Introduction
Chronic obstructive pulmonary disease (COPD) increases the risk of cardiopulmonary events, a composite term encompassing both cardiovascular (CV) and respiratory events.1–3 Cardiopulmonary risk has recently been defined as: the risk of serious respiratory and/or CV events in patients with COPD, including, but not limited to, COPD exacerbations, myocardial infarction, stroke, heart failure decompensation, arrhythmia and death due to any of these events.1,3 Increasing awareness of cardiopulmonary events experienced by patients with COPD is important since the cause of death among individuals with COPD is often due to both CV and respiratory causes.1 In fact, approximately one in five patients with COPD die from CV disease (CVD),4 an occurrence not limited to those with severe disease,5 and there is a 90% increase in mortality risk among those with both conditions versus COPD alone.6 Patients with COPD who have a history of CVD also have a higher risk of hospitalisation (for COPD and CVD) and an increased risk of frequent COPD exacerbations.7,8
The physiological link between acute COPD exacerbations and CV events is strong and may be bidirectional with the presence of one negatively impacting the other (eg CV events increase the risk of future exacerbations).9 COPD is an independent predictor for CVD, and may even be considered a distinct CV risk factor.10,11 More specifically, patients with COPD have 2–5 times the rate of CV comorbidities (eg hypertension, heart failure, cardiac dysrhythmia, ischaemic heart disease and arterial diseases) than those without COPD,12 and consistently have a higher CV risk than assessed using either QRISK3 or QRISK4.13,14 However, QRISK3 may underestimate CV risk in COPD, while QRISK4, despite including COPD, may not fully capture the acute and sustained risk associated with exacerbations.13–15 Pooled estimates from the EXACOS-CV multi-database studies also found a consistently increased and sustained risk of severe CV events following COPD exacerbations.15 Shared risk factors, including smoking, ageing, physical inactivity and socioeconomic disadvantage and the systemic effects of evolving COPD may contribute to CVD before COPD diagnosis, although whether events reflect unrecognised COPD or parallel disease remains uncertain.9,16–18 According to the Global Initiative for Obstructive Lung Disease (GOLD) 2026, multi-morbidity influences mortality and hospitalisations independently of the severity of airflow limitation and deserves specific treatment.19
Several studies have examined respiratory events, healthcare contacts and comorbidity burden before COPD diagnosis,20–23 but CV events during this period remain poorly characterised; most research on cardiopulmonary risk, including EXACOS-CV, has focused on established COPD or the period following a documented exacerbation.15,24–27 Although, guidelines recognise comorbidities, clear protocols for screening, risk assessment, and multidisciplinary management of cardiopulmonary risk in COPD are lacking.3 Our study addresses this knowledge gap in England, where COPD is suspected based on risk-factor exposure and persistent respiratory symptoms and confirmed by post-bronchodilator spirometry.28,29 Recurrent pre-diagnosis events or healthcare contacts may, therefore, provide opportunities for earlier assessment in appropriate patients. In contrast to previous research, we examined CV and respiratory events before the first recorded COPD diagnosis, characterising a less well-studied period in the patient journey and identifying potential opportunities for earlier clinical assessment. If cardiopulmonary events emerge months to years before formal diagnosis of COPD, there could be a need to prioritise systemic evaluation and proactive management of cardiopulmonary risk as early as the point of COPD diagnosis.
Addressing this gap can inform integrated pathways also improving early detection in at-risk patients.
Therefore, in the current study (ABACOS-RISK) we aimed to (i) quantify/estimate the prevalence and magnitude of cardiopulmonary events and risk among patients with COPD prior to diagnosis, and (ii) to compare these events, as well as healthcare resource utilisation (HCRU) attributed to them, to a matched cohort of patients without COPD in the UK.
Methods
Study Design
This was a retrospective, observational cohort study using secondary data of patients with COPD in England enrolled in the Clinical Practice Research Datalink (CPRD).30–32 The case identification period was from January 1, 2018, to December 31, 2019. Three cohorts were identified: overall COPD and incident COPD cohorts, and respective matched non-COPD controls. The overall cohort included patients with a history of COPD at any time during or before the case identification period.
The incident cohort included patients with no recorded COPD diagnosis before the case identification period (2018–2019), and the non-COPD matched controls included individuals without COPD matched in a 2:1 ratio to the COPD cohorts for age, sex and ethnicity. The index date was the first observed COPD diagnosis during the identification period for incident cohort or latter of January 1, 2018, or 1st observed COPD diagnosis for the overall cohort (Figure S1). The index date for the matched non-COPD cohort was set to their respective case.
