The categorisation of different causes of death in patients with chronic obstructive pulmonary disease (COPD) has not usually been regarded as an important topic, but with all-cause mortality and cause-specific mortality now being used as outcome measures in large multicentre clinical trials,^1,2 it is perhaps time to take stock of what we do and do not know about what patients with COPD actually die from. A number of studies that have addressed this issue over the years have, not surprisingly, found varying proportions of deaths ascribed to respiratory causes, lung cancer and cardiovascular disease (the three principal categories), with the results of any one study being highly dependent on both the source (and accuracy) of patient information and on the severity of underlying disease.^3–7

The past year has witnessed a flurry of papers and editorials covering a number of widely different aspects of mortality in COPD, with topics ranging from the confidence we can have in interpreting mortality data,⁸ the possible role of inhaled corticosteroids on cardiovascular mortality in COPD^9,10 and the relationship of inpatient mortality to hospital resources and staffing levels¹¹ to whether or not sex influences survival.^12,13 Over the past few years there has also been a growing realisation that we need to rethink the traditional outcome measures (especially those based on conventional lung function) in clinical trials of COPD and—perhaps learning from the experience of our cardiology colleagues in their clinical studies in cardiovascular disease—accept the need for large prospective trials based on hard clinical outcomes such as death.¹⁴ The TORCH (Towards a Revolution in COPD Health) trial¹ was the first major long-term study in COPD that took all-cause mortality as its primary outcome measure, and in this issue of Thorax McGarvey et al¹⁵ report on the activity of the Clinical Endpoint Committee (CEC) that was charged with categorising the cause of death and its relationship to COPD in patients who died during the course of the study (see p 411).

The importance of this paper is that it describes, for the first time, the methodology which members of the CEC used in adjudicating on specific causes of death, and the members of the committee are to be congratulated on reaching an agreed consensus in every single case. The reliability of these categorisations was assessed by blindly re-adjudicating 11% of the cases: identical categorisation was found in 83%. One of the issues which raised particular problems for the committee, and which was the cause of several of these discordances on re-adjudication, was the difficulty in distinguishing between COPD exacerbation and pneumonia as causes of death. As a result of this, one of the helpful recommendations to emerge from the paper is that future studies of COPD mortality should re-examine how best to classify COPD exacerbations that occur in the setting of pneumonia.

Another problem highlighted by the authors was the difficulty in defining cardiovascular deaths: whether “sudden death” should always be regarded as cardiac in origin and whether some true cardiovascular deaths might have been misattributed to respiratory causes. This is not just of academic interest, but is particularly relevant in the light of our developing understanding of the complex interrelationships between cardiovascular and COPD mortalities; the possible beneficial role of inhaled steroids on cardiovascular mortality in COPD (already alluded to earlier),¹⁰ the relationship between reduced forced expiratory volume in 1 s and death from ischaemic heart disease,^16,17 the role of systemic inflammatory mediators (such as C-reactive protein) in increasing the risk of cardiac death in patients with COPD¹⁸ and, more recently, the suggestion that statin usage might be associated with reduced mortality in COPD.¹⁹ Never has it been more relevant to try and obtain a clearer picture of precisely what our patients with COPD are actually dying from.

The authors also describe in great detail how they attempted to differentiate between deaths that could definitely be attributed to COPD regardless of the specific final fatal event, and deaths that were only “related” to COPD, defined as illnesses which would probably not have been fatal had COPD not been present. Using these definitions, 27% of all deaths were ascribed as being directly due to COPD and, overall, 40% of the deaths were judged to be definitely or probably related to COPD.

Agreeing a consistent approach to classifying the cause of death is not just a question of semantics. Different strategies to decrease mortality in COPD will depend crucially on which particular putative causation is being targeted: exacerbation frequency, development of pneumonia or prevention/control of cardiovascular comorbidities. In emphasising how important it is not to confuse the cause of death (cardiac vs non-cardiac vs all-cause) in cardiovascular studies, Lauer et al²⁰ quoted Miss Buttercup from Gilbert and Sullivan’s HMS Pinafore: “Things are seldom what they seem; skim milk masquerades as cream”. In reporting mortality statistics in any future large long-term studies of COPD (and perhaps these should now certainly include a prospective trial of statins!), perhaps the more appropriate literary quotation to bear in mind when defining and classifying the causes of death in COPD is that from Humpty Dumpty in Lewis Carroll’s Alice Through the Looking Glass: “When I use a word, it means just what I choose it to mean—neither more nor less.”.