Abstract
Background:
We identified the cost of care and clinical events using dipyridamole versus adenosine in pharmacological stress (ST)/single-photon-emission computed tomography (SPECT) myocardial perfusion testing.
Methods:
Commercial health plan members received adenosine or dipyridamole as an adjunct to ST/SPECT testing during the period January 1, 2006, through November 30, 2008. Propensity score matching techniques were used to compare risk-adjusted, test-related complications, symptoms and costs.
Results:
A total of 12,351 patients underwent ST/SPECT testing with dipyridamole and 59,969 with adenosine. Risk-adjusted outcomes analysis showed that patients receiving dipyridamole had a higher number of emergency room (ER) visits (0.65 vs. 0.23%, p < 0.001) and angina pectoris episodes (7.11 vs. 6.01%, p < 0.001). The likelihood of shortness of breath was significantly higher (6.63 vs. 5.77%, p < 0.001) in the adenosine group. One-day risk-adjusted, office-visit, outpatient hospital and other utilization costs for same day ST/SPECT testing were higher for the adenosine group. Risk-adjusted ER visit costs were higher for the dipyridamole group ($1276 vs. $1095, p < 0.001).
Limitations:
First the presence of a claim for a filled prescription does not indicate that the medication was consumed or taken as prescribed. Second, medications filled over-the-counter or provided as samples will not be observed in the claims data. Third, presence of diagnosis codes on medical claims are not positive presence of disease, as diagnosis codes may be incorrectly coded or included as rule-out criteria rather than actual disease. Finally, certain information is not readily available in claims data that could have an effect on study outcomes, such as certain clinical and disease-specific parameters.
Conclusions:
Differences in complications and symptoms may help identify a better-tolerated vasodilator drug (VD) for use in pharmacologic stress testing based on a patient's history and symptoms. Implementation of a data-based strategy for the selection of the most appropriate stress-testing adjunctive agent may be a cost-effective step for institutions and health plans.
Introduction
Myocardial perfusion imaging (MPI) is the most commonly used cardiac nuclear imaging technique. Through the evaluation of myocardial blood flow, this technique is useful to detect myocardial perfusion defects (ischemia or necrosis). The possibility of assessing different aspects including the presence, location, extent, and severity of perfusion defects allows MPI to play an important role in providing both diagnostic and prognostic information in patients with suspected or known coronary artery disease.
Image acquisitions can be obtained either by planar or single-photon emission computed tomography (SPECT) techniquesCitation1. SPECT imaging produces high resolution three-dimensional imagesCitation2 leading to a higher contrast resolution, a better differentiation of myocardial regions, and the possibility of detecting mild perfusion abnormalities.
Myocardial perfusion imaging for the evaluation of ischemic heart disease can be performed both at rest and under stress (exercise or pharmacologic stress). The continued growth and expanding role of stress MPI has encouraged the development of newer pharmacologic stressor agents that are simpler to administer, have an improved side-effect profile, and offer accessibility to patients who currently have contraindications to established agents. Approximately 50% of MPIs in the United States are carried out with pharmacologic stress testing (ST). Pharmacologic SPECT MPI is most commonly performed with coronary vasodilators, such as adenosine and dipyridamole. Following administration of such agents, normal coronary arteries dilate and myocardial blood flow significantly increases. In the region supplied by a stenotic coronary artery, myocardial blood flow does not change or may decrease, resulting in a heterogeneous myocardial blood flow pattern (perfusion defects).
The purpose of this study was to examine costs of care and clinical events that occurred on the same day, or within 1 day, of stress SPECT myocardial perfusion testing with dipyridamole compared with adenosine. In particular, we evaluated the rates of test-related complications (e.g., emergency room (ER) visits, inpatient (IP) events, atrioventricular (AV) block, other arrhythmias, and hypotensive episodes), rates of test-related symptoms (e.g., chest pain, shortness of breath, headache), and health services costs associated with complications or symptoms from dipyridamole versus adenosine treatment on the same day or within 1 day of ST/SPECT testing. We also completed sensitivity analysis by examining the costs of care and clinical events that occurred within 1 week or 1 month of ST/SPECT testing.
