INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The exacerbation is an important event in the progression of chronic obstructive pulmonary disease (COPD), because it leads to increased mortality (1) and further impairment of quality of life (2), both involving high economic cost (3, 4). Respiratory infections are suggested as the main cause of COPD exacerbation (5), but other factors have been related to this event. Daily wheeze and bronchitic symptoms (2) contributed to a higher frequency of exacerbations in a series of patients with COPD. Regarding physiologic parameters, muscle weakness was associated with high use of health services for COPD exacerbation in a case-control study (6), and gas exchange impairment and pulmonary hemodynamic worsening were predictive of hospitalization for a COPD exacerbation in another series of patients with COPD (7). FEV₁ was found to be a predictor of COPD admission in one study (8), whereas others did not find this association (6, 7, 9). Frequent past exacerbations (2) and quality of life (2, 9) have also been related to a higher frequency of exacerbations. Among external factors, influenza vaccination (10) has been shown to reduce the risk of admission whereas exposure to high levels of air pollution (11) has been related to a higher risk of COPD admission.

The role of other modifiable potential risk factors has hardly been studied. Pneumococcal vaccination reduces the risk of pneumococcal pneumonia (12), and respiratory rehabilitation relieves dyspnea and improves control over COPD (13), but their effects on exacerbation are not consistent among studies (12, 14). The role of smoking has only been addressed in one study (7) that found no differences in admission between current smokers and noncurrent smokers. Inhaled corticosteroids reduced the rate of exacerbations in a recent trial (15) and their severity in a smaller one (16), but another large trial did not find differences in the number of exacerbations between treated and placebo group (17). Long-term oxygen therapy (LTOT) (18, 19) is associated with better survival of patients with COPD but has not been studied in relation to COPD exacerbation.

The understanding of the relevant determinants of COPD exacerbation is still insufficient. All the mentioned studies but one (7) have been mainly focused on one single factor, without considering that the observed associations could be dependent on other determinants. In addition, several potential risk factors of exacerbation, such as adherence to medication, social support, physical activity, or type of physician, have not been previously investigated. Therefore, we estimated the association between hospitalization for a COPD exacerbation and a wide range of potential risk factors, both modifiable and nonmodifiable, after adjusting for sociodemographic and clinical factors, using a case-control approach.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Design and Subjects

We performed a case-control study in the framework of a cross-sectional study (20). For the cross-sectional study, a systematic sample of one out of two patients hospitalized for a COPD exacerbation in four tertiary hospitals in the Barcelona area over 1 yr (from May 1, 1997 to April 30, 1998) was identified. COPD diagnosis was established by the ward pulmonologist and based on medical history, current symptoms, and available pulmonary function tests, following the European Respiratory Society (ERS) guidelines (3). An exacerbation was defined when the patient reported an increase in dyspnea, in sputum production, or sputum purulence (21). To be a case in the present case-control study, a subject from the cross-sectional study had to have had a previous COPD admission in the same hospital in the period extending from 3 mo to 5 yr earlier. The latest admission within this period was used as the referent event for selection of control subjects. A list of potential control subjects was obtained from the hospital's records of discharges, selecting patients discharged with the diagnosis of chronic bronchitis, emphysema, or COPD (International Classification of Diseases [ICD-9] codes 491, 492, or 496) from the same hospital at the time that the corresponding previous case discharge occurred (taking a range of ± 1 mo). Control subjects needed to be clinically stable during the 3 mo previous to the present exacerbation of the case. Figure 1 illustrates the cases and controls selection process, adapted from the study of Pearce and coworkers (22). Potential control subjects were interviewed by telephone until one was found eligible and accepted to participate. If no control subject was found from the list, the case was not included in the present study. After the recruitment, one case and two control subjects were excluded because of a ratio FEV₁/FVC > 88% (3), and six cases and nine control subjects were excluded owing to positive bronchodilator tests. The protocol was approved by the ethics committees of the participating hospitals, and written informed consent was obtained from all patients.

View larger version (17K): [in this window] [in a new window]   Figure 1.   Selection of case and control groups. (Adapted from Reference 22 by permission.)

Measures

During the hospitalization cases were asked to complete a questionnaire and anthropometric measurements were made. At least 3 mo after hospitalization, and during a clinically stable period of COPD, cases performed a forced spirometric test and provided arterial blood to measure gas tensions. Control subjects completed all the tests during a visit close to the date of admission of the corresponding case.

Information about potential risk factors of COPD exacerbation was obtained through an extensive bibliographic search that has been detailed in a previous study (20). A list of all factors considered in the present analysis is shown in Table 1. Most of the content of the questionnaire was obtained from previously validated instruments, whereas some questions were developed and pilot tested. Detailed information about all variables, sources of questions, and methods for spirometry and blood gas measures is available in the Online Data Supplement.

