Embed Size (px)
Citation preview
In February 2013, GlaxoSmithKline (GSK) announced a commitment to further clinical transparency through the public disclosure of GSK Clinical Study Reports (CSRs) on the GSK Clinical Study Register.
The following guiding principles have been applied to the disclosure: Information will be excluded in order to protect the privacy of patients and all namedpersons associated with the study
Patient data listings will be completely removed* to protect patient privacy. Anonymized data from each patient may be made available subject to an approved research proposal. For further information please see the Patient Level Data section of the GSK Clinical Study Register.
Aggregate data will be included; with any direct reference to individual patients excluded
*Complete removal of patient data listings may mean that page numbers are no longer consecutively
numbered
Retrospective Administrative Claims Database Analysis of the Risk of Asthma Exacerbation, Asthma-Related Healthcare Utilization and Costs, and Adherence to Controller Therapy in Asthma Patients Receiving Fluticasone/Salmeterol Inhalation Powder vs. Mometasone Furoate Inhalation Powder
Study Report
Produced by:
USA
Commissioned by: PharmD, MS
GlaxoSmithKline Five Moore Drive Research Triangle Park, NC 27709 USA
Contributors:
USA (Principal Investigator)
USA GlaxoSmithKline, Research Triangle Park, NC, USA
Correspondence MSIA
Phone: Fax: Email:
Source of funding Funding for this research was provided to by GlaxoSmithKline.
Competing Interest has received research funding and consulting fees from GlaxoSmithKline.
Acknowledgements None
Date November 21, 2011
Table of Contents
1 INTRODUCTION .......................................................................................................................1
2 METHODS ................................................................................................................................2
2.1 Overview ......................................................................................................................................... 2
2.2 Data Source .................................................................................................................................... 2
2.3 Study Subjects ................................................................................................................................ 3
2.4 Patient Characteristics .................................................................................................................... 5
2.5 Propensity Score Matching............................................................................................................. 8
2.6 Outcome Measures ........................................................................................................................ 9
2.7 Analyses ........................................................................................................................................ 11
2.7.1 Patient Characteristics .......................................................................................................... 11
2.7.2 Study Outcomes .................................................................................................................... 12
2.7.2.1 Time to Event Outcomes................................................................................................ 12
2.7.2.2 Other Outcomes............................................................................................................. 12
2.7.3 Multivariate Analyses ............................................................................................................ 13
2.7.4 Statistical Significance ........................................................................................................... 13
2.7.5 Statistical Software................................................................................................................ 13
3 RESULTS ................................................................................................................................ 13
3.1 Overview ....................................................................................................................................... 13
3.2 Study Sample ................................................................................................................................ 14
3.3 Patient Characteristics of Un-Matched Patients .......................................................................... 14
3.4 Propensity Score Matching........................................................................................................... 15
3.5 Study Outcomes ........................................................................................................................... 16
i
3.5.1 Time to Event Outcomes ....................................................................................................... 16
3.5.2 Asthma-Related Healthcare Utilization ................................................................................. 17
3.5.3 Asthma-Related Healthcare Costs......................................................................................... 17
3.5.4 Adherence to Controller Therapy ......................................................................................... 18
3.5.5 Multivariate Analyses ............................................................................................................ 18
4 SUMMARY AND CONCLUSIONS .............................................................................................. 18
4.1 Summary ....................................................................................................................................... 18
4.2 Comparisons with Prior Studies ................................................................................................... 19
4.3 Limitations .................................................................................................................................... 21
4.4 Conclusions ................................................................................................................................... 21
5 REFERENCES .......................................................................................................................... 22
6 TABLES AND FIGURES ............................................................................................................. 25
7 APPENDIX A ........................................................................................................................... 43
ii
1 INTRODUCTION
The National Asthma Education and Prevention Program (NAEPP) Expert Panel and the Global Initiative
for Asthma (GINA) both recommend low- to medium-dose inhaled corticosteroids (ICS) as the preferred
initial treatment for patients with mild-to-moderate persistent asthma [1,2]. Results from numerous
studies have shown that compliance with ICS may be improved by reducing the frequency of dosing [3-
7]. Other studies have demonstrated that improved compliance with ICS in turn may result in improved
outcomes and reduced healthcare utilization and costs [8-10].
