Longitudinal changes in the nature, severity and frequency of COPD exacerbations

Patients

A total of 177 patients with COPD were recruited from the outpatients dept of the London Chest Hospital in the first 5 yrs of this 6‐yr study. In the first year, 99 patients were recruited consecutively and from then on those who subsequent withdrew or died were replaced. The recruitment criteria were an FEV₁ <70% predicted from sex, age and height, a FEV₁/forced vital capacity (FVC) ratio <70%, β₂ agonist reversibility <15% and/or <200 mL, the absence of asthma, bronchiectasis, bronchial carcinoma or other significant respiratory disease, and a willingness to participate in a long-term study. Of these 177 patients, 132 were selected for this analysis on the basis that they had recorded diary card data on a minimum of 365 days. The reasons for 45 patients failing to record sufficient data were withdrawal, death or inadequate completion of diary cards, combined with late enrollment. The 132 patients did not differ significantly from the 45 excluded patients in any of the characteristics reported in table 1⇓, except for a slightly lower PEF (152 versus 181 L·min⁻¹; Wilcoxon p=0.0179) and FVC (2.14 versus 2.44 L; p=0.039). This cohort has been the subject of previous publications, which investigated the relationship of COPD exacerbations to quality of life 1, inflammatory markers 3, lung function decline 5, time course of symptoms and lung function 6, respiratory viruses 7, fibrinogen 8 and bacterial colonisation 9. This is the first longitudinal analysis on the history of exacerbations with 6 yrs of data.

Enrollment

At recruitment, measurements were made of FEV₁, FVC and PEF by rolling seal spirometer (Sensor Medic Corp., Yorba Lindo, CA, USA), reversibility to 400 µg inhaled salbutamol and arterialised ear lobe blood gases (model 278 Blood Gas Analyzer; Ciba-Corning, Medfield, MA, USA) 10. A history was taken of smoking habits (years of smoking, current smoking status). Patients were asked about their symptoms of dyspnoea, sputum production, wheeze and cough. The patients were also asked about their long-term inhaled and oral steroid use. The study had ethics approval from the Ethics Committee of the East London and City Health Authority and patients provided written informed consent.

Monitoring and exacerbation

The patients were asked to record postmedication PEF (Mini-Wright Clement Clark International Ltd, Harlow, UK) and any increase over their normal, stable condition in symptoms of dyspnoea, sputum purulence or sputum volume (major symptoms) and colds (nasal discharge/congestion), wheeze, sore throat, cough (minor symptoms) on diary cards each morning. A total of 35 randomly selected patients also measured daily FEV₁ and FVC using a hand-held spirometer (Micro Medical Ltd, Chatham, Kent, UK). Exacerbation onset was taken as the first of two or more consecutive days with increase in either two or more major symptoms, or any one major symptom plus any minor symptoms 1, 3, 5, 6. Symptoms were disregarded in identifying onset if recorded continuous in the 5‐day period preceding a suspected exacerbation onset. Two subjects separately identified exacerbations (at clinic visits and at data entry) and then later resolved disagreements in diagnosis and timing. Exacerbations for which no or insufficient diary-card symptoms were recorded by the patient, were on occasion identified by hospital admission for an acute exacerbation of COPD or by questioning at clinic visits. Patients with increased symptoms were encouraged to contact the clinical team by telephone, and were seen prior to treatment generally within 48 h. The exacerbations seen by the clinical team were classified as “reported exacerbations” with those unseen termed “unreported exacerbations”. Records were kept of hospitalisation throughout, and from November 1996 onwards, records were available of the date of initiation and type of treatment prescribed to patients both at the authors' clinic and also that prescribed by the patient's general practitioner (GP).

Statistical analysis

Data are presented as mean (sd) or median (interquartile range (IQR)) and comparisons performed by unpaired t‐test, Wilcoxon matched-paired sign-rank test or a Chi-squared test, as appropriate. Differences in exacerbation frequency between groups were estimated using generalised linear models assuming a poisson distribution in the frequency, and thus 95% confidence intervals (95% CI) are reported rather than IQR.

