, especially among patients presenting with non-gastrointestinal symptoms.26^,27 Indeed, there has been an increase from 28 to 60% in the proportion of patients with coeliac disease where the diagnosis was made in primary care.25

The findings suggest that GPs should consider testing for coeliac disease when patients present often, especially with diarrhoea, and/or who are discovered to be anaemic, as this could lead to an earlier diagnosis of an important and easily treatable condition, thus relieving considerable morbidity. Further research is required to clarify the role of depression in the diagnosis of coeliac disease.