Clinical Relevance of Diagnosing COPD by Fixed Ratio or Lower Limit of Normal: A Systematic Review

Abstract

Background: Different spirometric criteria in diagnosing COPD have been advocated by different groups, debilitating adequate diagnosis and treatment of COPD. We reviewed the clinical relevance of fixed ratio and lower limit of normal (LLN) in diagnosing COPD and explored if modifying factors may affect their clinical relevance. Methods: Two reviewers independently searched PubMed and Embase for papers that compared both criteria on any clinically relevant outcome, published before June 1st, 2012, without any language restriction. Two reviewers independently extracted the study characteristics, including study design, population characteristics and diagnostic criteria used, and summarized the results of clinical relevance. Study quality was assessed by scoring forms for bias and level of evidence. Results: Of 394 studies retrieved, 11 studies were included, with a median of 1,258 participants. Although both criteria appeared related with various clinically relevant outcomes, we were unable to prefer one criterion over the other, with various performances of the criteria for different outcomes. Should the criteria disagree on diagnosis, an alternative diagnosis should be suspected, in particular in those (elderly) with less severe airflow limitation for whom the LLN appears a better criterion. The fixed ratio appears to perform better in subjects with more severe airflow limitation. Conclusion: In diagnosing COPD, severity of airflow limitation appears an important factor for choosing whether the fixed ratio or LLN. Disagreement between the criteria is suggestive for an alternative diagnosis. Future studies on clinical relevance should further reveal the criterion of choice, in order to improve adequate diagnosis and consequent treatments.

Keywords :

• COPD
• fixed ratio
• lower limit of normal
• spirometry

Introduction

Chronic obstructive pulmonary disease (COPD) is one of the leading chronic diseases worldwide with a major impact on patients’ health, health service use, and mortality (1). Improving the health and management of patients with COPD remains challenging. Recently, phenotype classifications, disease severity scores, and multidimensional prognostic indices have been developed to reveal clinically relevant outcomes and guide adequate care in those diagnosed with COPD (2, 3). However, misdiagnosis of COPD appears to be an overlooked factor that interferes with adequate care (4, 5). Depending on the diagnostic criteria used, prevalence of COPD has been reported up to 25% in individuals over 40 years, with underdiagnosis in more than half (6–11), whereas near fifty percent of those diagnosed by a physician appeared to be incorrectly diagnosed (5, 12). Misdiagnosis may result from both unawareness and inadequate performance or interpretation of spirometry, which is required to establish a diagnosis of COPD in those with chronic respiratory symptoms or those at risk (1, 13, 14).

Although numerous guidelines have raised awareness and championed the wider use and interpretation of spirometry over the last decades, they have also advocated different spirometric criteria for the diagnosis of COPD (15). Moreover, an adequate diagnosis not only depends on the clinician's skills, but also on the applied diagnostic criterion. Currently, most guidelines recommend to use the GOLD criteria to confirm airflow limitation, i.e. a fixed cut-off point of the forced expiratory ratio (forced expiratory volume in 1 second (FEV₁) / forced vital capacity (FVC)) <0.70 (1, 13, 14). However, this criterion may systematically overdiagnose in the elderly population and underdiagnose in the young (16, 17). An alternative, statistically derived, age- and gender-specific lower limit of normal (LLN) of the forced expiratory ratio is preferred by several investigators, as this may better reflect the physiologic properties of pulmonary function (16–18). Prevalence studies showed large discrepancy between the COPD rates based on these two criteria (5–9, 12). However, these studies do not reveal which criterion results in a more appropriate diagnosis and treatment. Moreover, the current debate on the appropriateness of these criteria remains unresolved and opinions between expert groups on which criterion to prefer are divided.

