Author + information
- Received December 31, 2001
- Revision received April 22, 2002
- Accepted May 20, 2002
- Published online August 21, 2002.
- Elizabeth M Mahoney, ScD*,* (, )
- Trevor D Thompson, BS‡,
- Emir Veledar, PhD*,
- Jovonne Williams, MS* and
- William S Weintraub, MD, FACC*,†
- ↵*Reprint requests and correspondence:
Dr. Elizabeth M. Mahoney, Emory Center for Outcomes Research, Emory West Suite 1N, 1256 Briarcliff Road, Atlanta, GA 30306, USA.
Objectives This study evaluated the cost-effectiveness of administering prophylactic intravenous (IV) amiodarone therapy to patients undergoing cardiac surgery according to their predicted risk of postoperative atrial fibrillation.
Background Atrial fibrillation (AF) is a common complication of cardiovascular surgery that is associated with a significant increase in hospitalization costs. Intravenous amiodarone has been shown to decrease the incidence of postoperative AF.
Methods All 8,709 patients who underwent coronary artery bypass grafting (CABG), 1,217 patients who underwent valve replacement and 624 patients who underwent CABG and valve replacement procedures (CABG + valve) from January 1, 1994, to June 30, 1999, at Emory University Hospitals were studied. Models predicting the risk of AF were developed using logistic regression; linear regression was used to estimate the influence of AF on hospitalization costs. Cost-effectiveness was evaluated for patient subsets identified according to their predicted risk of AF.
Results Postoperative AF rates were 17.7% for CABG, 24.6% for valve and 33.8% for CABG + valve. Using $5,000 as an acceptable cost per episode of atrial fibrillation averted, prophylactic IV amiodarone in CABG patients was not found to be cost-effective. Therapy would be recommended for roughly 5% of valve patients with a predicted risk of atrial fibrillation >45%, and roughly two thirds of CABG + valve patients who have a predicted risk of >30%.
Conclusions Cost-effectiveness of prophylactic IV amiodarone varies according to type of surgery and the predicted risk of atrial fibrillation. Older patients undergoing valve replacement, particularly those with a history of chronic obstructive pulmonary disease, and those undergoing concomitant CABG are likely to be the most appropriate candidates for IV amiodarone therapy in the perioperative period.
Atrial fibrillation (AF) occurs frequently after cardiac surgery, with reported incidences ranging from 20% to 50% of cases (1–7). Several strategies have been used to decrease the risk of this complication. The use of an oral amiodarone chlorhydrate (Cordarone) load before surgery has been shown to decrease the postoperative risk of AF (8). However, the difficulties associated with rapidly loading patients orally have raised the question of whether this dosing could be accomplished more readily with an intravenous (IV) load. This question was addressed in the Amiodarone Reduction in Coronary Heart (ARCH) trial, which showed that IV amiodarone could reduce the incidence of AF by 26% (9). The question remaining after the ARCH study is whether it is reasonable to give IV amiodarone to all patients having cardiac surgery or whether a more cost-effective approach would be to target this therapy to patients at or above a certain level of predicted risk of AF. The purpose of this study was to evaluate the cost-effectiveness of targeting IV amiodarone therapy to patients at varying levels of risk.
Data from 10,550 patients who underwent coronary artery bypass grafting (CABG) (n = 8,709), valve replacement (n = 1,217) and combined CABG and valve replacement procedures (CABG + valve) (n = 624) at Emory University Hospitals between January 1, 1994 and June 30, 1999 were used to develop two models for each patient population: 1) a model to predict the occurrence of AF and 2) a model to estimate the influence of AF on hospitalization costs. In addition, data from 25,975 patients who underwent coronary or valve surgery between 1980 and 1995 were used to estimate the influence of postoperative atrial fibrillation on five-year survival. Patients with a history of AF were excluded from the study.
Emergent procedure: performed in the setting of hemodynamic instability, acute ischemia or infarction. Defined by patient history: hypertension, diabetes, severity of angina, previous myocardial infarction (MI). Angina was categorized according to the Canadian Cardiovascular Society Classification (10). Congestive heart failure was categorized according to the New York Heart Association criteria (11).
