Chronic Obstructive Pulmonary Disease in UK Primary Care: Incidence and Risk Factors

Abstract

We evaluated the association of chronic obstructive pulmonary disease (COPD) with modifiable risk factors such as smoking and prescription medications, and investigated possible risk factors unique to patients who had never smoked. The UK General Practice Research Database was used to identify a cohort of patients with a first diagnosis of COPD (n = 1927) along with age- and sex-matched controls without COPD (n = 16 546). The incidence of COPD diagnoses and the risks associated with medication use, co-morbidities, and demographic factors, were estimated. The incidence of COPD was 2.6 per 1000 person-years (95% confidence interval [CI]: 2.5–2.7) among 40–89 year-olds. The risk significantly increased in current and former smokers (OR: 6.15 [95% CI: 5.41–7.00] and 3.45 [95% CI: 2.96–4.02]), respectively. The risk was significantly lower in former smokers than current smokers (OR: 0.61; 95% CI: 0.52–0.71). Current statin use was significantly associated with a reduced risk (OR: 0.45; 95% CI: 0.25–0.80). In never smokers, risk factors included advanced age and obesity. The risk in never smokers was more strongly related to paracetamol use (OR: 1.82; 95% CI: 1.33–2.49) than in current and former smokers (OR: 1.48; 95% CI: 1.18–1.86). In summary, COPD is associated with a range of cardiovascular and respiratory conditions and the risk is influenced by current and past medications. While the risk factors are similar in smokers and never smokers, some were unique to never smokers. Moreover, subjects who stopped smoking had a substantially lower COPD risk than those who continued smoking.

Keywords:

• COPD
• Epidemiology
• UK
• Incidence

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is a heterogeneous, chronic condition characterized by airflow limitation that is not fully reversible ([1]). Population-based studies have reported the prevalence of COPD to be 4.1% in the UK ([2]), 8–20% in Latin-American cities ([3]) and 5.9% in China ([4]). A study in UK primary care estimated the prevalence of diagnosed COPD to be 1–2% ([5]).

Most patients have a history of smoking, which is one of the major risk factors for COPD ([6]). However, it has been reported that approximately 25% of new COPD cases occur in non-smokers ([7]). Other conditions which have been associated with COPD include pneumonia, heart disease, malignancies, osteoporosis, musculoskeletal disorders, insomnia, peptic ulcer, migraine, sinusitis, depression, and cancer ([1], [8],9,10). Recent data has documented that drugs such as paracetamol or statins may influence the risk of COPD. Use of paracetamol has been shown to have a negative effect on lung function ([11]) and this may cause an increased risk of COPD. Several recent studies have shown that the use of statins may reduce morbidity and mortality in COPD ([12],13,14).

Few studies have estimated the incidence of COPD. A study using the UK General Practice Research Database (GPRD) found the overall annual incidence of COPD to be 7.2 per 1000 persons in the general population in 1993 ([15]). However, this study only examined patients aged 60–85 years. The primary aim of our study was to describe the characteristics of a cohort of patients newly diagnosed with COPD in the UK general population and their treatment. We also evaluated the association of COPD with modifiable risk factors such as smoking or prescription drug use and, given the proportion of cases which occur in non-smokers, we investigated risk factors for COPD in patients who had never smoked.

METHODS

Data source

The GPRD contains information entered by about 1500 primary care physicians and covers a population of 3,000,000 individuals, which is representative of the UK general population ([16]). The database includes the complete medical records of all individuals registered with participating primary care physicians, including demographics, diagnoses (recorded using Oxford Medical Information System [OXMIS] and Read codes that map to codes in the 8th edition of the International Classification of Diseases [ICD-8]), prescriptions (generated automatically from the system), and referrals. This information is anonymized and sent to the Medicines and Healthcare products Regulatory Agency (MHRA), which oversees quality control and management of data for use in research projects. Data from the GPRD have been validated in several studies ([17],18), including those of respiratory diseases and COPD ([5], [9], [15], [17], [19],18,19,20,21).

Study design

We identified a source population of individuals aged 40–89 years between 1 January and 31 December 1996 and included those who met the following criteria: at least 2 years' enrolment with the primary care physician, 1 year's computerized prescription history and no diagnosis of pulmonary fibrosis, kyphoscoliosis, asthma, COPD or cancer before the start date in order to avoid cases of rediagnosis of a pre-existing condition.

This source population of 808,513 patients was then followed from the start date until the earliest occurrence of one of the following endpoints: detection of potential cases of COPD using OXMIS/Read codes for obstructive lung diseases, chronic bronchitis and emphysema, meeting one of the exclusion criteria (such as cancer), reaching 89 years of age, death or reaching 31 December 1996. Following removal of all patient personal identifiers, the computerized records of all potential cases (n = 2351) were reviewed manually in order to identify non-cases, possible cases and definite cases of COPD.