The look-back period was a minimum of 12 months pre-index, extending to the full patient history to index date.
Data Source
Data were extracted from CPRD Aurum database (December 2023 build)31 with linkage to the Office of National Statistics (ONS) Index of Multiple Deprivation (IMD, 2019) and Hospital Episodes Statistics (HES) datasets. The CPRD Aurum database contains records on approximately 14 million patients registered in primary care in England (24% of the total English population).31 See Supplementary File 1 for additional details. Accurate identification of patients with COPD in the CPRD database has previously been confirmed.33 We focused on the UK since it is one of the few countries in the world where a patient journey can be followed through electronic health records (EHRs) from primary care (PC), secondary care, disease registries and death registries by using the patient’s unique patient identifier (NHS Number) particularly beneficial when conducting longitudinal research.34
Patients
All individuals included in the study were required to have a minimum of 12 months registration before the index date, be registered at a CPRD practice and aged ≥35 years on the index date and have acceptable research quality data (linkable to the HES admitted patient care and out-patient data sources). Patients with COPD were required to have a COPD diagnosis recorded prior to, or between, January 1, 2018, and Dec 31, 2019. COPD was identified using the ICD-10 codes for the HES dataset or using medcodes mapped to Snomed for CPRD Aurum.35,36 Individuals in the non-COPD control group were selected at random and matched without replacement.
Study Measures
Cardiopulmonary events of interest included both CV and respiratory events (Figure S2). CV events of interest included severe events such as diagnosis or hospitalisations for heart failure, myocardial infarction, cardiac arrest, emergency revascularisation, tachyarrhythmia, atrial fibrillation, and ischaemic stroke. Respiratory events include diagnosis or hospitalisation for COPD exacerbations among patients with COPD or lower respiratory tract infections (LRTIs) among incident COPD.
These events were identified using pre-specified medcodes in CPRD and ICD or OCS codes in HES, as appropriate (see Supplementary File 1 for further details). The primary outcome was the proportion of patients with ≥1 cardiopulmonary event in the overall and incident cohorts. Other outcomes included the proportion of patients with ≥1 healthcare contact, and associated costs for those with and without COPD.
HCRU variables included COPD-related PC consultations, outpatient visits, and inpatient admissions. PC visits were identified from CPRD Aurum consultation tables, excluding administrative contacts, outpatient appointments from the HES outpatient dataset, and inpatient admissions from the HES APC dataset. Average costs (British £) were assigned using the Unit Cost of Health and Social Care, and National Tariffs, as appropriate.37,38
Statistical Analyses
Baseline, prevalence and HCRU data were summarised descriptively: continuous variables using means and standard deviation (SD) or medians with interquartile ranges (IQR), and N (%) for categorical data. Missing values were quantified, not imputed, and variable-specific denominators were used. Cardiopulmonary events were quantified for all cohorts, overall and by event type (ie CV or respiratory) for the full look-back period in the incident cohort, and within 3 years pre-index among those with ≥3 years of pre-index data.
This provides an equivalent observation window for both cohorts and therefore standardises the opportunity for event ascertainment. Cardiopulmonary events were also described according to age, sex, region, IMD, and co-morbid asthma. The CV and respiratory components of the composite endpoint were not mutually exclusive; patients who experienced events in both categories were included in each component-specific analysis but counted only once in the overall cardiopulmonary composite.
The primary cohort of interest was the incident COPD cohort, with the larger overall COPD cohort analysed for further context.
Cardiopulmonary risk was compared between COPD and matched non-COPD cohorts using odds ratio (OR) and 95% confidence intervals (CI) overall and by event type, for all patients and those with at least 3 years of data. The primary analysis estimated real-world population-level differences in pre-diagnosis cardiopulmonary outcomes between the COPD and non-COPD groups. As such, models were adjusted for age, sex and other prespecified demographic characteristics.
Smoking, as a confounder and moderating factor, was not adjusted as the objective was to capture the overall population contrast, including differences in smoking exposure, rather than compare groups with equivalent smoking exposure. Among the incident cohort, a sensitivity analysis restricted the respiratory component of cardiopulmonary events to hospitalised LRTIs, which represent severe events, potentially indicative of exacerbations before a formal diagnosis of COPD. All analyses were descriptive, with no primary hypothesis testing related to treatment effectiveness between cohorts, so no adjustment for multiple comparisons or a priori sample size or power calculations were required.