As an undisputable reference standard in diagnosing COPD is currently lacking, a reasonable alternative for the appropriateness of a diagnosis would be the clinical relevance of such a diagnosis. In this review, we aimed to systematically assess how the two spirometric criteria for the diagnosis of COPD, i.e. fixed ratio and LLN, are related with or predict patient-reported outcome, disease progression and/or acute events, in any population. In addition, we aimed to assess if these relations depend on any modifying factors.

Methods

We performed a structured search in both Pubmed and Embase databases to retrieve all papers comparing the clinical relevance for different diagnostic criteria for COPD. Following study selection we extracted their study data and systematically assessed the quality of the study methods.

Eligibility criteria

Inclusion of the studies in this review was based on the following criteria: (a) the study compared the fixed ratio and LLN of FEV₁/FVC as criteria to diagnose COPD, and (b) the study analyzed associations between these diagnostic criteria and any clinically relevant outcome in terms of patient-reported outcome: symptoms, health status, and co-morbidity; disease progression, i.e. FEV₁-decline; and acute events, i.e. exacerbations, hospital admissions, and mortality.

Search strategy and study selection

We conducted the structured searches until June 1st, 2012 (Supplementary data). Our searches did not have any restrictions, in particular no time or language restrictions. Two authors (WD and NG) independently first screened all articles on title and abstracts and next assessed the full-text for their eligibility. In addition, we screened the bibliographies of all included articles and of Hoesein et al., who previously reviewed articles on the fixed ratio and LLN (19). Differences between both reviewers were resolved by consensus or in case of disagreement by a third party (JB).

Data extraction

WD and NG extracted the following data from each included article: (a) study design, (b) population characteristics –origin, selection, number, mean age, mean sex, and mean FEV₁ percentage from predicted –(c) diagnostic criteria and reference equations used, (d) COPD prevalence according to different criteria, (e) clinically relevant outcomes studied and their relations with the diagnostic criteria, and (f) modifications of the relations by setting, airflow limitation as measured by FEV₁, sex, age, and smoking status and/or pack years. Disagreements were solved by consensus.

Assessment of methodological quality

WD and NG independently evaluated potential bias at the outcome level using a scoring form adapted from Hayden et al., which was developed for evaluating bias in prognostic studies and consisted of six systematically judged bias areas: score 1-5 per area and level 1 being the lowest risk of bias (Supplementary data)(20). The level of evidence was scored according to the assessment scheme for prognostic studies from the Oxford Centre for evidence-based medicine: score 1–5; level 1 being the highest level of evidence (Supplementary data)(21). The scoring agreement was assessed by both correlation analysis and by calculating the mean scoring difference and its limits of agreement (22). Disagreements were solved by consensus.

Data synthesis

A narrative synthesis was employed to describe the current knowledge on the clinical relevance by different diagnostic criteria in COPD. Data synthesis was based on a description of all included papers as extracted above, and a summary of their results. If applicable, we compared the study results of those with similar outcomes studied, irrespective of analytical methods used. We constructed a detailed summary table for an overview of study characteristics, results and study quality. Based on the individual study results, we present the preferred diagnostic criteria in a table according to the outcomes studied. In addition, we summarize how associations between the diagnostic criteria and clinically relevant outcomes are modified by any confounders.

Results

Study selection

After full-text screening, we finally included 11 articles in our review, out of 394 articles retrieved by our search (Figure 1, Table 1). Five articles were excluded only after full-text assessment, as they did not show comparisons between the different diagnostic criteria, with regards to clinically relevant outcomes (23–27).

Figure 1.  Flowchart of articles progressing through the selection criteria.