A patient was classified as a definite case if a diagnosis of COPD was recorded and there was information about specific treatment for COPD (e.g., oral or inhaled corticosteroids, anticholinergic agents, β₂-agonists or theophylline) during the 2 months following diagnosis. A patient was classified as a possible case if no specific treatment for COPD followed diagnosis. A patient was classified as a non-case if, on manual review of the patient record, text in the comments field of the patient's medical records indicated that the disease started before the index date (i.e., was pre-existing), if a diagnosis other than COPD (such as bronchitis) was subsequently confirmed or if it was revealed that the patient had cancer.

Validation of COPD cases

To ascertain the validity of a COPD diagnosis, and to confirm it as a first-ever diagnosis, we sent questionnaires to the primary care physicians of a random sample of 195 anonymous, definite or possible cases identified from the manual review. Questions were asked regarding the patients' symptoms, the diagnostic procedures used, the patients' previous conditions, and complications following diagnosis.

Of the 195 questionnaires sent out, 176 (90%) were returned completed, of which 83 were for definite cases and 93 for possible cases. The completed questionnaires confirmed the diagnosis of COPD in 73 of our definite cases (88%) and 69 of our possible cases (74%). As the confirmation rate was lower for possible cases than for definite ones, we decided to study only the latter as final cases, giving a population of 1927 patients with an incident COPD diagnosis.

Control population

The same source population and eligibility criteria were used to select a control cohort matched by age (± 1 year) and sex, who had no record of COPD at any time (n = 16 546). An index date in 1996 was randomly assigned to each individual in the control cohort.

Data collection

The following data were collected from the GPRD: demographic data, smoking status, most recent weight and height for calculation of body mass index (BMI), alcohol intake, co-morbidities in the previous year, number of consultations, and admissions and referrals during the previous year. We also collected data regarding the use of inhaled and oral steroids, anticholinergic agents, β₂-agonists, theophylline, non-steroidal anti-inflammatory drugs (NSAIDs), aspirin, paracetamol, statins and influenza vaccination in the year before the index date. The use of prescription drugs was also evaluated according to whether it was ‘current’ (the most recent prescription lasted until the index date or ended in the 30 days before the index date); ‘recent’ (the supply of the most recent prescription ended between 31 and 365 days before the index date); ‘past’ (when the supply of the most recent prescription ended more than one year before the index date); or ‘no use’ (when there was never any recorded use before the index date).

Statistical analysis

The incidence of COPD was calculated as the ratio of the estimated number of cases over the total person-years within age and sex strata, weighting the number of computer-detected cases by the confirmation rate obtained in the validation study. The relative risk (RR) of newly diagnosed COPD associated with age and sex was also estimated. The distributions of co-morbidities, treatment patterns, and potential risk factors among patients with COPD and the control patients were analysed and compared by unconditional logistic regression in the nested case-control design. The odds ratio (OR) and 95% confidence intervals (CI) for each variable were estimated, adjusting for age, sex, number of visits, referrals, and hospitalizations, smoking, BMI and other co-morbidities. Both cases and controls who had not visited their primary care physician in the previous year were excluded from these analyses.

RESULTS

Incidence

The overall incidence of COPD among patients aged 40–89 years was 2.6 per 1000 person-years (95% CI: 2.5–2.7). The corresponding prevalence estimate was 0.4%. The incidence estimates for both males and females increased steadily until the age of 79. From 80–89 years, the incidence increased slightly in men, whereas it plateaued in women (Figure 1). The incidence of COPD was higher in men than in women in all age strata (Figure 1).

Figure 1 Incidence of chronic obstructive pulmonary disease in UK primary care by age and sex.

Behavioural and co-morbid risk factors

The distribution of characteristics of COPD cases in comparison with the control population (matched by age and sex) is shown in Table 1. The logistic regression model showed that the risk of being diagnosed with COPD was significantly higher in patients who had made greater use of health resources during the previous year.

Table 1 Co-morbidities and other baseline characteristics of COPD cases and controls, and their association with a new diagnosis of COPD