Statistical Analysis

A logistic regression model was used to obtain the odds ratio (OR) of hospitalization between cases and control subjects, after adjusting for potential confounders. First, a univariate analysis of each variable was carried out to determine its appropriate scale to be entered in the model. Second, sociodemographic and clinical variables were assessed and those with an independent statistically significant association with admission defined the "clinical model." Third, we adjusted the association between all variables of Table 1 and admission by the variables of the clinical model. Then, a multivariate model was obtained including all variables that after adjustment for the clinical model showed a p value < 0.25 (23), plus those considered a priori as clinically relevant. Finally, the most parsimonious model was fitted. The analysis was performed with Stata, release 5.0 (1997; StataCorp., College Station, TX).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 151 eligible cases were identified, but only 126 (83%) accepted to participate in the study. The reasons for nonresponses (n = 25) were: 20 refusals, one discharge before the interview, and four deaths. A comparison of responders and nonresponders showed a higher prevalence of LTOT utilization in nonresponders, although no differences in FEV₁, other severity markers, or prevalence of potential risk factors were found. We could select only 86 control subjects for the 126 cases, giving a total of 86 cases and 86 control subjects for the present analysis. There were no differences in sociodemographic and clinical variables between cases with (n = 86) and without (n = 40) a control (data not shown).

Table 2 shows that cases had a worse clinical status than control subjects, although almost no differences were found in sociodemographic factors. Having had three or more COPD admissions in the previous year (OR 3.82) and a lower level of percentage of predicted FEV₁ (OR 0.97 per percentual unit) were independently associated with admission for a COPD exacerbation, and constituted the clinical model.

For all factors listed in Table 1, crude and clinical model adjusted ORs were obtained, and those with a p value < 0.25 when adjusted by the clinical model are shown in Table 3. Oral corticosteroids and LTOT underprescription were found associated with an increased risk of admission, only the latter being statistically significant. Lack of influenza vaccination, LTOT utilization, LTOT overprescription, current smoking, and a better score in the physical scale of health-related quality of life were associated with a lower risk of COPD admission, although none of them achieved statistical significance. None of the remaining factors listed in Table 1 (including medical care, pneumococcal vaccination, respiratory rehabilitation, or compliance) was significantly associated with the admission.

In the final multivariate model (Table 4), having had three or more COPD admissions in the previous year, a lower level of FEV₁, and underprescription of LTOT were all found to be risk factors of hospitalization for a COPD exacerbation independently of other variables. Moreover, current smoking was negatively associated with admission. There were no relevant differences in the results between the nonconditional and the conditional logistic regression model. Results also were very similar after excluding never-smokers (n = 15): three or more COPD admissions in the previous year (OR 9.63, p = 0.005); FEV₁ (OR 0.96 per percentual unit, p = 0.001); underprescription of LTOT (OR 22.08, p = 0.009); and current smoking (OR 0.28, p = 0.018). Finally, results did not substantially change when the 40 cases initially excluded for not having a referent control were included in the final nonconditional model: three or more COPD admissions in the previous year (OR 4.50, p = 0.018); FEV₁ (OR 0.96 per percentual unit, p < 0.0005); underprescription of LTOT (OR 15.62, p = 0.014); and current smoking (OR 0.63, p = 0.269).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Despite having studied a wide range of both modifiable and nonmodifiable potential risk factors of COPD exacerbation, only a few clinical factors and one related to medical prescriptions were found to be independently associated with COPD admission. This is the first study integrating such a range of factors, and both positive and negative results deserve further consideration.

LTOT underprescription was found to be a strong independent risk factor of COPD admission. Moreover, LTOT utilization was associated with a lower risk of admission, independently of the clinical and sociodemographic variables. The Medical Research Council (MRC) study (18) did not find significant differences in the number of days spent in hospital due to exacerbations between oxygen-treated and control groups, although the investigators suggested that both groups had been much more supervised than the usual, and early detection of illness could have occurred. On the other hand, LTOT was found to be a risk factor of exacerbation in the univariate analysis of a previous study, although no association remained when adjusting for physiologic parameters, indicating that LTOT was probably a marker of severity in those patients (7). Our results are in agreement with the better survival found in patients treated with LTOT compared with a control group both in the MRC (18) and in the Nocturnal Oxygen Therapy Trial (NOTT) (19) studies, which included only patients with severe hypoxemia (PO₂ < 60 mm Hg and < 55 mm Hg, respectively). For COPD patients with moderate hypoxemia (56 to 65 mm Hg), one randomized controlled trial (24) did not find differences in survival among treated and control groups, but exacerbations were not studied. It could be argued against the present results that cases were more hypoxemic than control subjects because some effect of exacerbation could have remained. That would have led to a higher proportion of underprescription in cases. However, all cases in the group of underprescription had blood gases taken more than 3 mo after the exacerbation, and no significant differences were found in mean PO₂ among cases with tests performed previously to exacerbation and those with tests performed at least 3 mo after the exacerbation (61 mm Hg versus 64 mm Hg, p = 0.46). Although in two patients the lack of LTOT prescription was likely due to current smoking, its exclusion from the model still yields a high OR for LTOT underprescription (OR 22.23, 95% confidence interval [CI] 2.27 to 217.45). It could be said that COPD is poorly managed in the area where patients come from. However, a sustained effort has been made in Catalonia for the last 10 yr to study LTOT utilization patterns and to improve its effectiveness (25, 26).