Data collection and statistical analysis
This study is a retrospective analysis of prospectively collected data. Baseline demographic, clinical, angiographic and surgical characteristics, including complications, were recorded prospectively on standardized forms and entered into a computerized database. Similarly, follow-up outcome data were collected at one year and every five years postprocedure and during all cardiac rehospitalizations at Emory University Hospitals and entered into the same computerized database.
Development of the models to predict AF
Multivariable logistic regression was used to examine the influence of a variety of demographic, clinical, angiographic and surgical characteristics on the development of postoperative AF. Separate models were fit for each group (CABG, valve replacement, CABG + valve). Multiple imputation, implemented in S-plus, was used to impute missing covariate values. Potential nonlinear effects of each of the continuous predictor variables (age, ejection fraction, cross-clamp time, bypass time) on AF were checked using restricted cubic spline functions (piecewise polynomials). The ability of the logistic regression models to discriminate among patients with respect to their outcomes was measured using the c-index, equivalent to the area under the receiver-operating characteristic curve. The models were validated internally using bootstrap analysis (150 bootstrap samples).
Cost and cost-effectiveness analysis
Hospital charges were obtained from the UB-92 formulation of the hospital bill. Charges were reduced to costs using departmental cost-to-charge ratios. Professional charges were obtained from Current Procedural Terminology codes, which were converted to relative value units (RVUs) using the Resource Based Relative Value Scale (12); RVUs were summed and converted to dollars using the Medicare conversion factor. Hospital and professional costs were added together yielding total hospitalization costs.
Linear regression was used to estimate the influence of AF on total hospitalization costs after adjusting for the effects of significant clinical and demographic factors. In addition to an indicator for the development of postoperative AF, candidate predictors in these models included demographic, clinical and procedural characteristics
Decision analytic models (13) were developed to evaluate the cost-effectiveness of targeted IV amiodarone therapy for each of the three types of heart surgery patients based on in-hospital cost and event rate data and corresponding models previously described, published efficacy of IV amiodarone (26% reduction in atrial fibrillation rate) from the ARCH trial (9) and an assumed cost of amiodarone of $973 (1 g per day for two days). Results from the AF and cost-prediction models were used to provide patient-level inputs into the decision model used to evaluate cost-effectiveness. The model was specified by a set of definitions and equations that are available upon request. Sensitivity analyses were performed to assess the influence of varying the cost of therapy, efficacy of therapy and the cut-off probability of AF for initiating therapy on the cost-effectiveness ratio.
Cost-effectiveness was also estimated, taking a longer-term perspective in terms of cost per quality-adjusted life year gained. This involved making assumptions about the effect of postoperative AF on future costs, appropriate discount rate, utility and effect of events on utility, as well as the impact of the prevention of postoperative AF on future survival. Five-year survival in patients with and without postoperative AF was estimated using Cox proportional hazards regression (14), adjusting for the mean (continuous) and mode (categorical) of all covariates included in the model. Separate analyses were performed on data for 21,349 CABG patients, 3,275 valve replacement patients and 1,351 CABG + valve patients who underwent surgery between 1980 and 1995.
Clinical and angiographic characteristics of the patient populations are presented in Table 1. Valve replacement patients were the youngest, whereas CABG + valve patients were the oldest. The percentage of women was highest for valve surgery and lowest for isolated CABG. Previous MI was common in patients undergoing CABG. Hypertension was common overall although more common in CABG patients. Diabetes was most common in patients undergoing isolated CABG and least common in isolated valve replacement patients. Most of the cases were elective. Comorbidity of previous stroke, chronic obstructive pulmonary disease and renal insufficiency were noted in a minority of patients. Fewer than one in four patients was a current smoker. Ejection fraction was in the low-normal range on average. There were considerably more severe coronary obstructions in the CABG patients.
Mortality ranged from 2.6% in isolated CABG patients to 11.7% in CABG + valve patients. Atrial fibrillation occurred in 17.7% of isolated CABG cases, 24.6% of isolated valve cases and in 33.8% of CABG + valve cases. After adjusting for covariates, AF was associated with a 3.4-, 3.3- and 6.4-day increase in postprocedural length of stay for CABG, valve and CABG + valve patients, respectively.
