This multicenter retrospective cohort study aimed to evaluate the effectiveness of empirical anti-pseudomonal antibiotics in terms of the length of the hospital stay in patients with frequent COPD exacerbation. We hypothesized that anti-pseudomonal antibiotics would reduce the length of the hospital stay. However, after adjusting for known confounding factors, the length of the hospital stay was not significantly different between the non-anti-pseudomonal and anti-pseudomonal antibiotics groups.
Although the PA detection rate in the present study was not lower than that in previous studies conducted in other countries (22% in the present study vs 4–13%), our hypothesis was not supported [10, 26]. To the best of our knowledge, this study is the first to assess the effectiveness of anti-pseudomonal antibiotics in patients with recurrent COPD exacerbation. Our non-significant results are consistent with a previous prospective cohort study targeting hospitalized patients who had a positive culture result for PA in lower-tract specimens [27]. In this previous study, only 18% of patients were treated with anti-pseudomonal antibiotics and, after adjusting for confounding factors, inadequate initial antibiotic use was not associated with 12-month mortality. In contrast, inappropriate initial antibiotic use in the context of PA pneumonia has been reported as associated with increased 28-day mortality [28]. This difference in results may be due to differences in the proportion of patients with pneumonia. In a study of PA pneumonia, the 28-day mortality rate was 51%, which is comparable to that in other reports [29]. Although PA detection among patients with COPD has been reported as associated with extremely high 2-year all-cause mortality (23–41%), the 1-month mortality was 0% [10, 30]. Considering the low 1-month mortality in patients with COPD exacerbation with PA isolation, empirical anti-pseudomonal antibiotics may not have adequate power to change the prognosis during hospitalization.
Initial treatment with non-anti-pseudomonal antibiotics may spare anti-pseudomonal antibiotic use. The median time to the next hospitalization was approximately half a year, and the previous use of antibiotics was detected in only one-third of observations. In the present study, although at least 70 observations/cases had positive sputum-culture results for PA before admission, PA was covered by empirical antibiotics in only 29% of cases. The patient characteristics and length of the hospital stay were not substantially different between anti-pseudomonal and non-anti-pseudomonal antibiotics groups. Furthermore, a change in the treatment from non-anti-pseudomonal antibiotics to anti-pseudomonal antibiotics occurred in only 4% of cases. Thus, physicians may not need to administer anti-pseudomonal antibiotics based only on the number of recurrences. Currently, multidrug-resistant organisms are a global concern [31]. The judicious use of broad-spectrum antibiotics can reduce the number of drug-resistant microorganisms [32, 33]. Additionally, the use of broad-spectrum antibiotics is associated with a high cost of care and increased antibiotic-associated side effects, such as Clostridium difficile infection [34]. Therefore, a strategy of empirical narrow-spectrum antibiotic use in patients with recurrent COPD exacerbation may be an acceptable choice.
This multicenter retrospective study has numerous strengths. First, this study was based on daily clinical practice. In contrast to the GOLD 2020 report and a network meta-analysis of randomised controlled trials (RCTs), the antibiotics used in the present cohort were mainly intravenous antibiotics [6, 35]. This may be due to the large number of elderly patients requiring some assistance; the average age of the present cohort was 80 years, and only 175 of 965 (18%) cases were fully independent. The present study results may better reflect hospitalized patients in daily practice rather than selected patients in RCTs [8]. Second, the number of observations was large compared to that in previous RCTs. The number of included patients/observations was much larger than that in a meta-analysis of four RCTs assessing the effect of currently used antibiotics on the length of the hospital stay (965 observations in the present study vs 393) [8]. Furthermore, an ongoing RCT targeting patients with at least one hospitalization within the prior 12 months and the detection of PA in a sputum culture has an expected total sample size of 150 (for the comparison of the anti-pseudomonal antibiotics group and the placebo group) [36], which is far smaller than that in the present study. Thus, the present results may fill gaps not covered by these RCTs.
The main caveat regarding this study is that we only evaluated the length of the hospital stay. Although the length of the hospital stay is associated with the survival of hospitalized patients with COPD, as an outcome, it is short-term and soft. Based on the results of previous systematic reviews, hard outcomes such as in-hospital death and 30-day mortality could not be evaluated in the present study because of its sample size [7, 8]. Moreover, the time to the next exacerbation could not be evaluated because of unmeasured confounding factors, such as post-admission baseline treatments for COPD. In a previous retrospective cohort study with unmeasured confounding factors, antibiotic use was associated with improvements in the long-term mortality and time to the next exacerbation [37]. Further large-scale studies are warranted to evaluate hard short-term and long-term outcomes [38].
The present study had several other limitations. First, although our validation study showed a high predictive value for our patient selection strategy, the number of patients with COPD exacerbation was relatively small, considering that all of the hospitals were acute care and educational hospitals. Patient selection based on the ICD-10 code could have low sensitivity, and there could be many patients who should have been included in this study. In Japan, especially in our hospitals, payment is based on Diagnosis Procedure Combination: a system that reimburses hospitals based on the diagnosis code of hospitalized patients. Patients with a long length of stay or complications during hospitalization might be coded with diagnoses other than COPD exacerbation. Thus, the length of stay may have been right-truncated, and the effect of anti-pseudomonal antibiotics may have been skewed toward the null. Second, the interval between COPD exacerbations was not taken into account, and the variance correlation of the length of the hospital stay within each patient over multiple visits may differ from our expectations. We coped with this issue by using robust variances. Third, there may have been additional confounding factors. For example, because of the substantial amount of missing COPD stage data, we could not use it as a covariate; accordingly, the estimate of the effect of anti-pseudomonal antibiotics might have been skewed toward the null. Although we used other similar baseline covariates, such as the activities of daily living and home oxygen therapy use, an additional confounding factor may have altered the results. Further RCTs are needed to address unknown confounding factors.