Surprisingly, current smoking, compared with ex-smoking, was negatively associated with admission. A previous study (7) found no differences between smokers and noncurrent smokers in the risk of COPD admission. One possible explanation has been proposed by Anthonisen, who has suggested that severely ill patients with COPD spontaneously quit smoking, probably in response to their symptoms and disability, and do not do well afterwards (27). Prospective studies including a wider range of COPD severity would be helpful to clarify this issue.

Neither age nor sex was related to COPD admission, as in previous studies (2, 7, 9). Lower values of percentage of predicted FEV₁ were found associated with hospitalization for a COPD exacerbation. Some studies have reported a lack of association between this variable and COPD exacerbation or admission (6, 7, 9), whereas another has found FEV₁ to be predictive of COPD admission (8). The small number of subjects in some studies (6, 7) and the use of categorized FEV₁ instead of continuous in some analyses (7, 8) could explain these differences. Moreover, FEV₁ is one of the variables more consistently related to mortality of patients with COPD (28), which makes our results more coherent. To have three or more COPD admissions in the previous year was also found to be a risk factor of admission, as in a previous study (2). It could be argued that previous admissions is an intermediate variable and that its inclusion in the final model may lead to a certain degree of overadjustment. However, results for the remaining factors did not change when we excluded previous admissions from the model. Low body mass index (BMI) was associated with admission in the univariate analysis, in agreement with a previous study (7), and consistently with its relation to prognosis of COPD (29).

Oral corticosteroids were related to a higher risk of admission in the crude analysis. The lack of association in the adjusted models suggests that their intake could be interpreted as a marker of disease severity. Influenza vaccination could be acting as another marker of severity, because it was associated with a higher risk of admission in the crude analysis, contrarily to expected (10), whereas it did not show any relation in the adjusted models. Physical scale of health-related quality of life was related to COPD admission in the crude analysis, as in previous studies (2, 9), and maintained an association (although nonsignificant) in the clinical model adjusted analysis (Table 3). However, when we included dyspnea in this model, the relation was lost, suggesting that the association between physical quality of life and admission was confounded by dyspnea. Unfortunately, we have only available a retrospective measure of dyspnea, which is of doubtful validity, and for this reason has not been included in the analysis.

Inhaled corticosteroids did not show an association with COPD admission, according to a previous trial (17), and in contrast with another one (15). The effects of pneumococcal vaccination and respiratory rehabilitation could not be tested because of the small number of subjects and the distribution of such factors in our sample. Other variables related to medical care, such as the type of physician who controls the disease, were not associated with admission, this being consistent with the observed lack of association with mortality in a previous study (30). This is the first study assessing effects of poor compliance in patients with COPD, and more attention should be given to this issue even though we did not find an association with admission.

As in most epidemiologic studies, nonresponse (selection) bias can not be excluded. The comparison of responders and nonresponders among cases showed a higher prevalence of LTOT utilization in nonresponders, but information about the nonparticipating controls was not available. If the difference in LTOT utilization among responders and nonresponders of case and control groups was the same, the point estimate would go toward the null value, reinforcing our findings of LTOT being a protective factor. It is true that our results can only be extrapolated to patients with COPD who have suffered previous COPD admissions. However, in our opinion this is not a limitation, as emphasis on the attempt to make a valid comparison yields a stronger study design than the attempt to make subjects representative. The group of patients with COPD previously admitted is not a small population, and is responsible for high health expenditure.

One additional limitation of our study is that we have not used an operational definition of COPD exacerbation. In fact, it does not exist in epidemiologic studies nor in the clinical setting (31). Other studies have also used the COPD admission as a surrogate of exacerbation (6, 7, 9). In our study, an administrative database was used a posteriori to classify discharge diagnosis following ICD-9 in cases, with the aim of validating COPD diagnosis. Results showed that approximately 95% of patients were classified as having chronic bronchitis, emphysema, or COPD. In addition, patients excluded owing to high FEV₁/FVC ratio according to ERS criteria (3) meant only 1% of cases and 2% of controls. Therefore, misclassification, if any, is likely to be small and nondifferential. The exclusion of patients with a positive bronchodilator test could be argued, owing to the possible existence of bronchodilator reversibility in some patients with COPD. This exclusion criterion sought to avoid the inclusion of patients whose primary diagnosis could be asthma, because of potential differences between asthma and COPD in some of the variables examined, such as medical treatment or compliance. However, we must keep in mind that the interpretation of our results needs to be limited to COPD patients with negative bronchodilator test.

In summary, among a wide range of potential risk factors we have found that only previous admissions, lower FEV₁, and underprescription of LTOT are independently associated with a higher risk of admission for a COPD exacerbation. Caution is needed in the interpretation of these results, as the case-control approach should be followed by longitudinal studies. Further confirmation of such negative findings would suggest that the risk of exacerbation in patients with severe COPD is not affected by the studied factors, and that, possibly, most of the risk will be attributable to respiratory infections.