Multivariate predictors of the occurrence of postoperative AF are shown for CABG, valve and CABG + valve patients in Table 2. For CABG patients, variables associated with a significant increase in the risk of AF were increasing age, increasing bypass time, male gender and previous MI. The c-index for this model was 0.68 (the validated c-index was also 0.68), suggesting a moderate ability to predict the occurrence of postoperative AF. Factors associated with a significant increase in the risk of postoperative AF for valve patients were increasing age, chronic obstructive pulmonary disease and smoking status, which had a paradoxical effect whereby current smokers were found to have a significantly lower risk of AF than never smokers. This model had a c-index of 0.67 (validated c-index of 0.66). For CABG + valve patients, only two factors, increasing age and renal failure, were independently associated with an increased risk of postoperative atrial fibrillation. This model had a c-index of 0.65 (validated c-index of 0.64).
For each of the populations, the linear regression models revealed that AF was associated with a significant increase in costs though the estimated magnitude varied. For CABG patients, AF was associated with an increase of $3,783 ± 310, for valve patients, AF was associated with a $2,857 ± 932 increase and the estimated increase in cost for CABG + valve patients was $4,536 ± 1441.
Tables 3, 4 and 5⇓⇓⇓present results of cost-effectiveness analyses for CABG, valve replacement and CABG + valve patients, respectively. The columns to the far left in each table present the treatment threshold in terms of the predicted probability of AF above which therapy with amiodarone would be given. For CABG patients (Table 3), assuming a 0% threshold, 100% of the patients would be treated (column 2), and the event rate would drop from 17.7% (last row of column 3) to 13.1% (first row of column 3). The mean in-hospital cost per patient would also change, from $23,296 (last row of column 4) to $24,443 (first row of column 4). The average cost per episode of AF averted (column 5) for the strategy of treating all patients versus no patients is derived by dividing the difference in average costs of the two strategies by the difference in the two AF rates, i.e. ($24,443 − $23,296) / (17.7 − 13.1) = $24,934. The incremental cost per episode of AF averted for a strategy of treating patients at one cutoff level versus the next highest cutoff level is derived in an analogous manner. Column 6 shows the incremental (or marginal) cost per episode averted. Progressing from the bottom of column 6 towards the top, entries in column 6 describe the additional costs associated with preventing an event for patients in the next lowest level of risk. For example, the incremental cost per event averted moving from a 30% to a 25% probability of AF threshold for treatment is $17,479.
There is a great degree of variability in cost-effectiveness as the probability of AF in the targeted population varies. As the threshold probability of developing AF above which patients will receive therapy is increased, the risk of AF in the targeted patients rises, while the cost-effectiveness ratio falls. Because of the competition for limited healthcare resources, the consideration of additional healthcare expenditures should be based on the evaluation of whether the additional expense is worthwhile given the benefits accrued; therefore, it is the incremental cost-effectiveness ratios that must be evaluated with respect to their acceptability. The incremental cost per episode of AF averted for CABG patients ranged from $10,938 for the treatment of the highest-risk patients to $55,854 for the treatment of the lowest-risk patients. For valve replacement patients, the incremental cost per episode of AF prevented ranged from $4,219 for the highest-risk patients to $43,011 for the lowest-risk patients. For CABG + valve patients, the incremental cost per episode of AF prevented ranged from $69 to $39,698.
A two-way sensitivity analysis on cost of therapy and effectiveness of therapy for valve replacement placements is presented in Figure 1 . The marginal cost effectiveness for treating patients with a 25% versus a 30% predicted probability of AF is shown, as well as isobars of cost from $500 to $1,250 in $250 increments. The x-axis represents the effectiveness of therapy from 0.20 to 0.60, and the y-axis represents the cost-effectiveness ratio in dollars per episode of AF prevented. The cost-effectiveness ratio rises as the cost of therapy rises. The ratio falls as therapy becomes more effective.