R version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria) was used.
Ethics
CPRD has obtained ethical approval from the East Midlands – Derby Research Ethics Committee (21/EM/0265) for all standard observational research studies utilising anonymised data (see Supplementary File 1 for further details on CPRD). For an individual study, researchers are required to obtain CPRD Research Data Governance (RDG) approval to ensure that the study is in accordance with CPRD governance guidelines, the methods are appropriate and robust, and that the study objectives demonstrate a public health benefit. This study was conducted under CPRD RDG approval protocol #24_004479, which was reviewed and approved on 27 November 2024.
The approval allows analysis of anonymous electronic patient data without the need for written or oral consent. The administrative claims data analysis of our study was exempt from International Review Board review as it was a retrospective analysis of existing data (hence no patient intervention or interaction), and no patient-identifiable information was included in the claims dataset.
Individual informed consent was not required because the study used anonymised, routinely collected healthcare data; patients could opt out of research use of their data.
Results
Patients
Of 53,801,992 patients in the CPRD Aurum database, 34,272,170 had research-quality data eligible for HES/ONS linkage; and 427,044 and 77,461 met the eligibility criteria for inclusion in the overall and incident cohorts, respectively (Figure 1). The prevalence and odds of experiencing a pre-specified severe cardiopulmonary event was assessed in 426,854 patients in the overall cohort (matched to 853,708 non-COPD controls) and in 77,428 patients in the incident cohort (matched to 154,856 non-COPD controls) (Figure 1), of whom 70,034 (matched to 140,068 non-COPD controls) had data within 3 years pre-index comprising of the incident cohort with ≥3 years pre-index data. Figure 1Subject disposition.Flowchart sequence: From 53.8 million patients in CPRD Aurum, 427,044 met the eligibility criteria.
The final matched overall cohort comprised 426,854 patients with COPD and 853,708 controls without COPD (1:2 matching), including both previously diagnosed patients and those first diagnosed during 2018–2019. The incident cohort included 77,461 patients, of whom 77,428 were matched to 154,856 controls without COPD.Flowchart sequence: Patient selection from CPRD Aurum database to overall cohort and incident cohort with controls.Abbreviations: COPD, chronic obstructive pulmonary disease; CPRD, Clinical Practice Research Datalink; Dx, diagnosis; HES, hospital episode statistics; ONS, Office of National Statistics.
Baseline Demographic and Clinical Characteristics
The mean (SD) age of the incident COPD cohort and their matched non-COPD controls was 67.4 (13.6) years (Table 1). 52% of the groups were male, and the majority of patients identified as White (93.5% in both groups). After matching, patients with COPD were more likely to be current smokers (Tables 1 and S1).