Methods and data sources
Data sources and study population
HIPAA-compliant pharmacy and medical administrative claims data from a proprietary US health plan database were used for this study. For 2008, data for approximately 4 million individuals were available.
This study included commercial health plan members who received adenosine or dipyridamole as an adjunct to ST/SPECT testing during the period from 1 January 2006 through 30 November 2008. The date of the latest occurring test was defined as the index date. To gather 6 months pre-index and 1 month post-index information, the overall study period was from 1 July 2005 through 31 December 2008. Costs were adjusted using the annual medical care component of the Consumer Price Index (CPI) to reflect inflation between 2005 and 2008.
To be included in the final study sample, patients also had to meet the following inclusion criteria:
Age 18 and older
Continuous enrollment in their health plan for the 6 months prior to the index date and 30 days following the index date
Each cohort was categorized into four groups depending on the time interval between ST/SPECT MPI with dipyridamole/adenosine and SPECT testing: on the same day, within 1 day, within a week, and within a month.
Statistical analysis
Baseline characteristics were compared between the two patient cohorts, and descriptive statistics were displayed as means (±SD) and percentages. Differences between the cohorts were analyzed using t-testing, Mann–Whitney U-tests and chi-square tests.
Propensity score matching was applied to compare the risk-adjusted outcomes. Propensity score matching is a technique that aims at adjusting for selection bias in non-experimental, non-randomized, and retrospective studies like the present one. By using propensity score matching, each patient in the adenosine cohorts was ‘mirrored’ by a patient with similar pre-defined characteristics in the dipyridamole cohorts. The following characteristics were used to match: year of prescription, age, gender, region, clinical factors such as Deyo–Charlson Comorbidity Index, Elixhauser Index, number of unique medications, medication dispensings, pre-index cardiovascular or coronary artery disease, number of cardiac medications, and total healthcare costs at baseline.
Logistic regression was used to estimate propensity scores. Several interaction variables were attempted but they were not significant. Estimation power of the logistic model is determined with C-statistics. The ProbChoiceTM Algorithm is used for choosing the right matching technique. In particular we used mahalanobis, kernel, one-to-one, and radius matching techniques, and by following the guidelines set forth by Baser, determined that one-to-one matching created the best balance between the groupsCitation3. For more on propensity score matching, please see Stukel et al.Citation4.
Results
After applying the inclusion and exclusion criteria, 72,320 patients were eligible for the analysis. A total of 17% of this group comprised the dipyridamole cohort (n = 12,351), and 83% was in the adenosine cohort (n = 59,969).
As reported in , in both groups of patients received VD treatment, the majority of patients who underwent pharmacologic stress testing were 40–64 years old.
At baseline and before propensity score matching, patients in the dipyridamole group were more likely to be female and older, and more likely to live in the northeastern and western regions of the United States compared with those in the adenosine group. With regard to severity of illness, patients in the dipyridamole group were more ill than those in the adenosine group, as reflected by a higher Deyo–Charlson Comorbidity Index (p = 0.0019). The percentage of patients who had an Elixhauser score higher than 2, and who had a history of cardiovascular/coronary artery disease were slightly higher in the dipyridamole group: however, the minimal difference observed would suggest that a meaningful difference is unlikely. The total number of medications was higher in the adenosine group. Overall, there were no differences in the baseline healthcare expenditures between the two groups.
Since dipyridamole and adenosine groups differed in terms of demographic, socioeconomic and clinical factors, we applied the propensity score matching technique to compare incidence of test-related complications, symptoms and healthcare costs.
Summary statistics of baseline characteristics after propensity score matching are presented in . A total of 12,220 patients in the dipyridamole group were matched successfully with patients in the adenosine group in a 1:1 ratio. The C-score was 0.82, which shows the propensity score matching was able to eliminate the observed bias in the data set. We also used baseline medical cost in the propensity score model to control for possible unobserved bias.
Table 1. Codes used to identify ST, SPECT, adenosine and dipyridamole.
Table 2. Summary statistics of baseline characteristics.