	Cases (N = 1927)		Controls (N = 16 546)
	n	(%)	n	(%)	OR^‡	(95% CI)^‡
Age (years)
40–59	399	(20.7)	3261	(19.7)	N/A	N/A
60–69	585	(30.4)	5044	(30.5)	N/A	N/A
70–79	638	(33.1)	5606	(33.9)	N/A	N/A
80–89	305	(15.8)	2635	(15.9)	N/A	N/A
Sex
Male	1020	(52.9)	8431	(51.0)	N/A	N/A
Female	907	(47.1)	8115	(49.0)	N/A	N/A
Number of visits to primary care physician^†
≤1	136	(7.1)	2838	(17.2)	1.00	N/A
2–5	725	(37.6)	7812	(47.2)	1.62	(1.33–1.98)
> 5	1066	(55.3)	5896	(35.6)	2.16	(1.76–2.65)
Number of referrals†
0	693	(36.0)	8676	(52.4)	1.00	N/A
1–5	1056	(54.8)	7086	(42.8)	1.49	(1.33–1.67)
> 5	178	(92.0)	784	(4.7)	1.81	(1.46–2.25)
Number of hospitalizations†
0	1510	(78.4)	14382	(86.9)	1.00	N/A
1 or 2	332	(17.2)	1853	(11.2)	1.23	(1.06–1.43)
> 2	85	(4.4)	311	(1.9)	1.24	(0.93–1.67)
Smoking^§
Never	509	(26.4)	9786	(59.1)	1.00	N/A
Current	841	(43.6)	2502	(15.1)	6.15	(5.41–7.00)
Former	367	(19.0)	1920	(11.6)	3.45	(2.96−4.02)
BMI (kg/m^2) (most recent of previous 5 years)^§
< 20 (underweight)	131	(6.8)	470	(2.8)	1.83	(1.44–2.34)
20–24.9 (normal)	482	(25.0)	4036	(24.4)	1.00	N/A
25–30 (overweight)	438	(22.7)	4896	(29.6)	0.79	(0.68–0.92)
> 30 (obese)	153	(7.9)	1192	(7.2)	1.21	(0.98–1.50)
Co-morbidities^¥^†
Diabetes	91	(4.7)	1050	(6.3)	0.57	(0.44–0.72)
No drug treatment	28	(1.5)	286	(1.7)	0.63	(0.41–0.96)
Insulin only	5	(0.3)	81	(0.5)	0.44	(0.17–1.17)
Oral drugs only	53	(2.8)	615	(3.7)	0.59	(0.43–0.80)
Both treatments	5	(0.3)	68	(0.4)	0.34	(0.13–0.88)
Hyperlipidaemia	113	(5.9)	1316	(8.0)	0.70	(0.56–0.87)
Not treated	80	(4.2)	850	(5.1)	0.81	(0.63–1.05)
Treated with statins	14	(0.7)	189	(1.1)	0.52	(0.29–0.94)
Other treatments	12	(0.6)	162	(1.0)	0.64	(0.34–1.21)
Heart failure	136	(7.1)	272	(1.6)	2.38	(2.00–2.82)
Hypertension	458	(23.8)	4850	(29.3)	0.79	(0.70–0.90)
Respiratory infections	891	(46.2)	2453	(14.8)	3.97	(3.56–4.43)

BMI: body mass index; CI: confidence interval; OR: odds ratio.

† In the previous year.

‡OR and 95% CI were adjusted for all variables in the table using an unconditional logistic regression model.

§Smoking status was unknown in 10.9% of patients with COPD and 14.1% of the control cohort. BMI could not be calculated in 37.5% and 36.0% of patients with COPD and control cohorts, respectively.

^¥The reference category in each case was absence of the studied factor.

The risk of COPD was also significantly increased in current smokers (OR: 6.15; 95% CI: 5.41–7.00) and former smokers (OR: 3.45; 95% CI: 2.96–4.02). Patients who were underweight had a significantly increased risk of a new COPD diagnosis (OR: 1.83; 95% CI: 1.44–2.34), whereas those who were overweight had a significantly reduced risk (OR: 0.79; 95% CI: 0.68–0.92). Excessive alcohol consumption (> 35 units per week) was also significantly associated with a diagnosis of COPD (OR: 2.22; 95% CI: 1.59–3.09).

The risk of COPD was significantly reduced in patients diagnosed previously with diabetes (OR 0.57; 95% CI: 0.44–0.72) and hypertension (OR: 0.79; 95% CI: 0.70–0.90). Patients with hyperlipidaemia also had a significantly reduced risk of COPD (OR: 0.70; 95% CI: 0.56–0.87). However, when the type of treatment was considered, only patients receiving statins were significantly less likely to develop COPD than patients without hyperlipidaemia (OR: 0.52; 95% CI: 0.29–0.94). A history of respiratory infections (OR: 3.97; 95% CI: 3.56–4.43) or heart failure (OR: 2.38; 95% CI: 2.00–2.82) was more common in patients with COPD than those without COPD.

In our univariate analysis, we observed an apparent association between COPD and myocardial infarction, peptic ulcer, depression or osteoporosis, but this disappeared after adjustment in the multivariate model (data not shown). No association was found with inflammatory bowel disease (OR: 1.0; 95% CI 0.5–1.7) or dyspepsia (OR: 1.0; 95% CI: 0.9–1.1).

We conducted further analyses according to whether patients were current or former smokers, or had never smoked (Table 2 and Table 3). The data from the 1208 current and former smokers with COPD and 4422 controls showed that, compared with controls, current and former smokers with COPD were significantly more likely to have visited their primary care physician more than once in the year before diagnosis (OR for 2–5 visits: 1.71; 95% CI: 1.32–2.21, and for more than 5 visits: 2.59; 95% CI: 1.97–3.40), or to have had at least one referral to a specialist (OR: 1.54; 95% CI: 1.32–1.79, and OR for > 5 visits: 1.71; 95% CI: 1.27–2.32). Being underweight was significantly more common in current and former smokers who received a new diagnosis of COPD than in controls (OR: 1.69; 95% CI: 1.26–2.26). Compared with controls, COPD was significantly less common in those with a BMI of 25–30 kg/m² (OR: 0.75; 95% CI: 0.62–0.90), and was significantly more common in those who had a respiratory infection in the previous year (OR: 4.01; 95% CI: 3.46–4.65), or had received a previous diagnosis of heart failure (OR: 2.12; 95% CI: 1.63–2.75). A history of diabetes (OR: 0.55; 95% CI: 0.40–0.76) or hypertension (OR: 0.80; 95% CI 0.68–0.95) was significantly less common in newly diagnosed COPD patients than controls.