Of those for whom a risk assessment was recommended, 16.1% had a documented QRISK2 score. A similar pattern was noted in the overall cohort (Table S2).Table 1Demographic and Baseline Characteristics in the Incident COPD Cohort and the Matched Non-COPD Control CohortDemographic and Patient CharacteristicsIncident COPD Cohort N=77,428Non-COPD Controls N=154,856SMDAge, yrs Mean (SD)67.4 (13.6)67.4 (13.6)0.00 Median (LQ-UQ)68 (58–77)68 (58–77)0.00 35-44, n (%)4563 (5.9)9121 (5.9) 45-54, n (%)10,049 (13.0)20,094 (13.0) 55-64, n (%)16,759 (21.6)33,514 (21.6) 65-74, n (%)20,992 (27.0)41,902 (27.1) 75-84, n (%)16,893 (21.8)33,837 (21.9) 85+, n (%)8,242 (10.6)16,388 (10.6)Sex0.00 Male, n (%)40,239 (52.0)80,478 (52.0)EthnicityN=77,003N=154,0060.00 White, n (%)72,015 (93.5)144,030 (93.5) Black, n (%)1264 (1.6)2528 (1.6) Asian, n (%)2597 (3.4)5194 (3.4) Mixed, n (%)460 (0.6)920 (0.6) Other, n (%)667 (0.9)1334 (0.9)Cholesterol Total mmol/L*, mean (SD)N=54,2944.7 (1.2)N=105,3194.8 (1.2)0.13BMI category**N=62,660N=98,6600.04 Mean, (SD)28.0 (6.9)28.3 (5.8)0.44 <18.5, underweight, n (%)2976 (4.7)1552 (1.6) 18.5–24.9, healthy weight, n (%)19,370 (30.9)27,384 (27.8) 25.0–29.9, overweight, n (%)19,492 (31.1)37,628 (38.1) 30.0–39.9, obese, n (%)17,340 (27.7)28,379 (28.8) 40+, severely obese, n (%)3482 (5.6)3717 (3.7)Weight, KgN=62,674N=98,5630.05 Mean (SD)78.5 (21.1)79.6 (18.5)QRISK† Recommended, NN=57,103 Documented, n (%)9213 (16.1) High risk, n (%)3690 (40.1) Medium risk, n (%)3090 (33.5) Low risk, n (%)2433 (26.4)SmokingN=77,026N=153,2630.76 Current, n (%)28,122 (36.5)18,703 (12.2) Ex-smoker, n (%)36,361 (47.2)67,120 (43.8) Never-smoked, n (%)12,543 (16.3)67,440 (44.0)AlcoholN=68,104N=132,3800.11 Current, n (%)57,905 (85.0)115,707 (87.4) Ex-drinker, n (%)5570 (8.2)7840 (5.9) Never drank, n (%)4629 (6.8)8833 (6.7)IMD quintilesN=77,374N=154,7440.15 I -least deprived, n (%)12,195 (15.8)30,217 (19.5) II, n (%)14,492 (18.7)32,474 (21.0) III, n (%)14,825 (19.2)30,448 (19.7) IV, n (%)16,786 (21.7)30,532 (19.7) V – most deprived, n(%)19,076 (24.6)31,073 (20.1)Notes: *the most recent value 3 years prior to the index date within the relevant time window was reported. **the most recent value within the relevant time window. †high risk: QRISK2 >20%; medium risk: QRISK2 10–20%; low risk: QRISK2: <10%.Abbreviations: BMI, body mass index; COPD, chronic obstructive pulmonary disease; HDL, high density lipoprotein; IMD, Index of multiple deprivation; LDL, low density lipoprotein; LQ, lower quartile; SD, standard deviation; SMD, standardised mean difference; UQ, upper quartile.
Prevalence of Cardiopulmonary Events
Incident COPD and Matched Cohorts
Of 70,034 patients in the incident cohort with ≥ 3 years of pre-index data, 39,612 (56.6%) experienced a cardiopulmonary event within 3 years pre-index compared to 20.9% of matched controls (n=29,266/140,068); 23.8% (n=16,679/70,034) experienced ≥1 severe CV event (vs 10.4% of controls [n=14,543/140,068]) and 46.1% (n=32,293/70,034) experienced ≥1 respiratory event (vs 14.0% of controls [n=19,583/140,068) (Figure 2). A similar pattern was noted over the full look-back period (n=77,428); 59,010 patients (76.2%) had ≥1 pre-specified cardiopulmonary event at any time before COPD diagnosis compared to 45.6% of matched controls (n=70,615/154,856), with 30.8% (n= 23,823/77,428) experiencing ≥1 severe CV event (vs 16.9% [n=26,156/154,856]) and 68.8% (n=53,242/77,428) experiencing ≥1 respiratory event (vs 37.8% [n=58,454/154,856]) (Figure 2). The mean look-back period was 15.0 (SD 7.6) years for the incident COPD cohort and 12.3 (SD 7.5) for their matched controls.