Table 3. Summary statistics of baseline characteristics after propensity score matching.
As depicted in , after risk adjustment, patients in the dipyridamole group were more likely to have an ER visit on the same day of ST/SPECT testing than those in the adenosine group (0.65 vs. 0.23%, p < 0.001). The frequency of other test-related complications such as inpatient events, sinoatrial block, atrioventricular block, other arrhythmias, myocardial infarction, bronchospasm, stroke, hypotension, respiratory distress, and aminophylline administration were similar.
Table 4. Risk adjusted outcomes measures for the same day of ST/SPECT testing.
In terms of test-related symptoms, the percentage of patients with shortness of breath was higher in the adenosine group (6.45 vs. 5.61%, p < 0.006). The likelihood of angina pectoris, however, was more likely in the dipyridamole group (7.11 vs. 6.01%, p < 0.001). Differences in other test-related symptoms (headache, dizziness, atypical chest pain) were statistically insignificant.
Total risk-adjusted cost for the same day of ST/SPECT testing was higher for the adenosine group than the dipyridamole group ($1265 vs. $1163, p < 0.001). Risk-adjusted office visit costs, outpatient hospital costs and other utilization costs were also higher for the adenosine group ($1240 vs. $1096, p < 0.001; $1320 vs. $1271, p = 0.0124; $469 vs. $181, p < 0.005; respectively). When test-related complications associated with dipyridamole versus adenosine were analyzed, there were no significant differences between the two groups, except for ER visits that were significantly higher in dipyridamole patients (0.65 vs. 0.23%, p < 0.001). Risk-adjusted ER costs were also higher for this last group than for the adenosine group ($1276 vs. $1095, p < 0.001). There were no significant differences in inpatient hospital costs between the groups.
The results were similar when we compared the outcomes of patients in the dipyridamole and adenosine groups within one day of ST/SPECT testing: no differences were observed in most of the test-related symptoms, except for the likelihood of shortness of breath, which was higher for the adenosine group, and the frequency of angina pectoris, which was higher again in the dipyridamole group. Risk-adjusted ER visits and costs were higher for the dipyridamole group (0.65 vs. 0.23%, p < 0.001; $1276 vs. $1095, p < 0.001). The overall risk-adjusted medical costs and office visit costs were also higher in the adenosine group whereas the ER costs and outpatient hospital costs were lower.
Discussion
More than 7 million stress perfusion studies are performed in the United States annually, of which 44% are carried out using vasodilator stress agents. Both adenosine and dipyridamole are non-selective coronary vasodilators that are commonly used for stress perfusion imaging. In this report, we examined costs of care and clinical events associated with dipyridamole versus adenosine use in SPECT MPI.
After adjusting for baseline characteristics, both dipyridamole and adenosine had similar profiles regarding overt test-related complications and symptoms except for three variables: ER visits, shortness of breath and angina pectoris. ER Visits was the only test-related complication for which a significant difference was detected, being more frequent in dipyridamole patients. Regarding test-related symptoms, the likelihood of having shortness of breath on the same day or within 1 day of ST/SPECT testing was higher in the adenosine group while the likelihood of having angina pectoris was higher in patients who received dipyridamole. We do not have an accurate explanation for the higher frequency of shortness of breath in patients who received adenosine. However, this difference may be due in part to its mechanism of action (a direct activation of A1, A2b, and A3 receptors expressed in the coronary tree, the peripheral vascular system, and the lungs), the higher frequency of arrhythmias other than ‘paroxysmal ventricular tachycardia’ causing patients to feel ‘short of breath’, or perhaps an ‘anginal equivalent’ as suggested by the episodes of acute myocardial infarction as inpatient admission diagnoses not observed in the dipyridamole group.
