Association of Corticosteroid Dose and Route of Administration With Risk of Treatment Failure in Acute Exacerbation of Chronic Obstructive Pulmonary Disease FREE

Chronic obstructive pulmonary disease (COPD) affects more than 6% of adults in the United States, accounts for $32 billion in direct health care costs, and is the fourth leading cause of death.^1- 2 In 2006, there were approximately 600 000 hospital admissions for acute exacerbation COPD, making this 1 of the 10 leading causes of hospitalization nationwide.³

The mainstays of hospital care for patients with acute exacerbation of COPD include the provision of supplemental oxygen, short-acting bronchodilators, systemic corticosteroids, and usually antibiotics. Evidence supporting treatment with steroids is derived from a small number of randomized trials in which steroids were found to improve physiological measures of lung function, to reduce the risk of treatment failure, and to decrease length of stay in the hospital.^4- 9 Although the benefit of steroids appears clear, the optimal dose and route of administration of these agents remain uncertain.^10- 11 Based on their high bioavailability¹² and extrapolating from studies investigating asthma and other inflammatory conditions, clinical guidelines produced by leading professional societies in the United States, the United Kingdom, and Europe recommend treatment with low doses of steroids administered orally.^13- 15

We investigated the use of corticosteroids among patients hospitalized for acute exacerbation of COPD in a large network of US hospitals and compared the outcomes of those initially treated with low doses of steroids administered orally to those in whom steroids were initially administered at higher doses intravenously.

We conducted a retrospective cohort study using data from 414 hospitals that participate in Premier Incorporated's Perspective, a voluntary, fee-supported database developed for measuring quality and health care utilization. Participating hospitals are geographically diverse and similar to the composition of acute care hospitals nationwide and are predominantly small to midsize nonteaching facilities that serve a largely urban patient population. In addition to the information available in the standard hospital discharge file, Perspective contains a date-stamped log of all billed items, including medications, laboratory, diagnostic and therapeutic services, at the individual patient level. Three-quarters of hospitals that participate report actual costs derived from internal cost accounting systems, whereas others provide cost estimates calculated using Medicare cost to charge ratios. In 2006, the database contained the records of approximately 5.5 million hospital discharges, representing 15% of all hospitalizations nationwide.

Patients admitted to participating hospitals between January 1, 2006, and December 1, 2007, were included in our analysis if they were aged 40 years or older, had a principal diagnosis of COPD with acute exacerbation (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 491.21, 491.22, 493.22), emphysema (ICD-9-CM, 492.8), or if they had a principal diagnosis of respiratory failure (ICD-9-CM codes 518.81, 518.82, 518.84) combined with a secondary diagnosis of COPD with acute exacerbation or emphysema,¹⁶ and were treated with systemic corticosteroids during the first 2 days of hospitalization (Figure 1).

We excluded patients whose initial daily dose of corticosteroids fell outside a conventional range of treatment. This included patients who received less than 20 mg or more than 80 mg of oral prednisone and those treated intravenously at doses lower than 120 mg or higher than 800 mg of prednisone equivalents each day. Because mechanical ventilation initiated after hospital day 2 was one of our outcome measures, we excluded patients admitted directly to the intensive care unit (ICU). In addition, to focus on patients with less complicated acute exacerbation of COPD, we excluded those with a secondary diagnosis of pneumonia or pulmonary embolism. To reduce the likelihood that coding errors might be responsible for incorrectly classifying patients as having acute exacerbation of COPD, we further limited the analysis to patients whose attending physician was a family physician, general internist, hospitalist, pulmonologist, or intensivist as well as those assigned to an All Patient Refined-Diagnosis Related Group (APR-DRG version 15.0, 3M Corp, Minneapolis, Minnesota) consistent with COPD or respiratory failure. Finally, we excluded patients whose hospitalization lasted only 1 day and those who were transferred to or from another acute care facility because we could not accurately assess treatment with corticosteroid therapy. Permission to conduct the study was obtained by the institutional review board at Baystate Medical Center, where the study was conducted, and the need for written informed consent was waived.