A consistently high prevalence of cardiopulmonary events was recorded (>70%), irrespective of sex, IMD, region, or prior asthma (Figure 3). Figure 2Primary outcome: prevalence of pre-specified cardiopulmonary events* in patients with COPD and matched controls over the full look-back period and within 3 years pre-index (incident cohort). *a composite measure comprising pre-specified cardiovascular events (ie heart failure, myocardial infarction, cardiac arrest, emergency revascularisation, tachyarrhythmia, atrial fibrillation and ischaemic stroke) and/or respiratory events (ie COPD exacerbations or LRTI). †over the full look-back period. ‡Within 3 years pre-index.Grouped bar chart comparing the percentage of patients with at least one cardiopulmonary, severe cardiovascular or respiratory event before COPD diagnosis. Each category shows patients with incident COPD and matched controls without COPD, for both the full available look-back period and the 3 years before diagnosis. - Cardiopulmonary events: 76.2% of patients with COPD experienced an event versus 45.6% of controls over the full look-back period; 56.6% versus 20.9% within 3 years before diagnosis - Severe cardiovascular events: 30.8% versus 16.9% experienced an event over the full look-back period; 23.8% versus 10.4% within 3 years. - Respiratory events: 68.8% versus 37.8% experienced an event over the full look-back period; 46.1% versus 14.0% within 3 years.A group bar graph showing the prevalence of cardiopulmonary, severe cardiovascular, and respiratory events prior to COPD diagnosis.Abbreviations: COPD, chronic obstructive pulmonary disease; CP, cardiopulmonary; CV, cardiovascular; LRTI, lower respiratory tract infection.Figure 3Prevalence of pre-specified cardiopulmonary events* in patients with COPD over the full look-back period by sex, age, IMD, region and prior asthma history (incident cohort; N=77,428). *a composite measure comprising cardiovascular events (ie heart failure, myocardial infarction, cardiac arrest, emergency revascularisation, tachyarrhythmia, atrial fibrillation and ischaemic stroke) and/or respiratory events (ie COPD exacerbations or LRTI).Horizontal bar chart showing the percentage of patients with COPD who experienced a cardiopulmonary event prior to COPD diagnosis, by subgroup: - Gender: Male, 73.6%; female, 79.1%. - Age: 35–60 years, 66.5%; 61–70 years, 75.1%; 71–78 years, 79.4%; ≥79 years, 87.9%. - Index of Multiple Deprivation: Group I (least deprived), 76.4%; group II, 74.8%; group III, 74.5%; group IV, 75.9%; group V (most deprived), 78.8%. - Region: North East, 78.7%; North West, 80.9%; Yorkshire and the Humber, 81.6%; East Midlands, 77.7%; West Midlands, 68.6%; East of England, 81.2%; London, 73.7%; South East, 78.2%; South West, 76.5%. - Prior asthma: Without asthma, 73.4%; with asthma, 83.0%.A horizontal bar graph showing cardiopulmonary event prevalence before COPD diagnosis by subgroup.Abbreviations: COPD, chronic obstructive pulmonary disease; IMD, index of multiple deprivation; LRTI, lower respiratory tract infection.
Of those who experienced a severe CV event within 3 years pre-index (N=16,679), the most frequent were atrial fibrillation (58.1%; n=9,698), heart failure (46.1%; n=7,682) and myocardial infarction (15.1%, n=2511; Figure 4A), with a similar distribution observed for the full look-back (Figure 4B and Table S3). Figure 4Prevalence of individual severe CV events in patients (A) within 3 years pre-index for those with ≥3 years pre-index date (n=70,062) and (B) over the full look-back period (N=77,461, Incident cohort). *All respiratory events were LRTIs.Denominators for all estimates are provided in Table S3.Two panels show the breakdown of severe cardiovascular events before COPD diagnosis. Panel A: Among 16,679 patients with at least one severe cardiovascular event, within 3 years before COPD diagnosis 58.1% experienced atrial fibrillation, 46.1% heart failure, 15.1% myocardial infarction, 13.5% tachyarrhythmia, 8.1% ischaemic stroke, 4.7% emergency revascularisation and 2.5% cardiac arrest.
Panel B: Among 23,840 patients with at least one severe cardiovascular event over the full look-back period before COPD diagnosis, 49.9% experienced atrial fibrillation, 41.6% heart failure, 28.0% myocardial infarction, 21.0% tachyarrhythmia, 12.7% ischaemic stroke, 9.7% emergency revascularisation and 3.1% cardiac arrest.Two-panel bar chart showing the types of severe cardiovascular events experienced before COPD diagnosis among 16,679 patients with such events within 3 years of diagnosis (panel A) and 23,840 patients over the full look-back period (panel B).Abbreviations: AF, atrial fibrillation; COPD, chronic obstructive pulmonary disease; CP, cardiopulmonary; CV, cardiovascular; Emerg Revasc, emergency revascularisation; LRTI, lower respiratory tract infection; MI myocardial infarction.