In this study, the percentage of patients who experienced side-effects (complications + symptoms) following the administration of the vasodilator agent was lower compared to what has been reported in literature for adenosine (approx. 80%) and dipyridamole (approx. 50%)Citation5–6. It should be apparent, however, that most of the acute or short-term complications or symptoms would not be captured in the claims data. Alternatively, complications occurring immediately after the procedure (the most likely time given in proximity to drug administration) may not be easily captured from claims data either. Therefore, this data could be underestimating the complication/side-effect rates. Another reason could be the relatively younger and healthier patient population studied here: approximately 60% of the sample was younger than 65 years old and approximately 77% did not have a history of reported cardiovascular/coronary artery disease (CV/CAD). After propensity score matching, the percentage of patients with no history of CV/CAD was also approximately 77% in both groups: 77.12% in the dipyridamole group and 77.20% in the adenosine group.
Another interesting aspect analyzed in this study was the frequency distribution of ER admission diagnoses. In both groups, angina pectoris episodes and ‘other chest pain’ account by far for the majority of admission diagnosis. However, two differences can be highlighted. First, the higher frequency of ‘paroxysmal ventricular tachycardia’ as admission diagnosis in the dipyridamole group compared to adenosine, a phenomenon that could be related to the severity of CAD, to a reduced coronary flow reserve, or to the drug itself. The second difference is the higher frequency of ‘shortness of breath’ as an ER admission diagnosis in the adenosine group compared to dipyridamole. While no episodes of acute myocardial infarction as inpatient admission diagnoses were observed within the dipyridamole group, episodes of ‘acute myocardial infarction of anterolateral wall initial episode of care’, ‘acute myocardial infarction of other anterior wall initial episode of care’, and ‘subendocardial infarction initial episode of care’ were observed within the adenosine group, suggesting that some episodes of shortness of breath in this last group could also be episodes of myocardial ischemia presenting as anginal equivalents. According to these data, shortness of breath was not only more frequent in the adenosine group but also could have been of greater intensity to refer more frequently patients to ER.
ER costs were significantly higher in the dipyridamole group. We also note that a significantly higher percentage of ER visits was observed in the dipyridamole group. Different situations could have led to the patient’s referral to the emergency room more frequently, such as the occurrence of more intense or severe episodes, a higher difficulty to manage some of them, the higher percentage of angina pectoris episodes observed in these patients, and the higher frequency of arrhythmias. There could also have been some signs or symptoms that led to ER visits that were not analyzed in this study. Another contributing factor could have been the occurrence of significant ST-T changes on the electrocardiogram (ECG), with or without symptoms, that led to patient’s referral to the ER (a variable not analyzed in this study). It is not possible to accurately determine if the drug itself was responsible for this difference or whether individuals’ factors and/or factors related to the severity of the disease were also implicated.
Total medical costs were significantly higher in the adenosine group ($102 higher), and this could be explained by significantly higher ambulatory and other medical costs. However, it is interesting to point out that this statistically significant difference may be not meaningful in the real world, and, therefore, it might not be worth selecting the agent based on the related medical costs to each agent. Approximately 9 million MPI studies are performed each year in the United States, which would lead to a difference of only $70,000.
Our sensitivity analyses, which measured the outcomes in different time intervals (a week or a month rather than on the same day or within a day), yielded similar results. The difference of the medical cost between the two groups was $70 for the week and $81 for the month.
Limitations
While claims data are extremely valuable for the efficient and effective examination of healthcare outcomes, treatment patterns, healthcare utilization, and costs, claims data are collected for the purpose of payment and not research. Therefore, there are certain limitations associated with the use of claims data. First, the presence of a claim for a filled prescription does not indicate that the medication was actually consumed or that it was taken as prescribed. Second, medications filled over-the-counter or provided as samples by the physician will not be observed in the claims data. Third, presence of a diagnosis code on a medical claim is not positive presence of disease, as the diagnosis code may be incorrectly coded or included as rule-out criteria rather than actual disease. Finally, certain information is not readily available in claims data that could have an effect on study outcomes, such as certain clinical and disease-specific parameters.
It should be noted that neither propensity score matching nor regression adjustment addresses or resolves problems due to imbalances in unmeasured factors. It is quite possible that the outcomes of patients with the same observable characteristics can vary widely because of unobservable factors such as physician or practice prescribing patterns. There exist several methods to control for unmeasured factors such as instrumental variable approach, bounding approach and a difference-in-difference estimator. However, these estimators are also confounded by their own limitations.