Table 2 Co-morbidities and other baseline characteristics of COPD cases and controls who are current or former smokers, and their association with a diagnosis of COPD

	Cases (N = 1208)		Controls (N = 4422)		OR^‡	(95% CI)^‡
	n	(%)	n	(%)
Age (years)
40–59	327	(27.1)	1095	(24.8)	1.00	N/A
60–69	400	(33.1)	1542	(34.9)	0.91	(0.76–1.09)
70–79	356	(29.5)	1400	(31.7)	0.84	(0.69–1.02)
80–89	125	(10.3)	385	(8.7)	0.98	(0.75–1.29)
Sex
Male	636	(52.6)	2587	(58.5)	1.00	N/A
Female	572	(47.4)	1835	(41.5)	1.11	(0.95–1.29)
Number of visits to primary care physician^†
≤1	84	(7.0)	725	(16.4)	1.00	N/A
2–5	445	(36.8)	2001	(45.3)	1.71	(1.32–2.21)
> 5	679	(56.2)	1696	(38.4)	2.59	(1.97–3.40)
Number of referrals^†
0	436	(36.1)	2240	(50.7)	1.00	N/A
1–5	171	(55.5)	1953	(44.2)	1.54	(1.32–1.79)
> 5	101	(8.4)	229	(5.2)	1.71	(1.27–2.32)
Number of hospitalizations^†
0	975	(80.7)	3840	(86.8)	1.00	N/A
1 or 2	489	(15.6)	502	(11.4)	1.13	(0.92–1.39)
> 2	44	(3.6)	80	(1.8)	0.94	(0.60–1.45)
BMI (kg/m^2) (most recent of previous 5 years)^§
< 20 (underweight)	102	(8.4)	192	(4.3)	1.69	(1.26–2.26)
20–24.9 (normal)	352	(29.1)	1244	(28.1)	1.00	N/A
25–30 (overweight)	289	(23.9)	1482	(33.5)	0.75	(0.62–0.90)
> 30 (obese)	93	(7.7)	346	(7.8)	1.03	(0.78–1.37)
Co-morbidities^¥^†
Diabetes	53	(4.4)	304	(6.9)	0.55	(0.40–0.76)
Hyperlipidaemia	80	(6.6)	380	(8.6)	0.76	(0.58–1.00)
Heart failure	137	(11.3)	225	(5.1)	2.12	(1.63–2.75)
Hypertension	273	(22.6)	1237	(28.0)	0.80	(0.68–0.95)
Respiratory infections	570	(47.2)	697	(15.8)	4.01	(3.46–4.65)
Ex-smoker	367	(30.4)	1920	(43.4)	0.61	(0.52–0.71)

BMI: body mass index; CI: confidence interval; OR: odds ratio.

† In the previous year.

‡ OR and 95% CI were adjusted for all variables in the table using an unconditional logistic regression model.

§BMI could not be calculated in 30.8% and 26.2% of patients with COPD and control cohorts, respectively.

^¥ The reference category in each case was absence of the studied factor.

Table 3 Co-morbidities and other baseline characteristics of COPD cases and controls who had never smoked, and their association with a new diagnosis of COPD

	Cases (N = 509)		Controls (N = 9786)		OR^‡	(95% CI)^‡
	n	(%)	n	(%)
Age (years)
40–59	50	(9.8)	1778	(18.2)	1.00	N/A
60–69	142	(27.9)	2940	(30.0)	1.57	(1.12–2.21)
70–79	196	(38.5)	3385	(34.6)	1.76	(1.26–2.45)
80–89	121	(23.8)	1683	(17.2)	1.77	(1.23–2.55)
Sex
Male	270	(53.0)	4654	(47.6)	1.00	N/A
Female	239	(47.0)	5132	(52.4)	0.79	(0.65–0.97)
Number of visits to primary care physician^†
≤1	37	(7.3)	1591	(16.3)	1.00	N/A
2–5	184	(36.1)	4651	(47.5)	0.93	(0.72–2.03)
> 5	288	(56.6)	3544	(36.2)	2.07	(1.41–3.03)
Number of referrals^†
0	182	(35.8)	5111	(52.2)	1.00	N/A
1–5	269	(52.8)	4208	(43.0)	1.42	(1.15–1.75)
> 5	58	(11.4)	467	(4.8)	2.16	(1.51–3.10)
Number of hospitalizations^†
0	382	(75.0)	8545	(87.3)	1.00	N/A
1 or 2	99	(19.4)	1075	(11.0)	1.38	(1.07–1.78)
> 2	28	(5.5)	166	(1.7)	1.84	(1.15–2.94)
BMI (kg/m^2) (most recent of previous 5 years)^§
< 20 (underweight)	28	(5.5)	264	(2.7)	2.14	(1.36–3.38)
20–24.9 (normal)	125	(24.6)	2725	(27.8)	1.00	N/A
25–30 (overweight)	145	(28.5)	3339	(34.1)	0.91	(0.70–1.17)
> 30 (obese)	55	(10.8)	817	(8.3)	1.63	(1.15–2.31)
Co-morbidities^¥^†
Diabetes	30	(5.9)	658	(6.7)	0.56	(0.37–0.84)
Hyperlipidaemia	29	(5.7)	859	(8.8)	0.67	(0.45–0.99)
Heart failure	90	(17.7)	528	(5.4)	2.54	(1.92–3.36)
Hypertension	148	(29.1)	3149	(32.2)	0.74	(0.60–0.91)
Respiratory infections	225	(44.2)	1393	(14.2)	4.04	(3.33–4.91)