Overall COPD and Matched Cohorts
In the overall cohort (n=393,549), the prevalence of cardiopulmonary events was 56.2% (n=221,266/393,549) within 3 years pre-index and 82.5% (n=352,220/426,854) over the full look-back period (Table S4 and Figure S3). For the overall cohort 84,977 patients (21.6%) experienced ≥1 severe CV event and 195,375 (49.6%) experienced ≥1 respiratory event within 3 years pre-index, with corresponding prevalences of 31.6% (n=134,958/426,854) and 78.1% (n=333,334/426,854) over the full look-back. Of those who experienced a severe CV event within 3 years pre-index (n=84,977), the most common were atrial fibrillation (59.1%; n=50,275) and heart failure (43.6%; n=37,013), with similar prevalences noted for the full look-back (48.6% (n=65,533/134,958) and 40.5% (n=54,707/134,958), respectively; Table S4 and Figure S3).
Risk of Cardiopulmonary Events
Incident COPD and Matched Cohorts
Within 3 years pre-index, incident COPD cohort had higher odds of experiencing a cardiopulmonary event than matched controls (OR 4.9; 95% CI: 4.8, 5.0). Associations were evident for both severe CV events (OR 2.7; 95% CI: 2.6, 2.8) and respiratory events (OR 5.3; 95% CI: 5.2, 5.4) (Figure 5). The association was also observed over the full-look-back period (OR 3.8; 95% CI 3.8, 3.9) (Figure 5 and Table S3), again evident for both severe CV events (OR 2.2; 95% CI: 2.1, 2.2) and respiratory events (OR 3.6; 95% CI: 3.6, 3.7) (Figure 5).
Figure 5Risk of experiencing a pre-specified cardiopulmonary event* compared to matched patients without COPD for the full look-back period and within 3 years pre-index (incident cohort). Incident cohort (full look back): n=77,428; matched non-COPD control cohort: n=154,856. Incident cohort (with ≥3 years pre-index data): n=70,034; matched non-COPD control cohort (n=140,068). *a composite measure comprising cardiovascular events (ie heart failure, myocardial infarction, cardiac arrest, emergency revascularisation, tachyarrhythmia, atrial fibrillation and ischaemic stroke) and/or respiratory events (ie COPD exacerbations or LRTI).The odds of events were higher for patients with COPD before diagnosis than in matched controls without COPD.
Over the full available look-back period, the odds ratio was 3.8 (95% CI 3.8-3.9) for cardiopulmonary events, 3.6 (95% CI 3.6-3.7) for respiratory events and 2.2 (95% CI 2.1-2.2) for severe cardiovascular events. Restricting the analysis to incident patients with at least 3 years of pre-index data and their matched controls, the odds ratio was 4.9 (95% CI 4.8-5.0) for cardiopulmonary events, 5.3 (95% CI 5.2-5.4) for respiratory events and 2.7 (95% CI 2.6-2.8) for severe cardiovascular events.Forest plot comparing the odds of cardiopulmonary, respiratory and severe cardiovascular events in the incident cohort before COPD diagnosis with those in matched controls without COPD within 3 years of pre-index date and over the full lookback period.Abbreviations: CI, confidence interval; COPD, chronic obstructive pulmonary disease; CV, cardiovascular; LRTI, lower respiratory tract infection; OR, odds ratio.
In the sensitivity analysis including only hospitalized LRTIs, the odds of experiencing a severe respiratory event was substantially higher among patients with incident COPD than among matched non-COPD controls, both within 3 years pre-index (OR 9.2, 95% CI 8.8–9.6) and over the full look-back period (OR 5.8, 95% CI 5.7–6.0 Figure S5). The association for the severe cardiopulmonary composite was consistent with the primary analyses, within 3-years pre-index (OR 4.6, 95% CI 4.6–4.7) and over the full look-back period (OR 3.6, 95% CI 3.6–3.7).
Overall COPD and Matched Cohorts
For the overall cohort, patients with COPD also had higher odds of a cardiopulmonary event than matched controls within 3 years pre-index (OR 4.7; 95% CI: 4.6, 4.7) and for the full look-back period (OR 5.3; 95% CI 5.2, 5.3), (Figure S4 and Table S4).
HCRU and Cost
More patients with incident COPD accessed health care (PC, outpatient and inpatient services) compared to matched patients without COPD during the baseline period 1 year pre-index (Figure S6A). The cost/person during this period was 2.4 times higher for patients with COPD compared to non-COPD controls for PC access (£36.44 vs £15.20), 4.5 times higher for outpatient access (£98.18 vs £21.71) and 4.3 times higher for inpatient admissions (£473.57 vs £110.82) (Figure S6B).