Asmanex Twisthaler (mometasone furoate inhalation power [MF]) is an ICS that has been approved in
the US for once-daily maintenance treatment of asthma in patients aged 4 years and older; it is available
in two formulations (110 mcg, 220 mcg) [11] [“MF 110, MF 220”]. While MF is the only ICS currently
approved for once-daily dosing, there is evidence that many patients--especially those with mild
asthma--may use other ICS products only once daily (albeit on an off-label basis). There also is evidence
that once-daily dosing of other ICS products may be as effective as twice-daily administration [12]. The
extent to which outcomes differ in typical clinical practice for patients who receive MF once-daily versus
other available treatments for asthma is limited [13, 14].
Advair Diskus is a fixed-dose combination of fluticasone propionate (FP) and salmeterol (SAL)
administered via a specially designed self-actuated inhaler [15]; it is available in three formulations
(100mcg/50mcg [“FSC 100/50”], 250mcg/50mcg [“FSC 250/50”], 500mcg/50mcg [“FSC 500/50”]).
Advair is indicated for twice-daily maintenance treatment of asthma in patients aged 4 years and older.
FSC 100/50 is typically used for patients with less severe disease, who are more likely to be appropriate
candidates for (off-label) once-daily dosing. The objective of this study was to compare the risk of
asthma exacerbation, asthma-related healthcare utilization and costs, and adherence to controller
1
therapy costs in patients with asthma who receive FSC 100/50 versus MF in typical clinical practice,
using health insurance claims data.
2 METHODS
2.1 Overview
The objective of this study was to compare the risk of asthma exacerbation, asthma-related healthcare
utilization and costs, and adherence to controller therapy in asthma patients who receive FSC 100/50 or
MF in typical clinical practice. Using a large health insurance claims database, we identified all patients
with diagnoses of asthma who received either FSC 100/50 or MF. Risk of asthma exacerbation, asthma-
related healthcare utilization and costs, and adherence to controller therapy after the first prescription
for FSC 100/50 or MF were compared using propensity score matching to control for possible
differences in baseline characteristics. Specific measures of interest included the utilization of rescue
medications (short-acting beta agonists [SABAs], systemic corticosteroids [SCS]); asthma-related
emergency department (ED) visits, hospitalizations, and physician office visits; costs of asthma-related
care; and adherence with study therapy (FSC 100/50 or MF).
2.2 Data Source
Data for this study were obtained from the IHCIS National Managed Care Benchmark Database
(provided to by GSK). The database is comprised of information from enrollment files, and medical
and outpatient pharmacy claims from a variety of private health-care benefit plans covering (at the time
this study was initiated) approximately 17 million persons enrolled in over 30 health plans across eight
US census regions. The database contains information on age, sex, US census region, and dates of
claims for all patients. Data on race, socioeconomic status, and anthropometrics are not available. Each
2
medical claim includes the dates of service and International Classification of Diseases 9th Edition,
Clinical Modification (ICD-9-CM) diagnosis codes, and may include revenue codes (e.g., on hospital
claims) or Current Procedural Terminology, Version 4 (CPT-4) and Healthcare Common Procedure
Coding System (HCPCS) procedure codes (e.g., on professional service claims). Each outpatient
pharmacy includes the drug dispensed (in National Drug Code format), the dispensing date, and the
quantity and number of therapy-days dispensed. All claims include an estimated standardized cost,
calculated based on an algorithm designed to yield allowable payments (including member deductibles
and copayments) for healthcare services normalized across health plans, geographic areas, and calendar
time. All claims for an individual patient can be linked using a unique encrypted patient identifier. The
database is fully de-identified and HIPAA (Health Insurance Portability and Accountability Act of 1996)
compliant. Data for this study included claims for the period from January, 2004 through December
2008 (i.e., the most recent five years for which claims data are available) (“study period”).
2.3 Study Subjects
Study subjects were selected by first identifying all those in the database who had one or more medical
claims with a diagnosis (primary or secondary) of asthma (ICD-9-CM 493.XX) during the study period.
From among these persons, we then selected all those who had two or more outpatient pharmacy
claims for FSC 100/50, MF 110, or MF 220 during the study period; patients were stratified into
treatment groups based on the study drug received first. These patients constituted the “base sample”.
The date of the first prescription for FSC 100/50, MF 110, or MF 220 was designated the “index date”.
The 12-month (365-day) period prior to the index date was designated the “pre-index period”. The 90-
day period beginning with the index date was designated the “treatment identification period”. Patients
meeting any of the following criteria were excluded:
3
• No asthma diagnosis on index date or during pre-index period;
• One or more ICS prescription during pre-index period;
• <12 months of complete and uninterrupted claims history prior to index date;
• <3 months of follow-up;
• Any medical claims during study period with a diagnosis of:
o Chronic obstructive pulmonary disease (COPD) (ICD-9-CM CD-9-CM 491,
492, 494, or 496); or
o Respiratory tract cancer (ICD-9-CM 160-164, or 231);
• Any pharmacy claims during the study period for;
o Ipratropium; or
o Tiotropium; or
o Dornase alpha.