In addition to patient age, sex, race/ethnicity, and insurance status, we recorded the presence of up to 30 unique comorbidities using software provided by the Healthcare Costs and Utilization Project of the Agency for Healthcare Research and Quality based on methods described by Elixhauser et al.¹⁷ Information about race/ethnicity was recorded by admission or triage staff of participating hospitals using hospital-defined options. Furthermore, in an effort to better characterize the underlying severity of COPD, we calculated the total number of admissions for acute exacerbation of COPD in the year prior to the index hospitalization, and used ICD-9-CM secondary diagnosis codes to identify the presence of chronic pulmonary heart disease. For each hospital that participated in the study, we recorded bed count, teaching status, geographic region, and whether it served an urban or rural population.

We reviewed detailed pharmacy charges to determine both the route and dose of systemic corticosteroids administered during the first 2 days of hospital admission and to assess changes in steroid treatment that occurred throughout the hospitalization. Patients were categorized in the high-dose intravenous therapy group if their first recorded dose of corticosteroids was an intravenous preparation and if their maximum dose on day 1 or 2 was within the 120-mg to 800-mg range of the equivalent of prednisone, even if, as would be expected, their treatment was later transitioned to oral therapy.¹⁸ Patients initially treated with a maximum day 1 or 2 dose of oral cortiscosteroids between 20 mg and 80 mg of prednisone were categorized in the low-dose oral therapy group, even if later during their hospitalization steroid therapy was intensified. The first 2 days were initiated for patient categorization because we wanted to focus on patients whose steroids were initiated at or near the time of admission and because in administrative data sets, the duration of the first hospital day includes partial days that can vary in length. Furthermore, we recorded the total duration of steroid treatment administered during the hospitalization and noted any changes in therapy.

In addition to corticosteroids, we assessed many other treatments and tests administered during the first 2 hospital days, including anticholinergic bronchodilators, short- and long-acting β₂-agonists, antibiotics, and noninvasive positive pressure ventilation. Although categorized as being of uncertain benefit in most clinical guidelines, we also assessed the use of methylxanthine bronchodilators, mucolytic agents, chest physiotherapy, sputum studies, and spirometry. Furthermore, as indirect measures of symptom severity, we recorded use of arterial blood gas testing and treatment with loop diuretics and morphine.

Adapting an approach used in an earlier randomized trial of steroids in acute exacerbation of COPD, our primary outcome was a composite measure of treatment failure, defined as the initiation of mechanical ventilation after the second hospital day, death during the hospitalization, or readmission for COPD within 30 days of discharge.⁷ Secondary outcomes included hospital length of stay and hospital costs.

Summary statistics were constructed using frequencies and proportions for categorical data, and means, medians, and interquartile ranges for continuous variables. We compared the characteristics of patients who initially received high-dose intravenous therapy with those in whom steroids were administered at low doses orally. χ², z, or Kruskal-Wallis tests were used to assess the relationship between choice of therapy and the initiation of mechanical ventilation after the second hospital day, in-hospital mortality, 30-day readmission, the composite measure of treatment failure, length of stay, hospital costs, and any potential confounders.

We developed a series of multivariable regression models to assess the independent effect of initial choice of steroid administration on the composite measure of treatment failure, length of stay, and costs while adjusting for principal diagnosis, all patient and hospital characteristics, and other early diagnostic tests and treatments. Length of stay and cost were trimmed at 3 standard deviations above the mean, and the natural log-transformed values were modeled due to extreme positive skew. Generalized estimating equations (SAS PROC GENMOD; SAS Institute Inc, Cary, North Carolina) were used to account for the clustering of patients within hospitals. Logit-link models were used for treatment failure and identity-link models and for the log-transformed length of stay and cost.