A two-way sensitivity analysis on cost of therapy and the probability cutoff for treatment for valve replacement patients is presented in Figure 2. Therapy is assumed to be 26% effective in preventing episodes of AF. On the x-axis, the threshold for treatment varies from 0 to 0.50 in terms of predicted risk of AF. Points are computed as the incremental cost-effectiveness at each probability of AF threshold compared to a threshold 2% higher. Again, as the cost of therapy rises, the cost-effectiveness ratio also rises and as the probability cutoff for treatment rises, the cost-effectiveness ratio falls. Thus, a favorable ratio can be achieved by using a high treatment threshold of predicted risk of AF. However, as the threshold is raised, the proportion treated will fall.
For the long-term analysis, Cox regression proportional hazards model revealed a significant increased hazard of death for CABG patients who developed postoperative AF compared to those who did not (hazard ratio: 1.19, 95% confidence interval [CI]: 1.09 to 1.30) after adjusting for other significant factors. For valve replacement and CABG + valve patients, however, no effect of postoperative AF on long-term survival was found. The inability to estimate a survival advantage for patients who do not develop AF renders a long-term cost-effectiveness analysis for valve replacement and combined CABG and valve patients impossible.
For CABG patients, the incremental cost-effectiveness of targeted therapy measured in cost per quality-adjusted life year gained is presented in Table 6. The cases for patients with probability of AF thresholds from 0% to 10% to 45% to 50% were considered, assuming the cost of therapy to be $973. From the 26,247 patients who underwent CABG between 1980 and 1995, the five-year survival was 87.5% (95% CI 86.5% to 88.3%) without an episode of AF and 85.2% (95% CI 83.8% to 86.6%) with an episode of AF (p < 0.0001). Using an annual 3% discount rate, mean discounted life years over five years follow-up, adjusted for the influence of significant covariates using Cox regression, was 4.18 years for patients without AF and 3.92 years for patients with AF. No additional survival benefit of an AF free initial hospitalization was assumed beyond five years. It was assumed that the cost of medical care after discharge would be similar in the patients with and without a postoperative episode of AF, such that any difference in cost would be reflected in the hospital cost. Five models were developed. Utility was assumed to decrease for some of the models by 20% for the first year if there was an episode of AF, which would reduce quality-adjusted survival to 3.74 years; otherwise, the utility was assumed to remain 1.0. Therapy that decreases in-hospital events may or may not decrease future mortality. In the first two models, an event averted was assumed to increase survival to that of a patient without an event. In the last three models, the increase in survival attributed to an event averted was first assumed to be 50%, then 25% and finally 0% of that in models 1 and 2. In the final model, the effect of events averted is only to improve quality of life in the first year. The cost-effectiveness ratio varied widely depending on the assumptions made. For each model, the cost-effectiveness ratio fell as the treatment threshold in terms of predicted probability of AF was raised. The marginal cost-effectiveness comparing a 45% cutoff to a 50% cutoff probability of an event for the best case (model 2) was $47,553 and for the worst case (model 1) $174,089 per quality-adjusted life year (QALY)..
By developing a model to predict the occurrence of AF after cardiac surgery and a second model to estimate the influence of postoperative AF on hospitalization costs in the same patients, we were able to perform an analysis examining the cost-effectiveness of targeting amiodarone therapy to subsets of patients at incrementally lower levels of risk of AF. We performed these analyses separately for CABG, valve and CABG + valve patients owing to their different underlying risk and cost profiles. Because of the inability to estimate an increased hazard of death for valve replacement and CABG + valve patients who developed postprocedural AF, long-term cost-utility analyses could not be performed for these types of surgical patients. This is unfortunate because the short-term estimates in terms of cost per episode of AF prevented are difficult if not impossible to benchmark against other types of therapy. Nevertheless, the information presented in Tables 3 through 6, as well as Figures 1 and 2 can be useful in guiding the decision-making process regarding which patients should receive amiodarone. If an acceptable upper limit of cost per episode of AF averted or QALY gained is known, then all patients for whom the cost of preventing an event is less than this upper limit should receive therapy, given the limitations and assumptions of the models.