• One or more prescriptions during treatment identification period for:
o Leukotriene receptor antagonist (LTRA);
o Inhaled mast-cell stabilizer (IMCS);
o Methylxanthine (MTHL);
o IgE blocker (IgEB); or
o Non-study ICS or LABA (i.e., other than FSC 100/50 or MF).
• For patients with FSC 100/50 on the index date, one or more prescriptions for MF 110 or
MF 220 during the treatment identification period;
• For patients with MF 110 on the index date, on or more prescriptions for FSC100/50 or
MF 220 during the treatment identification period;
4
• For patients with MF 220 on the index date, on or more prescriptions for FSC100/50 or
MF 110 during the treatment identification period;
• Age <12 years or >65 years as of index date;
• Index date prior to January 1, 2005;
• On Medicaid or Medicare;
• Missing or invalid data for demographic or plan characteristics (see below); and
• Missing or invalid data on claims required to calculated pre-index characteristics or
study outcomes (e.g., days supply on pharmacy claims, payments information on
asthma-related claims).
Patients were censored if they had switched their study therapy or had received prescriptions for LTRA,
IMCS, MTHL, IgEB, non-study ICS or LABA after index date, and their censor date was defined as the day
prior to the switch or receipt of prescriptions listed above. The period beginning with the index date
and ending with either the last date for which complete claims data are available or the censor date,
whichever occurs first, was designated the “follow-up period”.
2.4 Patient Characteristics
Demographic and plan information to be captured for each patient included:
• Year of index date;
• Season associated with index date;
• Age (years) at index date;
• Gender;
• Census region (Midwest, New England, South, West); and
5
• Plan type (health maintenance organization [HMO], preferred provider organization [PPO], point
of service plan [POS], other).
Professional service claims were scanned to identify persons with specialist (allergist or pulmonologist)
visits during the pre-index period. Diagnosis codes on medical claims during the pre-index period were
also scanned to ascertain comorbidities of potential interest. The Charlson-Deyo comorbidity index was
calculated for each patient [16].
Procedure codes on professional service claims during the pre-index period were scanned to identify
persons receiving the following asthma and other respiratory-related services:
• Chest x-rays;
• Airflow tests; and
• Oxygen therapy (CPT E0425-E0480).
Pharmacy claims during pre-index period were scanned to ascertain the number of prescriptions for the
following medications:
• SABA;
• Asthma-related SCS (defined below);
• Antibiotics; and
• Other medications.
Copay associated to pharmacy claims for study medications on index date, and mean copay among
pharmacy claims during pre-index period were calculated for each patient.
6
Medical claims during the pre-index period were scanned to ascertain the following measures:
• Number of asthma-related hospitalizations;
• Number of asthma-related ED visits;
• Number of asthma-related physician office and hospital outpatient visits;
• Number of non-asthma-related hospitalizations;
• Number of non-asthma-related ED visits; and
• Number of non-asthma-related physician office and hospital outpatient visits.
Costs reported on medical and outpatient pharmacy claims during the pre-index period were
summarized as following measures:
• Costs of asthma-related hospitalizations;
• Costs of asthma-related ED visits;
• Costs of asthma-related physician office and hospital outpatient visits;
• Costs of asthma-related outpatient prescriptions;
• Costs of other asthma-related services;
• Total cost of asthma-related services;
• Costs of non-asthma-related hospitalizations;
• Costs of non-asthma-related ED visits;
• Costs of non-asthma-related physician office and hospital outpatient visits;
• Costs of non-asthma-related outpatient prescriptions; and
• Costs of other non-asthma-related services;
• Total cost of non-asthma-related services.
7
Hospitalizations were considered asthma-related if the primary discharge diagnosis on the hospital
inpatient claim is for asthma. Because primary and secondary diagnoses cannot be distinguished on
outpatient claims (including claims for ED visits), claims for ED visits and other outpatient services were
considered asthma-related if there was any diagnosis of asthma on the claim and all the other diagnosis
codes on the claim date were respiratory-related or there was one or more asthma-related procedure
code on the claim date. Claims for SCS were considered asthma-related if there was at least one claim
with a diagnosis code for respiratory diseases during the 7-day period ending with the day of the SCS
claim. In calculating asthma related costs, medical claims occurring during inpatient stays
(“confinements”) were classified based on the associated inpatient stay. Outpatient pharmacy claims
for SABAs, ICS, LTRAs, LABAs, MTHL, IMCS, and IgEB were also considered asthma-related.