We developed a nonparsimonious regression model for initial steroid therapy, using GEE modeling with a logit-link (SAS PROC GENMOD) to produce a propensity score for initial treatment with low-dose oral steroids. The model included all patient and hospital characteristics, attending specialty, all other early treatments and diagnostic tests, and selected interaction terms. The model produced a propensity score for each patient reflecting his/her probability of initial treatment with low doses of steroids administered orally. This score was then included as a covariate in the previous models for treatment failure, length of stay and costs. Then, using a greedy 5 to 1 digit match algorithm, we matched each patient who received initial low-dose oral therapy with a patient in the high-dose intravenous group with a similar propensity for oral treatment.¹⁹ The matched cohort was evaluated for differences for each of the potential confounding factors, and conditional logistic regression was used to assess the association between choice of steroid use and treatment failure, while adjusting for any remaining differences between groups (P < .05).

To address the threat of residual confounding by indication, in which sicker patients might preferentially be given higher-dose intravenous treatment, we developed a model combining an ecologic, or grouped treatment variable, with individual-level covariates and outcomes.²⁰ This is an adaptation of the instrumental variable approach, a technique with an extensive track record in econometrics and growing use in health care.^21- 22 Using this approach, each patient was assigned a probability of treatment with low-dose oral steroids that was equivalent to the rate of low-dose oral therapy at the hospital where they received care—the group rate. We then repeated the GEE model for treatment failure, substituting the hospital rate of treatment in place of individual treatment. This approach attempts to answer the question, “Is treatment at an institution where low-dose oral steroids are used more frequently associated with better or worse outcomes, regardless of how a particular patient is treated?” Grouping treatment at the hospital level in this way has the advantage of largely bypassing the issue of confounding by indication at the patient level, yet in contrast to a traditional ecologic analysis is able to account for patient-level covariates and outcomes.

As a sensitivity analysis to assess the effect of modeling structure, we repeated the final models in a hierarchical structure with SAS PROC GLIMMIX including a random hospital effect.

All significance tests were 2-sided, with a .05 significance level. The study was powered to detect an odds ratio (OR) of 1.2 or larger. With approximately 80 000 cases, 92% in 1 treatment group and adjusting for clustering within hospitals, we had more than 90% power to detect a treatment failure difference from 10% to 11.8%. All analyses were carried out using SAS software version 9.1.

Of 79 985 patients who met our enrollment criteria (Figure 1), 71 628 (90%) had a principal diagnosis of COPD and 8357 (10%) had a principal diagnosis of respiratory failure. The median age was 69 years, 61% were women, 73% were white, and Medicare was the most common form of health insurance (Table 1). Hypertension, diabetes, heart failure, and depression were the most frequently recorded comorbidities, and chronic pulmonary heart disease was present in 8% of patients. Overall, 17% had 1 admission for COPD in the year before the index hospitalization, and 12% had 2 or more. Sixty-six percent of patients were cared for at nonteaching facilities, 19% at hospitals with fewer than 201 beds, and 27% at hospitals operating more than 500 beds. Eighty percent were admitted directly from the emergency department, and general internists and family physicians were the attending of record for 86% of cases (Table 2). The median length of stay was 4 days, median costs were $5021, and 3 out of 10 patients were hospitalized for 6 days or longer. A total of 941 patients (1.2%) had mechanical ventilation initiated after the second hospital day, 1080 (1.4%) died during the hospitalization, and 6911 (9%) were readmitted for COPD within 30 days of discharge. Overall, 8671 patients (11%) experienced the composite measure of treatment failure.

In total, 73 765 patients (92%) were initially treated with high doses of steroids administered intravenously, while 6220 (8%) began low doses of steroids given orally. Expressed in prednisone equivalents, the median total dose administered during the first 2 hospital days was 600 mg in the high-dose intravenous group, and 60 mg among those given lower-dose steroids orally. The 25th to 75th percentile day 1 and 2 dose ranges in the 2 groups of patients was 350 mg to 781 mg in the high-dose intravenous group and 40 mg to 120 mg in the low-dose oral group. Across the 388 hospitals that contributed 20 or more patients to the study, the use of oral steroids as an initial treatment strategy varied modestly. The median rate of initial low-dose oral steroid treatment was 5%, rates varied from 3% to 9% across the 25th to 75th percentile, and 1% of hospitals started half or more of patients with low doses of steroids orally (Figure 2).