Conclusions
The purpose of this project was to examine the risk-adjusted, test-related complications, symptoms and costs in patients who received adenosine or dipyridamole as an adjunct to ST/SPECT testing.
Using a retrospective analysis of health insurance claims data and controlling for confounders such as demographic factors, comorbid conditions and baseline healthcare utility, we showed interesting differences among the variables analyzed. The percentage of patients who experienced side-effects following the administration of dipyridamole or adenosine was lower compared to what has been reported in literature. Both VD agents had similar profiles regarding overt test-related complications and symptoms except for ER visits, shortness of breath, and angina pectoris, and for healthcare costs. The likelihood of shortness of breath, office visit costs and total medical costs were higher in the adenosine group, whereas the likelihood of angina pectoris, ER visits and ER costs was higher in the dipyridamole group. Among ER admission diagnoses, ‘other and unspecified angina pectoris’ and ‘other chest pain’ were the most common in both groups. Paroxysmal ventricular tachycardia was significantly more frequent with dipyridamole, whereas shortness of breath continued being more frequent with adenosine as ER admission diagnosis.
Future research could stratify the analysis by age (<65, ≥65 years), as well as by presence or absence of CV/CAD history, to determine if the lower frequency of test-related symptoms or complications observed in this study were related to a relatively young and healthy population, and if the additional differences observed in ER admission diagnosis were also related to patient’s characteristics or to the drugs.
Further analysis in this area could be useful in selecting the appropriate vasodilator agent based on patient’s medical history and symptoms, in the absence of any specific contraindication, and could have meaningful cost-related implications. Patients receiving dipyridamole presented a small but statistically significant higher rate of angina pectoris episodes and episodes of paroxysmal ventricular tachycardia among ER admission diagnoses, as well as higher rates of ER visits and ER costs. Further analysis may help determine if these higher rates seen with dipyridamole are related to the patient’s medical history and the severity of CAD or to the drug itself, and to see if there is a meaningful basis for choosing between both agents in the real world. If drug-related, choosing the other agent in patients who have never presented those signs and/or symptoms could avoid the request for additional studies.
It would also be valuable to extend the research by controlling not only for observable differences in the claims data, but also for unobservable ones using an instrumental variable approach. One would use co-pay and/or physician or practice-prescribing patterns to control for hidden bias and see how results are sensitive for these factors.
Transparency
Declaration of funding
The study was supported by Astellas Pharma US, Deerfield, IL, USA.
Declaration of financial/other relationships
O.B. was sponsored by i3 Global and M.L.M. is Medical Director of Medical Affairs at i3 Global. H.Y. has no relationship to be declared. J.S. was an employee of Astellas Pharma US by the time the article was written. S.K. is an employee of Astellas Pharma US.
Acknowledgments
We appreciate the thoughtful review of this manuscript by R.G. McAllister, MD, Executive Director of Medical Affairs at i3 Global.
References
- Baser O, Kothari S, Monsalvo ML, et al., eds. Braunwald’s Heart Disease: A Textbook of Cardiovascular Medicine, 8th edn. Philadelphia, PA: Saunders Elsevier, 2008
- Introduction to SPECT. http://www.physics.ubc.ca/∼mirg/home/tutorial/intro.html. Accessed on 11 November 2008
- Baser O. Too much ado about propensity score models? Comparing methods of propensity score matching. Value Health 2006;9:377-85
- Stukel TA, Fisher ES, Wennberg DE, et al. Analysis of observational studies in the presence of treatment selection bias: effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA 2007;297:278
- Henzlova MJ, Cerqueira MD, Hansen CL, et al. American Society of Nuclear Cardiology, imaging guidelines for nuclear cardiology procedures. A report of the American Society of Nuclear Cardiology quality assurance committee: stress protocols and tracers. J Nucl Cardiol 2009
- Johnston DL, Daley JR, Hodge DO, et al. Hemodynamic responses and adverse effects associated with adenosine and dipyridamole pharmacologic stress testing: a comparison in 2,000 patients. Mayo Clin Proc 1995;70:331-6