Both short-term cost-effectiveness and longer-term cost-utility analyses suggest that the administration of IV amiodarone to CABG patients is not cost-effective. Using $50,000 as the maximum amount that society would be willing to pay per QALY gained, treatment of the small fraction of patients at the very highest level of risk results in cost utility estimates, which are only slightly below $50,000 per QALY under the most favorable assumptions. Estimates of incremental cost per episode of AF averted for the highest risk CABG patients are >$11,000. Although there are no benchmarks for an acceptable cost per episode of AF prevented, $5,000 may seem a reasonable amount in terms of an individual’s willingness to pay to avert an episode of AF. For valve replacement and CABG + valve patients, if society were willing to pay $5,000 per episode of AF prevented, then therapy would be recommended for roughly 5% of valve replacement patients and two thirds of CABG + valve patients, assuming underlying rates of AF equal to those observed for this study. If the incidence of AF is higher than that observed in this study, assuming that the overall covariate effects for the prediction of AF remained the same, the proportion of patients treated at any given predicted probability threshold would increase. An analysis based on modified logistic regression models, for which the overall incidence of AF was adjusted upwards to 30%, 40% and 45% for CABG, valve and combined CABG + valve patients, respectively (by adjusting the intercept while keeping the originally estimated parameters for the covariate effects in the logistic regression models), yielded marginal cost-effectiveness ratios that differed little from those in presented here, although the percent of valve patients who would be treated at a 45% predicted probability of AF threshold increased to 40% and the percent of combined CABG and valve patients who would be treated at a 30% threshold was 87% (results not presented). Nomograms are presented online in the Appendix that may be used, with some caution, to evaluate the probability of developing postoperative AF for populations with both the observed and adjusted (higher) underlying risk of AF. The nomograms may be used to predict risk in individual patients. Accordingly, for a 50-year-old patient undergoing CABG + valve surgery with no history of renal insufficiency, the predicted risk of AF would be 16%. For this patient, amiodarone would probably not be cost-effective. However, for a 60-year old patient with renal insufficiency undergoing CABG + valve surgery, the predicted risk would be 37%. In this patient, amiodarone therapy may be considered cost-effective.
AF is a common occurrence after cardiac surgery, with estimates of incidence ranging from approximately 20% to 50% (1–7). Atrial fibrillation has been shown to be associated with prolonged length of hospital stay in this study as well as in previous studies (1,2,15). In this study we have shown, as have Aranki et al. (1), that AF is associated with increased hospital costs. Although the heart rate of AF can generally be easily controlled pharmacologically and AF can be converted to sinus rhythm both pharmacologically as well as by electrical cardioversion, AF tends to recur. Furthermore, AF may be complicated by thromboembolic events. Thus, there would seem to be a role for effective prophylactic therapy. This was examined in a randomized trial of oral amiodarone by Daoud et al. (8). These investigators reported a decrease in the incidence of AF after coronary surgery from 32 of 64 (50%) in the control group to 16 of 64 (25%) in the group who received several days of oral amiodarone pre-operatively. Because of the logistical difficulty of administering amiodarone orally over several days, IV amiodarone is a more practical alternative. The efficacy of IV amiodarone in preventing postoperative AF was studied in the ARCH trial (9). Patients were treated with 1 g per day of IV amiodarone for two days beginning within 3 h of entering the surgical intensive care unit. Atrial fibrillation was noted in 67 of 142 (47%) patients on placebo versus 56 of 158 (35%) on amiodarone (p = 0.01), a 26% reduction in risk. Because amiodarone was clearly effective in the ARCH trial, the important question of whether all patients should be treated or whether therapy should be targeted to patients at higher risk was addressed in the present cost-effectiveness analysis.
Although this study takes a societal perspective overall, hospital costs had a hospital perspective and professional costs a payer perspective, which were used as proxies for societal costs. The cost-effectiveness ratio should be from a societal perspective because it is the responsibility of society as a whole to prevent cardiovascular events because society would have to pay the cost of these events (16). However, the hospital would have to pay the cost of the drug and, depending on reimbursement scheme, may derive little or no benefit from events prevented (17).
This study was limited by the moderate ability of the logistic regression models to predict the occurrence of postoperative AF (c-indexes 0.65–0.68). An improved ability to identify high-risk patients would result in more favorable cost-effectiveness ratios. This study is also limited by the assumption that costs beyond the initial hospitalization are equivalent in patients who do and do not develop postoperative AF. If AF also increases costs postdischarge, the cost-effectiveness of any prophylactic treatment would increase.
