Thursday 9 December 2010
Diagnostic strategies for chronic obstructive pulmonary disease
Promotor: Prof.dr K.G.M. Moons, prof.dr Th.J.M. Verheij, prof.dr A.W. Hoes
Defence: 9 December 2010
Earlier studies suggest that primary care patients with complaints of persistent cough frequently have underlying, unrecognised chronic obstructive pulmonary disease (COPD). Adequate detection of COPD in these patients is highly warranted, because targeted interventions (notably smoking cessation programmes) exert prognostic benefit by reducing the incidence of exacerbations and improving quality of life Nevertheless, much is unknown about the diagnostic value of tests that are usually included in the diagnostic workup in suspected COPD in primary care, including symptoms and signs. Whether and to what extent COPD can correctly be excluded or recognised by history taking and physical examination is unknown, as is the added diagnostic value of spirometry conducted by general practitioners (GPs). The independent value of other potentially useful diagnostic tests, such as level of C-reactive protein and reversibility testing after an oral corticosteroid test, has also not been studied. Persistent cough is one of the most frequently presented complaints in primary care, further underlining the need for evidence-based and efficient diagnostic strategies in these patients. Therefore we studied the diagnostic value of different (combinations of) diagnostic tests in middle aged and elderly patients suspected of COPD, i.e., those consulting their GP because of persistent cough and aimed to develop a diagnostic strategy applicable to these patients in daily practice.
In chapter two, a systematic search in electronic medical databases was reported, with the aim to review the existing evidence in the literature on the diagnostic value of items from history taking and physical examination in patients suspected of COPD. Studies were included after using predefined in- and exclusion criteria and their methodological quality was assessed using the Quadas (Quality Assessment of Diagnostic Accuracy Studies) criteria. Only six studies could be included of the 45 articles that were selected based on title and abstract. The history items dyspnoea, wheezing, previous consultation for wheezing or cough, age and smoking, and the physical examination items wheezing, forced expiratory time, laryngeal height and prolonged expiration were reported to have be of diagnostic value. None of the items were studied in more than three studies and study populations were very heterogenic. The reference test for COPD in five of the six studies concerned the ‘old definition’ of obstructive lung disease, and not COPD according to current “GOLD” guidelines. We concluded that there was insufficient evidence in the published literature to quantify the true value of history taking and physical examination in suspected COPD.
In chapter three, a large prospective diagnostic study in Dutch primary care, performed between 2006 and 2009 (FRESCO study: From REspiratory Studies to COPD) is presented in detail. In the FRESCO study, 400 subjects, older than 50 years who consulted their GP with persistent cough were included. All participants underwent an extensive diagnostic work-up, including standardised history taking and physical examination, as well as secondary care lung function tests. An expert panel determined the presence or absence of COPD and/or asthma (reference test). If participants had physician diagnosed asthma before entry in the study, or insufficient spirometry results, they were excluded. This resulted in 353 participants for the analysis on the prevalence of COPD and asthma reported in chapter three. 102 participants (29 %) were diagnosed with COPD, of which 14 had both COPD and asthma. The severity of COPD was GOLD stage I (mild) in 71 subjects, stage II in 29 (moderate), and stage III (severe) in 3 subjects. Asthma (without COPD) was diagnosed in 23 (7 %) participants. We concluded that in patients over 50 years old who consult their GP for persistent cough, previously undetected COPD or asthma is frequently present.
In Chapter four a study quantifying the diagnostic value of patient history taking and physical examination, and the added value of spirometry performed by the GP, and C-reactive protein (CRP) levels in primary care patients presenting with persistent cough was reported. For this analysis, the data of all 400 participants of the FRESCO study (described in chapter three) were included, of which 118 (30 %) had COPD according to the expert panel. History taking and physical examination items with independent diagnostic value were increasing age, male gender, current smoking, > 20 pack years of smoking, a history of cardiovascular disease, wheezing complaints, diminished breath sounds and wheezing on auscultation. A multivariable model combining these items had an area under the receiver operating characteristic curve (ROC area) of 0.79 (95% confidence interval (95% CI) 0.74 to 0.83). We developed and internally validated a simplified scoring rule for COPD including these history and physical examination items. In patients with a low score (n = 202, 51 % of total), COPD prevalence was 8 % (negative predictive value 92 %). Adding the post bronchodilator ratio of the forced expiratory volume in one second and the forced vital capacity (FEV1/FVC ratio) obtained by the GPs’ spirometry (cut off value < 0.7) increased the ROC area to 0.86 (95% CI 0.83 to 0.90) and substantially improved diagnostic risk classification. CRP levels had no relevant added value, however. We concluded that history taking and physical examination are useful for exclusion of COPD in patients presenting persistent cough in primary care, and that FEV1/FVC ratio adds substantial diagnostic value in contrast to CRP.