Discussion
The novelty of our study derives from its examination of the prevalence of cardiopulmonary events pre-COPD diagnosis (and matched controls), in a large, real-world setting in a nationwide linked EHR cohort of incident cases of COPD in England. Within the 3 years preceding diagnosis, cardiopulmonary events were recorded in over half of the incident COPD cohort, who had approximately five-fold higher odds of an event than matched controls. Across the full look-back period, the proportion increased to more than three-quarters, with approximately four-fold higher odds than controls.
Atrial fibrillation and heart failure were the most commonly experienced severe CV events. Although atrial fibrillation, heart failure and myocardial infarction are common in patients with COPD,12,16 the novelty of our findings lies in demonstrating that this CV burden is already evident before the first recorded COPD diagnosis. CV events occurring alongside recurrent LRTIs or other acute respiratory presentations, may identify patients who were previously undiagnosed.
These patterns should not be regarded as diagnostic of COPD in isolation, but may provide clinically relevant opportunities for respiratory assessment, including quality-assured spirometry where appropriate.
Our findings are important since evidence on cardiopulmonary risk particularly pre-COPD is scarce, and CV risk factors are inadequately monitored, undertreated, and poorly controlled in people living with COPD.3,39 This is particularly relevant given the National Health Service Long Term Plan’s identification of CVD as a clinical priority and the single biggest condition where lives can be saved over the next 10 years.40 The occurrence of cardiopulmonary events up to three years before COPD diagnosis, underscores a critical gap that justifies prioritizing systematic evaluation and proactive cardiopulmonary risk management as early as the point of diagnosis while accelerating early detection of COPD in at-risk patients. Our results reinforce the need for earlier diagnosis of COPD and management of cardiopulmonary risk through proactive therapeutic interventions to prevent exacerbations and reduce both cardiopulmonary risk and mortality, as well as the need to enhance integration between cardiologists, pulmonologists and PC providers.41
Traditionally viewed as separate entities, evidence is accumulating to suggest a complex bi-directional relationship between COPD and CVD.9,16,42 The results of our study support the “syndemic” model proposed by Fabbri et al,17 which suggests that COPD exists as part of a network of interacting chronic diseases with common risk factors and exacerbating pathways amplifying disease burden and complicating management, necessitating early detection and comprehensive care. Shared exposures and biological pathways, including smoking, systemic inflammation, oxidative stress and hypoxaemia, may contribute concurrently to pulmonary and CV disease before COPD is formally diagnosed.9,16,18 Irrespective of the underlying mechanism, many have found increased risk of CVD among patients with COPD than those without;10,12,15,43–45 people with COPD being up to 2.5 times more likely to have CVD than those without,12,43,44 and approximately 10 times more likely to experience a serious CV event in the first 7 days following a COPD exacerbation.15 Our study differed from these as we examined risk of cardiopulmonary events, a novel composite endpoint comprising both severe CV and respiratory events, by category as well as individually, before COPD diagnosis.
We found a consistently high prevalence of pre-COPD cardiopulmonary events, irrespective of socioeconomic status (ie IMD) and UK geographical region. The fact that COPD does not discriminate in this regard is counter-intuitive since CVD risk correlates negatively with socioeconomic status,46 and warrants further study. Cardiopulmonary events were also common in patients with and without comorbid asthma (83.0% and 73.4%, respectively), although asthma–COPD overlap, diagnostic misclassification, and the independent CV risk associated with asthma may have influenced these estimates.
Approximately 2/3 of younger adults (aged 35-60 yrs) also had evidence of severe cardiopulmonary events pre-COPD diagnosis, suggesting that it is never too early to start assessing cardiopulmonary risk. Unfortunately, the presence of CVD in COPD may be ignored by attending physicians who focus on their own disease of interest. Further direction from guidelines on how, for whom, and when to assess cardiopulmonary risk would be helpful.
A recent multidisciplinary expert framework identified the absence of structured screening algorithms, defined referral pathways and clear cross-specialty responsibilities as key gaps in cardiopulmonary risk management. Proposed actions included earlier respiratory and cardiovascular risk assessment, routine spirometry and preventive care, timely treatment optimisation, and primary care-coordinated referral and follow-up across respiratory and cardiology services.3
We also noted significant under-utilisation of the QRISK assessment tool in routine clinical practice (documented for only 16.1% of our COPD cohort for whom it was recommended), and high rates of healthcare contact and cost prior to diagnosis. The former highlights a potential gap in current CV risk assessment protocols within the context of COPD management, suggesting underuse or under-recording rather than necessarily absent CV-risk consideration. Formal assessment can guide risk-factor management and preventive treatment, particularly in patients with recurrent respiratory presentations.20,47 However, because QRISK3 may underestimate COPD-specific risk and available tools may not fully capture exacerbation-associated risk, scores should complement clinical judgement and comprehensive cardiopulmonary assessment.13,15,47 The substantially greater HCRU and costs observed before COPD diagnosis demonstrate the pre-diagnosis burden extends beyond clinical outcomes and suggests missed opportunities for integrated care, and the potential for significant cost savings to the UK health service.