Indices of exacerbation severity

Exacerbation severity with respect to lung function and symptom count (the sum of the binary coded presence or absence of seven respiratory symptoms) was assessed over a 51‐day period. Baseline was taken as the mean of a parameter over days 14–8 preceding exacerbation onset, as no significant changes in lung function and symptoms were seen over this time period. The change in any parameter associated with exacerbation was taken as the difference between baseline and the day of onset. Recovery was the time from onset for a 3‐day moving average to equal or exceed the baseline. A moving average was used to avoid false early recoveries when lung function improved for just a single day, but then remained below baseline for a few more days. The authors analysed recovery from exacerbation at 35 days because in most clinical studies a patient not having an exacerbation for 4–6 weeks would be considered stable. Exacerbation nonrecovery was taken as recovery taking >35 days 6.

To assess changes over time of these indices, cross-sectional, generalised linear models were fitted using the xtgee command in Stata 5.0 (Stata Corporation, Texas, USA). These models examine time variations independently of cross-section variations in panel data 11. The distribution of the dependent variable was specified after inspecting histograms of the data and the independent variable was time in years.

Exacerbation frequency and symptom composition

Each patient was considered to have started on the same day (day 1=November 1, 1995), whether enrolled on or after that date, to avoid bias from patients recruited later during the study who may have had very high or low exacerbations frequencies. The subsequent 6 yrs were then divided into 24 quarter-yr periods. Within each period, the number of exacerbations was divided by the number of patients involved, and multiplyed by four, to give an annual exacerbation frequency. The symptom percentage was calculated by dividing the number of exacerbations with a given symptom by the total number of exacerbations and multiplying by 100. Linear regression was then used to assess time trends in frequency and symptoms. Quarter year periods were chosen as a compromise between year long periods, in which the number of patients at the start and end of the year would differ markedly, and between month long periods in which too few exacerbations would occur to sensibly calculate the symptom percentages. The analysis was repeated with allowance for seasonality, as the findings could potentially be biased by enrollment during the winter or withdrawal just before winter when exacerbations are more common. The exacerbation frequency was calculated as above, but without aligning the start dates of each patient. Sine and cosine terms with a year period, in addition to the trend term, were then fitted by the regression technique described above.

The authors chose not to analyse changes over time in the interval between exacerbations using cross-sectional models, since interval data will over estimate exacerbation frequency towards the end of the study period, when completed intervals for patients with infrequent exacerbations are less likely to be available relative to patients with frequent exacerbations. With the approach adopted by the authors, it was not sensible to analyse differences between groups, such as smoking status or drug therapy, as too few patients would be available in the later stages of the study, nor was the study designed with these analyses in mind.

Patients

The 132 patients (91 male, 31 female) studied had moderate-to-severe COPD (table 1⇑). Of these, 119 patients took inhaled steroids daily (1.53 (sd 1.1) mg·day⁻¹ beclomethasone equivalents) and 12 patients took a mean 5.91 (3.0) mg·day⁻¹ of oral prednisolone; 10 patients used both oral and inhaled steroids. The subgroup of 35 (31 male, four female) patients who recorded daily FEV₁ and FVC were similar to the others for the characteristics reported in table 1⇑, except they had a higher PEF (p=0.001) and a higher percentage of males (p=0.003).

Exacerbation frequency, changes over time

The patients participated in the study for a median of 918 (IQR: 666–1.365) days. During the 6 yrs of the study, seven of the 132 patients had no exacerbations and eight patients had just one. There were a total of 1,111 exacerbations of which 971 (87%) were identified from the symptom data recorded by the patients on their diary cards, 123 (11%) by questioning about symptoms at clinic, six by hospital admission alone, eight by treatment by the authors or GP alone, and three were recorded as an exacerbation but other data lost. In total, 511(46.0%) of all exacerbations were reported and seen by theclinical team. The odds ratio of an exacerbation being reported to the authors clinic team, was 0.84 (95% CI: 0.78–0.91; p<0.001) relative to the preceding year. However, the odds ratio of an exacerbation being treated by the authors or the patient's GP, relative to the prior year, did not change over time (0.93; 95% CI: 0.84–1.03; p=0.149).

The median exacerbation rate was 2.52 (IQR: 1.34–3.93) exacerbations·yr⁻¹. The exacerbation frequency was 3.43·yr⁻¹ (95% CI: 2.7–4.2), 0.75·yr⁻¹ (0.08–1.43) higher (p=0.029) in patients (n=38) with severe COPD (FEV₁ <30% pred and FEV₁/FVC <70% pred; Global Initiative for Chronic Obstructive Lung Disease (GOLD) category III), compared with those (n=94) with moderate COPD (FEV₁ ≥30% pred and <80% pred and FEV₁/FVC <70% pred; GOLD category II) whose frequency was 2.68 (2.35–3.01). The FEV₁ in the two groups were 0.67 L (sd 0.14) and 1.22 L (0.4), respectively (p<0.001). There was no significant difference in exacerbation frequency between GOLD IIA and GOLD IIB.