Chapter five addressed the question whether the previously developed model for COPD including symptoms and signs had added value above the physician’s own clinical assessment. This analysis was performed in 357 participants of the FRESCO study, of whom 104 (29 %) had COPD. The GPs estimated their (subjective) probability that COPD was present in a patient, after history taking and physical examination, without knowledge on results of spirometry or the diagnostic model. The added diagnostic value of the previously developed diagnostic model beyond the physicians’ estimation was quantified by the increase of the ROC area, and the improvement in classification across clinically relevant diagnostic risk categories: < 20 % (low probability of COPD) and ≥ 20 % (possible COPD). Adding the model’s estimated probability of COPD to the GP probability estimate increased the ROC area from 0.75 (95% CI 0.70 to 0.81) to 0.84 (95% CI 0.80 to 0.89)). Reclassification results implied that in 35 more patients (184 instead of 219), COPD could correctly be excluded without performing spirometry, and that less COPD cases were missed (13 instead of 19). We concluded that a diagnostic model based on signs and symptoms has additional diagnostic value beyond the physicians’ own clinical assessment in patients suspected of COPD.
In Chapter six the diagnostic value of a test treatment with oral prednisolone for exclusion or diagnosing COPD was studied, in 233 participants of the FRESCO study, of whom 75 had COPD (30 %). All subjects used a 14 day prednisolone test of 30 milligram per day including before and after measurement of the post bronchodilator FEV1. A person was defined a ‘responder’ to the prednisolone test if the increase of FEV1 exceeded 200 ml or 12 % of the baseline value. In participants with COPD, the proportion of responders was 20 % (n = 15), and in those without COPD 10 % (n = 15) (p value 0.03). Being a responder was univariably associated with COPD (OR 2.4 (95% CI 1.1 to 5.2)). After adjustment for age, gender and current smoking the multivariable OR was 2.0 (95% CI 0.8 to 5.0) and the prednisolone test did not increase the ROC area (from 0.78 (95%CI 0.72 to 0.85) to 0.79 (95%CI 0.72 to 0.85)). We concluded that in primary care patients suspected of COPD a response to a prednisolone test is suggestive of COPD, but has no value in addition to more easily obtainable diagnostic information.
Chapter seven included a cost-effectiveness analysis comparing different diagnostic strategies for COPD in primary care patients presenting persistent cough, using Markov modelling and Monte Carlo simulation. Three diagnostic strategies were compared: 1. history and physical examination ; 2. history, physical examination and the FEV1/FVC ratio and 3. history, physical examination and hospital lung function tests. All three strategies resulted in similar expected health benefits (QALYs), whereas the expected additional costs (per patient) of strategy 2, and 3 compared with strategy 1, were € 70 and € 264 respectively. Therefore, strategy 1, which is the most basic diagnostic strategy, was the preferred strategy in our model. This was demonstrated in all scenarios that were analysed, indicating the robustness of this finding. Also, strategy 1 has the largest probability of being cost-effective independent of the cost-effectiveness threshold applied. However, for commonly applied cost-effectiveness threshold values in the range of € 20,000 - € 100,000 per QALY strategies 2 and 3 also have a substantial probability of having the most favourable cost-effectiveness. We concluded that more extensive diagnostic strategies for COPD, which may have higher diagnostic accuracy (hence, lead to less misclassification), do not necessarily increase long term health benefits but will increase costs.
In chapter eight we discussed implications for clinical practice and research, in the view of our main findings. Our developed diagnostic model for COPD deserves further validation, in which consultation of practitioners is essential, before it could be used to safely exclude COPD based on symptoms and signs only. However it could already be helpful to raise awareness of possible COPD in patients with persistent cough. To enhance implementation of our model in daily medical practice would require integration in the electronic patient record.
Diagnosing early COPD has advantages and disadvantages. On the one hand, it is beneficial because it enhances (more) appropriate treatment of patients, and could help smokers to quit. On the other hand, there is still no clear evidence on the effects on prognosis, of an early diagnosis and treatment of mild COPD, except for smoking cessation. Studies in primary care patients with mild COPD could provide useful answers.
Next, to perform spirometry and to interpret the results is not straightforward. There is currently large debate on the appropriate thresholds values to define presence or absence of COPD as the normal values of spirometry results vary with gender and age. Therefore continuing education of GPs or other health care workers who perform spirometry will remain needed. It also underlines the need to interpret the results of spirometry only in combination with symptoms and signs.