Table 1.  Characteristics of studies that analyzed the associations between diagnostic criteria and clinical relevance

Study reference	Year	Population	Design	Number	Male (%)	Age (mean)	FEV1%	Criteria ratio	Post-BD	LLN reference
Progression
Akkermans ^35	2012	Referred from family doctor to diagnostic centre, Netherlands	prospective 3.4 yr	3,324	45%	>40 (58)	80%	Fixed, LLN	Yes	a, b, c, d, e, f
Acute events
Guder^5	2012	COPD diagnosed by family doctor, Netherlands	prospective 4.2 yr	405	55%	>65 (72)	82%	Fixed, LLN, expert diagnosis	Yes	b, c, g
Lizak^36	2011	^$CHF, Poland	prospective	178	83%	n/a (52)	n/a	Fixed, LLN	n/a	n/a
Mannino (1)^31	2007	General, USA (CHS)	prospective <11yr	4,965	43%	>65 (n/a)	n/a	Fixed, LLN	No	f
Mannino (2)^28	2012	General, USA (NHANES III)	prospective 12–18 yr	13,847	48%	>25 (n/a)	n/a	Fixed, LLN*	No	f, from study
Patient-reported outcomes
Ko^37	2008	Social centres, China	cross-sectional/retrospective 1yr	1,008	27%	>60 (74)	87%	Fixed, LLN	Yes	h
Lamprecht^32	2011	General, Austria (BOLD)	cross-sectional	1,258	54%	>40 (n/a)	95%	Fixed, LLN	Yes	f
Wang^33	2011	General, China	cross-sectional	2,055	n/a	n/a	n/a	Fixed, LLN	n/a	i
Acute events and patient-reported outcomes
Cerveri^29	2008	General, Europe (ECRHS)	prospective 9 yr	6,249	47%	20-44 (n/a)	106%	Fixed, LLN	No	from study
Garcia-Rio^30	2011	General, Spain (EPI-SCAN)	cross-sectional/retrospective 1yr	885	58%	40-80 (58)	101%	Fixed, LLN	Yes	j
Izquierdo^34	2012	Matched COPD(GOLD), prim. care, Spain (CONSISTE)	retrospective 1 yr	1,248	72%	>40 (63)	58%	Fixed, LLN	Yes	B

ECRHS: European Community Respiratory Health Survey; EPI-SCAN: Epidemiologic Study of COPD in Spain; CONSISTE: Consecuencias Sistémicas de la EPOC; NHANES: National Health and Nutrition Examination Survey; CHS: cardiovascular health study; BOLD: Burden of Obstructive Lung Disease. ^$CHF: cardiac heart failure, on list for transplantation; n/a not available; *LLN = FEV₁/FVC <LLN and FEV₁ <LLN.

^a Swanney MP, et al. Thorax 2008; 63(12):1046–1051; ^b Quanjer PH, et al. Eur Respir J Suppl 1993; 16:5–40; ^c Falaschetti E, et al. Eur Respir J 2004; 23(3):456–463; ^d Kuster SP, et al. Eur Respir J 2008; 31(4):860–868; ^e Brandli O, et al. Thorax 2000; 55(2):173–174; ^f Hankinson JL, et al. Chest 2010; 137(1):138–145; ^g Enright PL, et al. Chest 2006; 129(2):384–392; ^h Ip, et al. Chest 2006; 129(2):384–392; ⁱ Cerveri I, et al. Thorax 2008; 63(12):1040–1045; ^j Roca J, et al. Bull Eur Physiopathol Respir 1986; 22(3):217–224.

Study characteristics

Table 1 summarizes the study characteristics of all included articles: One study reported progression as its outcome, four studies acute events, three studies patient-reported outcomes, and three studies both acute events and patient-reported outcomes. The mean number of participants was 1,258 (interquartile range 885 to 4,965), with a weighed 48% of males (range: 27% to 83%), and of those reported a mean age of 62 years (range 58 to 74), and a mean FEV₁ of 93% from predicted (range 58% to 106%). Six studies reported the use of post-bronchodilator spirometry values. All studies compared the fixed ratio to LLN, but one study also included an additional cut-off point for the FEV₁ as percentage from predicted for both criteria (28). Six studies recruited patients from the general population (28–33), whereas three studies specifically recruited patients based on (risk for) COPD diagnosis (5, 34, 35), one study recruited chronic heart failure patients (36), and one study recruited from social gatherings at community centres (37).