There was no significant difference in the exacerbation frequency of those 35 patients who recorded FEV₁, and FVC(3.17·yr⁻¹; 95% CI: 2.4–3.9) and the other 97 patients (2.80·yr⁻¹; 2.5–3.1; p=0.286). There was no significant difference in exacerbation frequency between the 64 patients who withdrew early from the study (median 3.06·yr⁻¹; 2.4–3.8) and those who were still participating at the end of the study in November 2001 (2.7·yr⁻¹; 2.3–3.1; p=0.307). There was also no difference in exacerbation frequency between the 19 patients who died and the other 113 patients (2.77 versus 2.92·yr⁻¹; p=0.724).

Annual exacerbation frequency for all 132 patients remained constant during the study, changing by only −0.025·yr⁻¹ (95% CI: −0.065–0.015; p=0.208) from a starting value of 2.88·yr⁻¹. The exacerbation frequency was also constant if allowance was made for seasonality (−0.025·yr⁻¹; −0.054–0.003; p=0.080), although there were significant winter peak to summer trough changes of 1.28 exacerbations·yr⁻¹ (p<0.001). Over time, symptoms at exacerbation of both sputum purulence and volume rose by 4.12%·yr⁻¹ (p=0.004) and 5.25%·yr⁻¹ (p=0.001) respectively, whilst wheeze fell by 2.8%·yr⁻¹ (p=0.004) (fig. 1⇓). There was no significant change in the other symptoms (table 2⇓).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.—

Percentage of exacerbations of a) dyspnoea, b) sputum purulence, and c) sputum volume, with the patient recording a major symptom on the day of onset, over time. Patients were assumed to have started the study on the same day. Circle size indicates the number of exacerbations recording in that quarter-year.

Exacerbation severity, changes over time

Table 3⇓ shows estimates of the annual change in the time-course indices and their value at the start of the study in November 1995. At the study start, symptom count at baseline was estimated as 0.36 and rose to 2.23 at exacerbation onset. Over time, symptom count did not rise significantly during the baseline period or at exacerbation onset, but their difference did increase by 0.05·yr⁻¹ (p=0.047). FEV₁ at baseline declined significantly by −34.5 mL·yr⁻¹ (p<0.001) but there was no change over time in FEV₁ at exacerbation onset. FVC did not change significantly over time at baseline or onset. At baseline PEF fell by 2.7 L·min⁻¹·yr⁻¹ (p<0.001) but PEF did not change at exacerbation onset over time (p=0.249).

Recovery in FEV₁, FVC and symptom count from exacerbation took significantly longer each year, by 0.55, 0.85 and 0.32 days·yr⁻¹, respectively (all p<0.001). No change was seen in recovery of PEF. Similar results were found for exacerbations that were untreated, with recovery respectively taking 0.34, 1.39 and 0.39 days·yr⁻¹ longer (all p<0.013). Again, changes in PEF recovery were nonsignificant.

Complete recovery to baseline levels did not always take place: 121 of 1,043 (11.6%) never recovered in PEF, 100 of 1,045 (9.6%) in symptom count, 22 of 221 (10.0%) in FEV₁ and 15 of 199 (7.5%) in FVC. Table 3⇑ also shows that there was no significant change in the number of exacerbations not recovering within 35 days over time.

Drug treatment and hospitalisation, changes over time

During the whole study, there were 65 hospitalisations for acute exacerbation of COPD, representing 5.9% of the 1,111 exacerbations. The odds ratio of hospitalisation increased by 1.26 each yr (95% CI: 1.084–1.459; p=0.002). The median length of stay in hospital was 11 days (IQR: 7–15 days) and did not change over time. Data collected over the last 5 yrs of the study showed that of 918 exacerbations, 223 (24.3%) were treated with oral steroids and 525 (57.1%) with antibiotics. There was a significant increase in treatment of exacerbations with oral steroids, as the odds ratio for treatment was 1.12 (95% CI: 1.00–1.25; p=0.051) relative to the previous year, but a fall in treatment with antibiotics, as the odd ratio was 0.91 (0.82–1.00; p=0.052).