The extension of the in-hospital cost-effectiveness analysis to a long-term cost-utility analysis involved multiple additional assumptions, some of which are not verifiable. The cost-effectiveness ratio may be quite sensitive to these assumptions, as Table 6 suggests. However, with cost-effectiveness expressed in cost per quality-adjusted life year gained, it is possible to benchmark the cost-effectiveness of this therapy against other forms of therapy; this is not possible when measuring cost-effectiveness in terms of cost per episode of AF prevented.
On the other hand, the cost-effectiveness ratios expressed in cost per event averted were highly data driven, and event rates and costs (exclusive of drug costs) were based on actual data. Drug efficacy was based on the literature (8,9). Drug efficacy and costs were the two primary variables examined by sensitivity analysis.
Another potential limitation is the assumption that the efficacy of the drug is independent of the probability of an event. Perhaps in low-risk patients it prevents a lower or a higher proportion of events. Two additional assumptions underlying this cost-effectiveness model are that 1) the cost of an episode of AF in treated patients is equal to the cost of an episode of AF in untreated patients and 2) the cost of AF in high-risk patients is equal to that in low-risk patients.
Through the examination of patient level data and creating a risk profile, the ability to examine cost-effectiveness according to predicted level of risk has been demonstrated. Although a similar approach has been used previously in a study of lipid lowering (18) and the use of platelet antagonists in the setting of high-risk percutaneous coronary intervention (19), this is the first study to examine the cost-effectiveness of targeting patients undergoing cardiac surgery for IV amiodarone treatment according to their predicted level of risk using inputs derived from an empirical data set. One other study reported advantages of a selective approach for the prevention of AF after coronary surgery using a retrospective analysis of data from a clinical trial comparing sotalol versus placebo, although a formal cost-effectiveness analysis was not conducted (20).
Although this study was based on a single center experience, the population resembled that seen in clinical trials and registry studies (1–9). This study has shown quantitatively and in a real patient population what is often stated: As the level risk rises, therapy to decrease events can become more cost-effective. From a societal or policy standpoint, studies such as this may help in the development of informed guidelines for the use of expensive therapies. Although application of these results to decision making for individual patients must be undertaken with some caution, this study shows that for low-risk patients, IV prophylactic amiodarone probably is not cost-effective, whereas for higher risk patients, particularly older patients undergoing valve surgery or patients undergoing CABG + valve surgery, perioperative IV amiodarone may be considered cost-effective in preventing AF.
To view the nomograms, please see the August 21, 2002 issue of JACC at http://www.cardiosource.com/jacc.html.
☆ This work was supported by a grant from Wyeth-Ayerst Laboratories.
Presented in part at the American College of Cardiology 51st Annual Scientific Sessions, March 2002.
- atrial fibrillation
- Amiodarone Reduction in Coronary Heart trial
- coronary artery bypass grafting
- CABG + valve
- combined coronary artery bypass graft and valve replacement
- myocardial infarction
- quality-adjusted life year
- relative value unit
- valve replacement
- Received December 31, 2001.
- Revision received April 22, 2002.
- Accepted May 20, 2002.
- American College of Cardiology Foundation
- Aranki S.F.,
- Shaw D.P.,
- Adams D.H.,
- et al.
- Caretta Q.,
- Mercanti C.A.,
- DeNardo D.,
- et al.
- Guarnieri T.,
- Nolan S.,
- Gottlieb S.O.,
- Dudek A.,
- Lowry D.R.
- Campeau L.
- The Criteria Committee of the New York Heart Association
- Becker E.R.,
- Mauldin P.D.,
- Bernadino M.E.
- Weinstein W.C.,
- Cox D.R.
- Tamis J.E.,
- Steinberg J.S.
- Gold M.R.,
- Siegel J.E.,
- Russell L.B.,
- Weinstein M.C.
- Weintraub W.S.,
- Thompson T.D.,
- Culler S.,
- et al.
- Kloter Weber U.,
- Osswald M.H.,
- Buser P.,
- et al.