When compared with those in the high-dose intravenous group, patients treated with low-dose oral steroids were marginally older, included a lower proportion of white patients, and were less likely to have private insurance (Table 2). Patients treated with low doses of orally administered steroids had a greater number of comorbidities, including diabetes, heart failure, anemia, and renal failure. When compared with those initially treated intravenously, they were less likely to receive early treatment with antibiotics, methylxanthine bronchodilators, to undergo arterial blood gas analysis, and to receive noninvasive ventilation in the first 2 hospital days (Table 2). Treatment with low-dose orally administered steroids was more common in the Northeast, at larger hospitals, and those with teaching programs. A total of 1.4% (95% confidence interval [CI], 1.3%-1.5%) of patients initially treated with intravenous steroids died during the hospitalization and 10.9% (95% CI, 10.7%-11.1%) experienced the composite treatment failure outcome, whereras 1.0% (95% CI, 0.7%-1.2%) of orally treated patients died during the hospitalization and 10.3% (95% CI, 9.5%-11.0%) experienced the composite outcome. A total of 1356 patients (22%) initially treated with low-dose oral steroids were later switched to intravenous therapy. More than 92% of patients received treatment throughout their entire hospital stay, including 90% receiving low-dose and 92% receiving high-dose treatment intravenously.

Overall, 99.7% of patients treated with low-dose steroids administered orally were successfully matched to a patient with a similar propensity who was treated with high-dose steroids administered intravenously (Table 3 and Table 4). In this propensity-matched cohort, the majority of covariates were well balanced, and after conditional regression was applied to control for remaining differences between the groups, patients treated with oral steroids had a lower risk of treatment failure (OR, 0.84; 95% CI, 0.75-0.95), shorter length of stay (ratio, 0.90; 95% CI, 0 0.88-0.91), and lower hospital costs (ratio, 0.91; 95% CI, 0.89-0.93; Table 5).

In models that adjusted for patient, hospital, and physician characteristics, including the propensity for treatment with low doses of steroids given orally, the early use of other treatments and diagnostic tests, and selected interaction terms, the risk of treatment failure among patients given low doses of steroids orally was not significantly different from those treated with high-dose steroids intravenously (OR, 0.93; 95% CI, 0.84-1.02). Patients treated with low doses of steroids administered orally had shorter lengths of stay (ratio, 0.92; 95% CI, 0.91-0.94) and lower costs (ratio, 0.93; 95% CI, 0.91-0.94; Table 5).

Finally, when individual patients were assigned a probability of initial treatment with low-dose steroids given orally equal to the hospital rate where they received care, each 10% increase in the hospital rate of low-dose oral steroids was associated with no change in the odds of treatment failure (OR, 1.00; 95% CI, 0.97-1.03; Table 5).

Estimates of treatment effect for oral vs intravenous steroids from hierarchical models were virtually identical to those obtained from GEE modeling.

In this large observational study, we found that, in sharp contrast to the recommendations contained in leading clinical guidelines,^13- 15 the vast majority of patients hospitalized for acute exacerbation of COPD were initially treated with high doses of corticosteroids administered intravenously. This practice does not appear to be associated with any measurable clinical benefit and at the same time exposes patients to the risks and inconvenience of an intravenous line, potentially unnecessarily high doses of steroids, greater hospital costs, and longer lengths of stay.

Following several small trials conducted over more than a decade, the effectiveness of corticosteroid therapy for patients hospitalized with acute exacerbation of COPD was convincingly established in 1999 through a multicenter trial at the Department of Veterans Affairs (VA) in which a total of 271 patients were randomized to receive placebo or 2 weeks or 8 weeks of steroid treatment.⁷ The protocol called for 125 mg of methylprednisolone to be given intravenously every 6 hours for 72 hours, followed by 60 mg of prednisone once daily orally on days 4 through 7. Investigators found that corticosteroid treatment led to more rapid improvements in lung function, a reduced risk of treatment failure (primarily in the form of decreased need for retreatment with steroids following discharge), and shorter hospital stays, and they found that prolonged steroid tapers had no clinical benefit.