Prevalence of COPD varied according to the diagnostic criteria and population of interest, ranging from 4.3% to 78%; COPD prevalence was higher according to LLN only in the single study on young participants (29); overall and discordant diagnoses were more prevalent amongst selected and aged populations (Table 2). Table 2 also shows that all-cause mortality was studied four times; (respiratory and/or all-cause) hospitalizations five times; any hospital use twice; exacerbations twice; FEV₁–decline once. Patient-reported outcomes included dyspnea and co-morbidity three times; cough, sputum, or a combination of respiratory symptoms twice; wheeze, dyspnea medication, activity of daily living, and two health-related quality-of-life questionnaires once.

Table 2.  COPD prevalence and clinical relevance according to the different diagnostic criteria

	Fixed%	LLN%	Discordant	Outcomes studied	Method & statistics	Results	Bias	CEBM
Progression
Akkermans	27.6%	15.9%	11.7% fixed	FEV1-decline	(B) Random coefficient regression	Discordant similar to non-obstructed	2	1b
Acute events
Guder	60.2%	35–43%	17–25%	All-cause death, exacerbation, resp. hospital	(A) Univariate survival	Death: expert diagnosis better; Exacerbation: similar;Hospital: fixed ratio similar to little worse	2	1b
Lizak	n/a	n/a	n/a	All-cause death, all-cause/cardiac hospital	(A) X^2, Anova, K. Meier	Hospital: LLN stronger association	5	4
Mannino (1)	42.1%	19.3%	22.8% fixed	All-cause death, “COPD” hospitalization	(C) Survival*	Death: FEV1>80%; discordant similar to non-obstructed.Death: FEV1<80%; discordant similar to obstructed.Hospital: discordant in between.	2	1b
Mannino (2)	8.7%	6.5%	3.2% both	All-cause death, (respiratory death)	(C) Survival^#*	Discordant in between (LLN+ discordant unstable results)	2	1b
Patient-reported outcomes
Ko	25.9%	12.4%	13.5% fixed	Cough, sputum, wheeze, dyspnoea	(A) Logistic regression	Fixed ratio associated with dyspnoea only. LLN associated with both wheeze and dyspnoea (larger odds ratio).	2	2b
Lamprecht	24.2%	15.8%	9.5% both	Cough, phlegm, dyspnoea.Comorbidity (prevalence)	(B) X^2, Wilcoxon^#	Cough and phlegm: discordant similar to non-obstructed.Dyspnoea: similar to obstructed.Heart disease: discordant similar to obstructed.	2	2c
Wang	7.6%	6.9	0.7% fixed	Respiratory symptoms	(B) n/a	Discordant are all asymptomatic elderly	4	4
Acute events and patient-reported outcomes
Cerveri	4.3%	9.4%	5.1% LLN	Respiratory ED visit and/or hospitalization,dyspnoea medication, chronic cough/phlegm	(B) Two level Poisson regression	Hospital use/symptoms: LLN+discordant similar to obstructed. Medication: LLN+discordant in between	3	1b^
Garcia-Rio	10.2%	5.5%	4.6% fixed	Exacerbations, any hospitalization , SGRQ, EQ-5D, LCADL, mMRC,.Comorbidity (prevalence)	(C) Anova*	SGRQ: discordant in betweenEQ-5D: discordant in between/similar to mild obstructed.Rest: discordant in between/similar to non- & (mild) obstructed.	2	2b
Izquierdo	78%	59%	19% fixed	ED visits, (exacerbation-) hospitalization.Comorbidity (prevalence)	(B) Anova	Discordant similar to obstructed	2	2b

Discordant subjects are defined as presence of airflow obstruction according to one criterion but absent according to the other. If applicable, they are compared to the obstructed and non-obstructed subjects in the results section.