BMI: body mass index; CI: confidence interval; OR: odds ratio.

† In the previous year.

‡OR and 95% CI were adjusted for all variables in the table using an unconditional logistic regression model.

§BMI could not be calculated in 30.6% and 27.0% of patients with COPD and control cohorts, respectively.

^¥The reference category in each case was absence of the studied factor.

We also evaluated patients who had never smoked (509 patients with COPD and 9786 controls) (Table 3). Like with the current and former patients, we found that factors significantly associated with COPD in never smokers included frequent visits to the primary care physician (OR for more than 5 visits in the year before diagnosis: 2.07; 95% CI: 1.41–3.03), at least one referral to a specialist (OR for 1–5 visits: 1.42; 95% CI: 1.15–1.75), or at least one hospitalization (OR: 1.38; 95% CI: 1.07–1.78). Female never smokers were significantly less likely than males to receive a diagnosis of COPD (OR: 0.79; 95% CI: 0.65–0.97). As with the current and former smokers, being underweight was significantly more common in never smokers who received a new diagnosis of COPD than in controls (OR: 2.14; 95% CI: 1.36–3.38).

Again, we observed that never smoker patients with COPD were significantly more likely than controls to have had a respiratory infection in the previous year (OR: 4.04; 95% CI: 3.33–4.91) or to have received a previous diagnosis of heart failure (OR: 2.54; 95% CI: 1.92–3.36). Within this subgroup of never smokers, a prior diagnosis of diabetes (OR: 0.56; 95% CI: 0.37–0.84), hyperlipidaemia (OR: 0.67; 95% CI: 0.54–0.99) or hypertension (OR: 0.74; 95% CI: 0.60–0.91) was significantly less common in newly diagnosed COPD patients than controls.

Risk factors which were unique to patients who had never smoked included obesity (OR: 1.63; 95% CI: 1.15–2.31) and advanced age (OR for 60–69 years old: 1.57 [95% CI: 1.12–2.21]; OR for 70–79 years old: 1.76 [95% CI: 1.26–2.45]; OR for 80–89 years old: 1.77 [95% CI: 1.23–2.55]).

A key finding of our study was that the risk of COPD was significantly decreased in those who ceased smoking compared with those who continued to smoke (OR: 0.61; 95% CI: 0.52–0.71).

Drug treatment

The patterns of specific drug treatment in the COPD and control groups before the index date were evaluated (Table 4). Patients in the COPD cohort were significantly more likely than controls to be current or recent users of respiratory drugs (anticholinergic agents, theophylline, β₂-agonists, or inhaled corticosteroids); this was particularly marked in short-term users (data not shown).

Table 4 Use of anti-inflammatory drugs, statins and influenza vaccination in COPD cases and controls, and their association with a first diagnosis of COPD