Although direct evidence from head-to-head trials is limited, most clinical guidelines recommend treatment with 20 mg to 60 mg of prednisone once daily administered orally rather than the higher doses of intravenously administered methylprednisolone as described in the VA trial. These recommendations are supported by the high bioavailability of orally administered corticosteroids,¹² small trials demonstrating the efficacy of low-dose oral steroids for treating acute exacerbation of COPD,^6,9 and multiple studies among patients with acute exacerbations of asthma showing equivalency between oral- and intravenous-based strategies.^23- 25 Moreover, when compared with intravenous therapy, oral administration has several advantages, including lower nursing costs, decreased risk of line-related infections and complications, and reduced pain and immobility.^26- 30

When treating acute exacerbation of COPD specifically, the oral and intravenous routes of steroid administration were recently compared in a randomized trial involving 435 patients at a single hospital in the Netherlands.³¹ Investigators there found that 60 mg of prednisone administered daily produced a similar clinical benefit whether administered intravenously or orally; however, patients treated by the intravenous route incurred greater hospital costs. Our findings extend those of De Jong and colleagues³¹ by asking a more pertinent question for physicians practicing in the United States; namely whether higher doses of steroids given intravenously offer any advantage over low-dose oral therapy. Because high-dose intravenous therapy is so common and because patients with COPD are hospitalized frequently for exacerbations, our findings have a significant potential to alter practice, potentially reducing costs, complications, and lowering the risks of steroid-associated adverse events.³²

What factors might explain the large discrepancy between the recommendations found in guidelines and current physician practice? Among a number of possibilities, we suspect that a lack of knowledge of pharmacokinetics, a gut instinct that higher doses are more effective than lower doses, and possibly utilization review programs that require the presence of an intravenous line to justify continued hospitalization at an acute care level are all implicated. In addition, some patients may initially receive high-dose intravenous therapy after the real or perceived failure of low-dose oral therapy started in the ambulatory setting.

The primary limitation of this study is its observational design. Treatment assignment was not random, and physician decisions regarding the choice of high-dose intravenous and low-dose oral therapy may have been influenced by initial illness severity. We found some evidence to support this view; on the one hand, patients in the low-dose oral therapy group were older and had a greater number of comorbidities; however, they were also less likely to undergo early arterial blood gas analysis and less likely to receive early treatment with antibiotics. In addition, although we recorded information about a large number of patient, hospital, physician, and treatment factors, our study was based on highly detailed hospital claims, not direct clinical measurements, and therefore we were not able to adjust for measures of lung function, such as the forced expiratory volume in the first, second, or the partial pressure of carbon dioxide in arterial blood. Furthermore, although we used robust statistical techniques to minimize the threat of selection bias, including propensity adjustment and matching, these approaches are primarily designed to control for imbalances in the distribution of measured confounders. To address our concern about residual biases due to unmeasured factors, we applied a form of instrumental variable analysis in which all patients treated at the same hospital were assigned an identical probability of treatment with low-dose oral steroids equal to the hospital rate of initial oral therapy. This approach is attractive because, although it does not control for differences in the management skills at hospitals that might be associated with the choice of therapy, it largely overcomes the problem of confounding by indication at the patient level. Finally, our analysis was limited to patients initially cared for outside of the intensive care unit, and our results should not be generalized to more severely ill patients.

As an initial treatment strategy for patients with acute exacerbation of COPD, low doses of steroids administered orally were not associated with worse outcomes than high doses of steroids administered intravenously. In light of the greater risks and higher costs associated with high-dose intravenous treatment, opportunities may exist to improve care by promoting greater use of low-dose steroids given orally. Given the large numbers of patients hospitalized with COPD each year in the United States, a clinical trial comparing these 2 approaches to management would be valuable.