n/a: not available; (A): GOLD+ group compared to <LLN group, (B): discordant group compared to concordant goup(s), (C): discordant group compared to concordant goup(s) and according to disease severity (FEV₁%); * adjustments for both sex and age; ^# many patients with pulmonary restriction, which are analysed in a separate group; SGRQ: St. George respiratory questionnaire; EQ-5D: EuroQoL 5 Dimensions; LCADL: London Chest Activity of Daily Living; mMRC: modified medical research council dyspnoea scale; ED: emergency department; Anova: analysis of variance; CEBM: level of evidence; ^Due to the design, less than 80% of participants at baseline could be assessed for their outcomes.

Methodology and quality

Three studies compared fixed ratio diagnosed subjects directly to LLN diagnosed subjects, five studies analyzed discordant subjects (i.e. presence of airflow limitation according to one criterion but absent according to the other) as a separate group, and three studies analyzed discordant subjects as a separate group including stratification by disease severity as measured by FEV₁ (Table 2). The studies analyzing the discordant subjects separately provided information on agreement between the criteria and whether those subjects resembled obstructed or non-obstructed subjects (Figure 2). Three studies compared the criteria by survival analysis, three studies by analysis of variance, by regression models in three studies, and by Chi-square in two studies (Table 2). Adjustments for both sex and age were performed only in three studies (28, 30, 31), whereas the group of subjects with restrictive lung function were separated only in two studies (28, 32). Acute events and progression were studied in prospective cohorts, whereas patient-reported outcomes were mostly studied in cross-sectional and/or retrospective cohorts, with the exception of one prospective cohort of young participants.

Figure 2.  Airflow limitation according to the fixed ratio (FR) — and lower limit of normal (LLN) —.

Five studies scored level 1b for evidence, three level 2b, and one level 2c. Two studies for which we had reports from conferences only scored level 4 (33, 36). There were no inter-scorer (WD and NG) disagreements. In addition, bias was mostly scored level 2. The two studies that were reported from conferences only scored a bias level 4 and 5. Correlation analysis for scoring mean bias at the decimal level revealed a Pearson's r of 0.97, whereas the mean difference of scoring bias was 0.16 (95% limits of agreement: -0.22 to 0.55). Two disagreements on rounding bias scoring were solved by consensus.

Performance of diagnostic criteria

Preference for any diagnostic criterion varied across outcomes (Table 3). According to the outcomes studied, the fixed ratio performed better than LLN and/or discordant subjects had similar outcome as the obstructed subjects for 5 outcomes. LLN performed better and/or discordant subjects were similar to non-obstructed subjects for 9 outcomes, including two outcomes from low-evidence studies. In addition to these outcomes, one study on mortality preferred both criteria depending on the level of airflow limitation (31).

Table 3.  Preference of diagnostic criteria according to different outcomes studied

	GOLD better(or discordant (fixed+) similar to obstructed)	In between(discordant (fixed+) between non- /obstructed)	LLN better (or discordant (fixed+)similar to non-obstructed)	No preference(no discordant analysisor statistics)
Progression
FEV1-decline	–	–	Akkermans	–
Mortality
All-cause death	Mannino(1)^(for FEV1 <80%)	Mannino(2)^(for FEV1 <80%)	Mannino(1)^(for FEV1 >80%)	Guder, [Lizak]
Health service use
(Any) hospitalization	Izquierdo	Mannino(1)	Garcia-Rio, [Lizak]	Guder
Exacerbation	–	Garcia-Rio	–	Guder
Any hospital use	Izquierdo	–	Cerveri	–
Patient-reported outcome
Respiratory symptoms	Lamprecht	Garcia-Rio	Cerveri, Lamprecht, Ko, [Wang]	–
Activity of daily living	–	–	Garcia-Rio	–
Dyspnoea medication	–	Cerveri	–	–
Health status	–	Garcia-Rio	–	–
Comorbidity	Lamprecht ^(CVD), Izquierdo	Garcia-Rio ^(CVD)	–	–