	Cases (N = 1927)		Controls (N = 16 546)		OR^†	(95% CI)^†
	n	(%)	n	(%)
Use of oral steroids^‡
None	1451	(75.3)	15531	(93.9)	1.00	N/A
Current	194	(10.1)	260	(1.6)	5.86	(4.68–7.34)
Recent	155	(8.0)	216	(1.3)	4.67	(3.64–5.98)
Past	127	(6.6)	539	(3.3)	2.04	(1.63–2.55)
Duration of oral steroid use
None	1451	(75.3)	15531	(93.9)	1.00	N/A
< 1 month	128	(6.6)	47	(0.3)	22.12	(15.08–32.43)
1 month–1 year	43	(2.2)	78	(0.5)	3.39	(2.18–5.26)
> 1 year	23	(1.2)	135	(0.8)	1.55	(0.95–2.52)
Use of NSAIDs^‡
None	780	(40.5)	6990	(42.2)	1.00	N/A
Current	199	(10.3)	1719	(10.4)	0.86	(0.71–1.03)
Recent	322	(16.7)	2429	(14.7)	0.95	(0.82–1.11)
Past	626	(32.5)	5408	(32.7)	0.91	(0.81–1.03)
Duration of NSAID use
None	780	(40.5)	6990	(42.2)	1.00	N/A
< 1 month	50	(2.6)	427	(2.6)	0.91	(0.65–1.26)
1 month–1 year	69	(3.6)	605	(3.7)	0.74	(0.56–0.99)
> 1 year	80	(4.2)	687	(4.2)	0.94	(0.72–1.22)
Remaining	948	(49.2)	7837	(47.4)	0.92	(0.83–1.03)
Paracetamol use^‡
None	633	(32.8)	7805	(47.2)	1.00	N/A
Current	436	(22.6)	2329	(14.1)	1.53	(1.32–1.78)
Recent	351	(18.2)	2510	(15.2)	1.09	(0.93–1.27)
Past use	507	(26.3)	3902	(23.6)	1.30	(1.14–1.49)
Duration of paracetamol use
None	633	(32.8)	7805	(47.2)	1.00	N/A
< 1 month	101	(5.2)	574	(3.5)	1.39	(1.08–1.79)
1 month–1 year	129	(6.7)	792	(4.8)	1.32	(1.05–1.65)
> 1 year	206	(10.7)	963	(5.8)	1.84	(1.51–2.24)
Remaining	858	(44.5)	6412	(38.8)	1.21	(1.08–1.37)
Use of aspirin^‡
None	1518	(78.8)	13575	(82.0)	1.00	N/A
Current	260	(13.5)	1969	(11.9)	0.95	(0.81–1.12)
Recent	70	(3.6)	462	(2.8)	1.05	(0.79–1.40)
Past	79	(4.1)	540	(3.3)	1.03	(0.79–1.36)
Duration of aspirin use^§
None	1518	(78.8)	13575	(82.0)	1.00	N/A
≤ 1year	125	(6.5)	810	(4.9)	1.04	(0.83–1.30)
> 1 year	135	(7.0)	1159	(7.0)	0.89	(0.72–1.09)
Use of statins^‡
None	1905	(98.9)	16199	(97.9)	1.00	N/A
Current	15	(0.8)	273	(1.6)	0.45	(0.25–0.80)
Recent	1	(0.1)	36	(0.2)	0.34	(0.05–2.61)
Past	6	(0.3)	38	(0.2)	1.37	(0.50–3.72)
Duration of statin use^§
None	1905	(98.9)	16 199	(97.9)	1.00	N/A
≤ 1 year	8	(0.4)	144	(0.9)	0.42	(0.19–0.92)
> 1 year	7	(0.4)	129	(0.8)	0.48	(0.21–1.09)
Influenza vaccination^‡
None	1059	(55.0)	9862	(59.6)	1.00	N/A
Current	82	(4.3)	530	(3.2)	1.37	(1.05–1.80)
Recent	528	(27.4)	4288	(25.9)	1.18	(1.04–1.35)
Past	258	(13.4)	1866	(11.3)	1.23	(1.05–1.45)

CI: confidence interval; NSAIDs: non-steroidal anti-inflammatory drugs; OR: odds ratio.

† OR and 95% CI were adjusted for matching variables (age and sex), life habits, and co-morbidities included in the first logistic model (Table 1).

‡Use was defined as ‘current’ when the most recent prescription lasted until the index date or ended in the 30 days before the index date; ‘recent’, when the supply of the most recent prescription ended between 31 and 365 days before the index date; ‘past’, when the supply of the most recent prescription ended more than one year before the index date; and ‘no use’, when there was no recorded use ever before the index date. For influenza vaccinations, ‘current’ refers to the vaccination taking place within the 30 days preceding the index date, ‘recent’, within 31 and 365 days before the index date, and ‘past’ when the vaccination took place more than 1 year before the index date.

^§< 1 month, and 1 month–1 year categories were combined because of their small size.

A history of oral steroid use was significantly more common in newly diagnosed COPD patients than controls, particularly if this was within the 30 days before the index date (OR: 5.86; 95% CI: 4.68–7.34). The most frequent indication for oral steroids was respiratory disease. The risk of COPD was significantly increased within this subgroup, particularly when steroids had been used in the 30 days preceding the index date (OR: 60.9; 95% CI: 33.4–109.3). In the subgroup of patients who were prescribed oral steroids to treat other diseases, the risk was smaller and only significant when steroids had been used within the previous 30 days (OR: 3.5; 95% CI: 1.7–7.4).

Current users of paracetamol had a significantly increased risk of COPD (OR: 1.53; 95% CI: 1.32–1.78), and this risk was slightly higher among long-term than short-term users (Table 4). The use of aspirin or other NSAIDs in the previous year was not significantly associated with COPD (Table 4). The estimated OR for long-term (> 1 year) use of cardioprotective aspirin was 0.89 (95% CI: 0.72–1.09).

Receiving the influenza vaccination was significantly associated with COPD (Table 4). The risk was slightly higher if the patient had been vaccinated within the month before index date (OR: 1.37; 95% CI: 1.05–1.80). Current use of statins was significantly associated with a reduced risk of COPD (OR: 0.45; 95% CI: 0.25–0.80), even after short-term use.