The cardiopulmonary-associated costs incurred by our COPD cohort prior to diagnosis, compared to those without COPD, were 2 times higher cost/person for PC access and approximately 4 times higher for both outpatient and inpatient appointments, indicating an already resource-intensive population. The substantial HCRU and per-person costs associated with undiagnosed COPD may provide opportunities for targeted respiratory assessment and present a powerful financial incentive for payers and commissioners to implement programs for earlier disease recognition and intervention. Prospective studies should evaluate the clinical effectiveness and cost-effectiveness of earlier case-finding and integrated cardiopulmonary assessment.
Limitations
Limitations of our study include those common to observational studies, including missing data, event observation and referral bias. Observation bias may have inflated relative cardiopulmonary risk estimates in patients with more frequent healthcare interactions with greater opportunity to diagnose and record comorbidity. Some pre-diagnosis events may have prompted COPD investigation or represented manifestations of unrecognised disease, consistent with previously reported missed diagnostic opportunities.20 Furthermore, cardiopulmonary events were identified using prespecified EHR codes which may have introduced misclassification bias through incomplete ascertainment or inaccurate classification of outcomes, and COPD was similarly identified from routinely recorded diagnostic codes and could not be confirmed by post-bronchodilator spirometry for all patients, creating the potential for diagnostic misclassification.
LRTIs were included as respiratory events in incident COPD as exacerbations before COPD diagnosis may be coded as LRTIs; however, their inclusion may have increased the estimated prevalence of severe respiratory events. Therefore, a sensitivity analysis was conducted including hospitalised LRTIs only which showed consistent findings. Bias may also have been introduced due to differing index dates between COPD cohorts and controls.
However, common sources of bias in comparative analyses were mitigated to some extent by matching COPD and control cohorts. Undiagnosed COPD in some controls may have attenuated between-group differences. As with all observational studies, residual confounding is possible despite matching COPD and control cohorts on key variables.
However, estimates may also have been underestimated in these patients due to referral to secondary care. Smoking, including intensity and pack-years, was not adjusted for and may confound the observed associations because it is related to both COPD and cardiopulmonary disease. Our objective was to estimate the total, real-world COPD-associated burden rather than the association independent of smoking between the COPD and matched non-COPD populations, including differences in smoking history.
In addition, this study was conducted using healthcare data from England which may limit generalisability of the findings to other countries and health systems. Our findings also demonstrated correlation rather than a definitive causal relationship. However, a strong correlation (as noted in the current study) provides significant value by revealing predictable patterns among variables, enabling precise forecasting and the identification of hidden interdependencies, even when a direct cause-and-effect relationship has not been established.
The study draws credibility from the inclusion of a large number of patients, exceeding initial feasibility expectations, considered more than adequate to support the descriptive aims of the study, including stratified analyses and reporting across relevant sub-groups. The use of CPRD Aurum represents a key strength, as it enables linkage with HES and ONS datasets, providing results generalisable to the wider COPD population in the UK through the use of routinely collected clinical data. Future avenues of research could include identification and assessment of broader risk factors (eg hyperlipidaemia, transient ischemic attack), as well as lifestyle factors including smoking and family history of CVD.
Conclusion
Cardiopulmonary events are common during the three years preceding a first recorded COPD diagnosis, highlighting a potential opportunity for earlier recognition of COPD and assessment of CV risk. Recurrent lower-respiratory presentations, persistent respiratory symptoms or otherwise unexplained cardiopulmonary events, particularly in patients with relevant exposure histories, could prompt consideration of targeted clinical review and quality-assured spirometry, rather than population-wide screening. These observational findings are hypothesis-generating and may help inform the development and prospective evaluation of pathways for earlier COPD recognition and integrated cardiopulmonary assessment against the risks of overdiagnosis and overtreatment.