Each study represents a preference for that given outcome. Studies with CEBM #4 are between [ ]. Between ( ) are specific outcome restrictions. Discordant is based on fixed+ and LLN-, except for Cerveri (29): discordant based on fixed- and LLN+; Guder (5), Lizak (36) and Ko (37) did not separately analyze a discordant group; Mannino2 (28) and Lamprecht (32) separated out subjects with restrictive pulmonary function; Mannino2 (28), Garcio Rao (30) and Mannino1 (31) adjusted for both sex and age.

From studies that analyzed the discordant subjects as a separate group, the discordant subjects had worse outcome than non-obstructed subjects but better than obstructed patients for 7 outcomes. From studies that did not analyze discordant subjects as a separate group, no criterion was preferred for 3 outcomes. As mortality and hospitalizations where studied more frequently, we summarized their relation with the diagnostic criteria (supplementary data) to reveal the diversity of the related studies in terms of populations and analytical methods.

Modifying factors

The modifying effect of the severity of airflow limitation on the relation between diagnostic criteria and outcome was directly or indirectly analyzed in 3 studies (28, 30, 31). However, one (retrospective) study only stratified for airflow limitation in the obstructed subjects, not for discordant subjects, and hence does not reveal the effect of airflow limitation itself (30). The two other studies revealed that in terms of mortality, discordant elderly subjects with a FEV₁ >80% of predicted were similar to non-obstructed subjects, whereas discordant (both elderly and young) subjects with a FEV₁ <80% of predicted were about similar to obstructed subjects (28, 31). In other words, for (elderly) subjects with a FEV₁ >80% of predicted LLN appears a better criterion to predict mortality, whereas for subjects with a FEV₁ <80% the fixed ratio appears more appropriate. The severity of airflow limitation did not affect the relation between the two criteria and hospitalization.

Both age and smoking had been studied for their modifying effect in 1 study only. Different age groups revealed similar relations between the criteria and mortality (28), whereas smoking did not affect the relation between the criteria and FEV₁-decline (35). Clinical setting and sex had not been studied for their modifying effect.

Discussion

Although a few articles have been previously discussed for their clinical relevance, this is the first article to systematically review the clinical relevance of diagnosing COPD according to both fixed ratio and LLN criteria (19). Based on 11 included studies, we were unable to decide on the preferred criterion, with various performances of the criteria for different clinically relevant outcomes. LLN tends to better reflect FEV₁-decline and respiratory symptoms, whereas fixed ratio might be better associated with co-morbidity. Studies analyzing the discordant subjects as a separate group offered more specific results on which criterion to prefer.

The relation between mortality and diagnostic criteria appeared to depend on the level of severity of airflow limi­tation, with a preference for LLN in those subjects with less severe airflow limitation and vice versa. Furthermore, the severity of airflow limitation may have affected other study results as well. For exacerbations and hospitalizations, discordant subjects seemed to have comparable outcomes to those subjects that had similar level of airflow limitation, either obstructed or non-obstructed subjects (30, 34). The one study that explored FEV₁-decline, revealed a very similar decline between discordant and non obstructed patients. The severity of airflow limitation of discordant patients, as measured by the baseline FEV₁, appeared to be intermediate from the airflow limitation of the non obstructed and obstructed patients (35).

In general, discordant subjects tended to have worse outcomes than non-obstructed subjects but better than obstructed patients. The association of discordant subjects with clinically relevant outcomes may not necessarily be due to COPD. Mannino et al. reported that discordant subjects with more severe airflow limitation had an increased risk for death, with the causes of death differing from that for obstructed patients (28). For those with less airflow limitation (FEV₁ >80% of predicted) the diagnosis of COPD has been questioned before in population based studies (38, 39).