Drug treatment in current and former smokers

We also evaluated the patterns of specific drug treatments on the index date in COPD patients and controls who were current and former smokers (data not shown). Current and former smokers in the COPD cohort were significantly more likely than controls to be current or recent users of respiratory drugs (anticholinergic agents, theophylline, β₂-agonists, or inhaled corticosteroids), oral steroids (OR: 5.12; 95% CI: 3.51–7.47) and paracetamol (OR: 1.48; 95% CI: 1.18–1.86). The current, recent or past use of aspirin or other NSAIDs was not found to be significantly associated with COPD in current or former smokers. The use of statins in current and former smokers was significantly less common in newly diagnosed COPD patients than in controls (OR: 0.32; 95% CI: 0.14–0.73).

Drug treatment in never smokers

Like with the current and former smokers, we found that factors significantly associated with COPD in patients who had never smoked included current or recent use of respiratory drugs and a history of oral steroid use (particularly if this was within the 30 days before the index date; OR: 4.86; 95% CI: 3.21–7.33).

Current use of paracetamol was significantly more common in newly diagnosed COPD patients than in controls (OR: 1.82; 95% CI: 1.33–2.49) as was recent (OR: 1.53; 95% CI: 1.12–2.11) or past (OR: 1.49; 95% CI: 1.12–1.97) use. The risk of a COPD diagnosis in never smokers was found to be more strongly related to paracetamol use than in current and former smokers. The current, recent or past use of aspirin or other NSAIDs was not significantly associated with COPD. Again, the use of statins in never smokers was significantly less common in newly diagnosed COPD patients than in controls (OR: 0.60; 95% CI: 0.25–1.49).

DISCUSSION

Our observational study of data from UK general practice has revealed that the overall incidence of COPD among patients aged 40–89 years is 2.6 per 1000 person-years. We have shown therefore, that 0.2–0.3% of the UK population registered with a GP was diagnosed with COPD for the first time during a one-year period.

Results in context

To date, a limited number of studies have estimated the incidence of COPD based on data from primary care. Our results are similar to those from population studies that have reported the annual incidence of COPD to be 0.2% in 40–64 year-old males in Finland ([22]), 0.5% in a longitudinal, epidemiological study of a Polish population aged 19–70 years ([23]) and an estimated 1.5% (defined mainly by symptomatic criteria) per year in 40–59 year-old males in the Netherlands ([24]).

Another GPRD study found an overall annual incidence of COPD of 7.2 per 1000 persons in the general population ([15]). However, that study included only patients aged 60–85 years, whereas we also included patients aged 40 years and above, which will have lowered the overall incidence of COPD. Nevertheless, the incidence that we reported was lower even for the older age ranges, and this is most likely because we included only ‘definite’ cases where a recorded diagnosis of COPD was accompanied by information about specific treatment for COPD during the two months following diagnosis.

Effect of smoking

As expected, we found a significant association between COPD and smoking, with the highest risk in current smokers. Cigarette smoking is the most important risk factor for developing COPD. Cigarette smoking causes inflammation of lung tissue by increased infiltration of inflammatory cells such as macrophages, lymphocytes, neutrophils and eosinophils ([25], [26]). These inflammatory cells release proteinases such as matrix metalloproteinases which cause the alveolar destruction and emphysema that is characteristic of COPD ([27], [28]).

We also found that 26% of new COPD cases were in patients who had never smoked, which is in line with previous findings ([7], [29]). We further identified risk factors for COPD which were unique to patients who had never smoked. These included obesity and advanced age (greater than 60 years old). The link between obesity and COPD is increasingly being recognised, and a potential cause is the abnormal adipose tissue function in obesity that leads to altered expression and secretion of inflammatory factors and thus to systemic inflammation ([30]). In addition, excessive fat mass in obese individuals can also have direct, deleterious effects on lung function. Other factors that have been shown previously to increase the risk of COPD include pneumonia, musculoskeletal disorders, osteoporosis and malignancies ([1], [8],9,10).

As expected, our data show that the risk of COPD is decreased in those who cease smoking compared with those who continue to smoke. This supports the guidelines from the National Institute of Clinical Excellence (NICE) which advise that smoking cessation should be encouraged in those in whom COPD is suspected ([31]).

BMI

There is some ambiguity over whether low body weight is a risk factor for COPD or is a consequence of established COPD. Increased blood levels of tumour necrosis factor-α (TNF-α) occur in patients with COPD ([32]). TNF-α is known to be elevated in various human diseases associated with weight loss and so the elevated levels observed in COPD may be responsible for the weight loss that frequently occurs in COPD ([32]).

Low BMI is also a marker of poor prognosis in COPD ([33],34,35). Our analysis shows a significant association between a new COPD diagnosis and low BMI. This was true for both current and former smokers and never smokers and may in part reflect diagnosis of COPD late in the course of the disease, when the patient is already markedly affected by the disease.

Heart failure

There is limited published information on the frequency and relevance of prior cardiovascular diagnoses associated with incident COPD. Our study identified an association between COPD and heart failure. It is a possibility that the association is the result of sharing common risk factors such as smoking and not a causal relationship. COPD and heart failure are both characterized by systemic inflammation and increases in C-reactive protein ([1], [36],37,38). Plasma concentrations of proinflammatory cytokines such as TNF-α and interleukin (IL)-1 and IL-8 are increased, promoting muscle degradation and increasing the production of reactive oxygen species ([38]). In addition, patients with COPD or heart failure share a sedentary lifestyle, which also promotes abdominal obesity and systemic inflammation in the early stages of disease ([38]).

Hyperlipidaemia and statin use

We found that patients (both smokers and never smokers) with hyperlipidaemia had a significantly reduced risk of receiving a new COPD diagnosis. However, of those with hyperlipidaemia, only those receiving statins had a significantly reduced risk of receiving a new COPD diagnosis. Although there is little evidence in the literature that statins are preventative against COPD, several studies have documented that statins are associated with a reduction in COPD-related morbidity and mortality ([12],13,14), possibly through their anti-inflammatory effect ([39],40,41,42,43).

Other drug treatment

Current use of oral steroids in both smokers and never smokers was associated with an increased risk of COPD, particularly when the steroids were prescribed for respiratory indications in the month before COPD diagnosis. It is a possibility that the association with current steroid use is the result of the physician using steroids as a treatment trial in cases of suspected COPD.

Clinical implications

There is a need for the primary care physician to identify those patients who are at risk of COPD in order for them to be managed appropriately. Our data has indicated that the risk of a new COPD diagnosis is lower in former smokers than current smokers supporting the recommendation that smoking cessation should be encouraged in the ‘at-risk’ patients ([31]). The association identified between existing cardiovascular symptoms and new COPD diagnoses suggests that the management of cardiovascular symptoms may be of benefit in patients at risk of COPD. Further support for this is provided by the observation that statin use reduces the risk of a new diagnosis of COPD.

The association with low BMI perhaps indicates that diagnosis of COPD occurs late in the course of the disease, when the patient is already markedly affected by the disease. It may, therefore, be appropriate to routinely screen patients who are most at risk of the disease, such as smokers over the age of 49. Indeed, a recent screening study indicated that COPD is still underdiagnosed in primary care; it revealed that 10–20% of smokers over the age of 40 may have undiagnosed COPD, with the prevalence of undiagnosed disease rising sharply in individuals aged 50 years and over ([44]). Misdiagnosis of COPD in primary care is also common, with particular diagnostic confusion between COPD and asthma ([45]). Our data also indicate that it may be prudent to avoid paracetamol use in patients at risk of COPD, especially in current or former smokers. A link between paracetamol (acetaminophen) use and COPD has also been observed by others and the mechanism has been suggested to be related to a paracetamol-induced increase in oxidative stress in the lung tissue of susceptible individuals ([46]).

Strengths and limitations

The strength of our study is that we used data collected prospectively from a large representative sample of the general population registered with primary care practices in the UK, and identified newly diagnosed COPD in that setting. However, our study only evaluates the incidence of COPD from primary care consultations so may be an underestimation of the true incidence of COPD. A limitation of this data is that it evaluates incidence only over a one-year period. Increasing awareness of COPD in primary care since this date will likely have influenced the number of diagnoses. However, late diagnosis as well as misdiagnosis of COPD is still common in primary care ([44], [45]). As COPD is a slow, progressive disease, clinical diagnosis often does not occur until extensive damage has occurred. By this stage, the patient may have visited the primary care physician on several occasions about symptoms related to COPD. We aimed to address this issue by adjusting our results for the number of visits to the primary care physician.

Our results are also dependent on the diagnostic behaviour of the primary care physician and thus some factors considered to be risk factors or co-morbid conditions may in fact be consequences of the established disease. However, such late diagnosis does occur in COPD and we feel our results represent real-life practice. A further consideration is that as the diagnosis of COPD was based on the clinical judgment of the physician, the definition is likely to vary between individuals. Diagnosis by spirometry can identified COPD patients with less severe disease and minimal symptoms, but implementation of this methodology is likely to be hindered by practical thresholds in primary care ([47]).

CONCLUSIONS

The present study establishes the overall incidence of COPD among patients aged 40–89 years to be 2.6 per 1000 person-years. We have shown that the risk of a COPD diagnosis is higher in current smokers and those with a low BMI. A 40% reduction in the risk of COPD was observed in former smokers compared with current smokers. In patients who have never smoked, risk factors include advanced age and obesity. COPD is associated with a range of cardiovascular and respiratory conditions and the risk of a diagnosis is also influenced by current and past prescription medications.

Declaration of interests

This study was supported by an unrestricted research grant from AstraZeneca. LAGR has also received consulting fees from AstraZeneca. Saga Johansson is an employee of AstraZeneca R&D Mölndal, Sweden, and Mari-Ann Wallander was an employee of AstraZeneca R&D Mölndal at the time the research was conducted. All authors take responsibility for the content of this paper.

ACKNOWLEDGMENTS

We thank Dr. Catriona Turnbull from Oxford PharmaGenesis™ Ltd, who provided medical writing support funded by AstraZeneca R&D Mölndal, Sweden.