Furthermore, the better association between fixed ratio and co-morbidity, including cardiovascular disease, may hence be due to misdiagnosis instead. Altogether, for discordant subjects (usually elderly males) an alternative diagnosis should thus be suspected, in particular for those elderly with less severe airflow limitation for whom the LLN appears a better predictor of mortality. This approach may result in a more adequate diagnosis and a better treatment of the disease causing the symptoms. In those with more severe airflow limitation, the fixed ratio may ensure a diagnosis and access to relevant treatments.

Strengths and limitations

The many studies retrieved in this systematic review provide a diverse mixture of populations and outcomes studied, which would ensure its translation at large. Other strengths of this study include the extensive and systematic search in two databases, the independent study retrieval and following data extraction and the appraisal of study quality by adequate scoring systems for both bias and level of evidence (20, 21).

However, several limitations apply. The variety of outcomes studied, the methodological heterogeneity and different study populations including varying degrees of airflow limitation resulted in different preferences and hampered any valid meta-analysis. In addition, as the fixed ratio introduces sex and age bias, one might argue that these should be accounted for in the analysis, depending on the outcomes studied (40, 41). Since the discordant subjects tend to be elderly males in particular, ignoring sex and age adjustments would increase its associations with outcomes like mortality and thus favor the fixed ratio.

Consequently, it may be questioned if the results of several studies may be less valid as these did not account for one or both of these two factors (5, 29, 32–37). However, restricted subjects were counted as non-obstructed subjects in most studies. As restricted subjects tend to have worse clinically relevant outcomes than non-obstructed subjects, failing to separate these subjects would reduce differences between non-obstructed and discordant subjects, disfavoring the fixed ratio (28, 32).

Conclusions and research directions

This systematic review shows that both the fixed ratio and LLN criteria are related with clinically relevant outcomes, but does not reveal a clear preference for one criterion. Limited information is available for factors that may modify the relation between diagnostic criteria and clinical relevance. In discordant subjects an alternative diagnosis should be suspected, in particularly in those with less severe airflow limitation. The LLN may prevent misdiagnosis and overtreatment in these subjects, whereas the fixed ratio may ensure access to relevant treatment in those with more severe airflow limitation.

To prevent misdiagnosis and improve adequate treatment, future studies regarding the clinical relevance of the diagnostic criteria are necessary to confirm which criterion to favor, preferably by separately analyzing the discordant subjects. Those studies should include analyses for the modifying effect of the severity of airflow limitation in particular, but also sex, smoking and age. Another aim would be to compare the clinical relevance of these diagnostic criteria to the diagnosis from physicians, who base their diagnosis also on symptoms and risk factors.

Declaration of Interests Statement

Van Dijk received a university grant (RvB08.066.51196/GE) by Radboud University Nijmegen Medical Center, Nijmegen, the Netherlands; Tan attended medical advisory board meetings and provided continuing medical education for GlaxoSmithKline, and received funding for research from unrestricted educational grants from GlaxoSmithKline, Pfizer, Boehringer Ingelheim, AstraZeneca; Bourbeau received research funding via the Research Institute of the McGill University Health Centre, from AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Merck, Novartis, Nycomed, Pfizer and Thera technologies; and has served on speakers, consultation panels and/or advisory boards for the above listed pharmaceutical companies; Gupta does not have any conflict of interest. There are no sources of funding related to this paper to be disclosed.

The authors alone are responsible for the content and writing of the paper as follows: Van Dijk: Study initiation, concept and design, literature search, data collection and interpretation, drafting manuscript, and guarantor of manuscript content, taking responsibility for the integrity of the work performed; Gupta: Literature search, data collection and interpretation, revision of manuscript; Tan: Study concept and design, data interpretation, and revision of manuscript; Bourbeau: Study initiation, concept and design, data interpretation, and revision of manuscript.

Acknowledgements

We wish to thank Danielle Bastien for critically reviewing the text for English writing.