In calculating baseline patient characteristics, all pharmacy claims other than study medications and all
medical claims occurring on the index date were assumed to have occurred prior to the index
prescriptions for FSC 100/50 or MF and therefore were counted in the pretreatment period.
Costs were estimated using an estimated standardized cost, calculated by I3 Ingenix UHC/IHCIS based on
an algorithm designed to yield allowable payments (including member deductibles and copayments) for
healthcare services normalized across health plans, geographic areas, and calendar time.
2.5 Propensity Score Matching
Patients in the FSC 100/50 group were matched to those in the MF group, either MF 110 or MF 220,
using propensity score matching [17, 18]. Propensity scores were calculated for all members by
8
estimating a logistic regression model with treatment group (MF vs. FSC 100/50) as the dependent
variable, and pretreatment characteristics as independent variables (MF was used as the reference
category, as it was the smallest cohort). Propensity scores for each subject were defined as the
predicted probability (range: 0 - 1) of being in the MF group, conditional on the observed values of the
other characteristics. Matched pairs of MF and FSC 100/50 subjects were identified using nearest
neighbor matching. When patient characteristics were not well matched for treatment groups after
propensity score matching (p value less than .05 for difference between groups, see Section 2.7 below
for detail), these unmatched characteristics were added to the matching criteria one at a time. Starting
with one of the unmatched characteristic, patient matching was repeated until groups were well
balanced on all baseline characteristics. Note that different matched population could result depending
on the first unmatched characteristics added. When more than one well balanced matched populations
were obtained, the largest population was selected. To assess the quality of propensity-score matching,
the pre-index characteristics of matched samples were compared using appropriate statistical tests, as
described in Section 2.7 below. Also, the distribution of propensity scores across groups was arrayed
and compared graphically (i.e., via histogram) to assess the degree of overlap of the propensity score
distributions.
2.6 Outcome Measures
Measures of interest included the following assessed during the follow-up period:
• Time to first asthma-related ED visit;
• Time to first asthma-related hospitalization;
• Time to first asthma-related ED visit or hospitalization; and
9
• Time to first asthma-related exacerbation, defined as an asthma-related ED visit,
asthma-related hospitalization, or receipt of asthma –related SCS.
The other measures of interest included the following assessed over follow-up period:
• Number of SABA prescriptions;
• Number of asthma-related SCS claims;
• Number of asthma-related office/outpatient visits;
• Costs of asthma-related medications;
• Costs of asthma-related ED visits or hospitalizations;
• Other asthma-related costs;
• Total costs of asthma-related care;
• Costs of asthma-related medications excluding FSC 100/50 or MF;
• Total costs of asthma-related care excluding FSC 100/50 or MF;
• MPR for study therapy (FSC 100/50, MF); and
• Refill rate for study therapy.
Hospitalizations were considered asthma-related if the primary discharge diagnosis on the hospital
inpatient claim was for asthma. Because primary and secondary diagnoses cannot be distinguished on
outpatient claims (including claims for ED visits), claims for ED visits and other outpatient services were
considered asthma-related if there was any diagnosis of asthma on the claim and all the other diagnosis
codes on the claim date were respiratory-related or there was a code for one or more asthma-related
procedures on the claim date. Chest x-rays, airflow tests, and oxygen therapy were considered asthma-
related procedures. Claims for SCS were considered asthma-related if there was at least one claim with 10
a diagnosis code for respiratory diseases during the 7-day period ending with the day of the SCS claim.
In calculating asthma related costs, medical claims occurring during inpatient stays (“confinements”)
were classified based on the associated inpatient stay. Outpatient pharmacy claims for SABAs, ICS,
LABAs, LTRAs, MTHL, IMCS, IgEB were also considered asthma-related.
The MPR was defined as the sum of the number of therapy-days supplied on all FSC 100/50, MF 110, or
MF 220 dispensed from the index date to end the follow-up period divided by the sum of the number of
days between the first and last such prescription during follow-up and the number of days on the last
such prescription. The refill rate was defined as the number of prescriptions for FSC 100/50, MF 110, or
MF 220 dispensed from the index date to the end of the follow-up period divided by the duration of
follow-up.
2.7 Analyses
In the original analysis plan, copay associated to pharmacy claims for study medications on index date,
and mean copay among pharmacy claims during pre-index period were not included in the list of patient
characteristics. After reviewing the results based on the original analysis plan, the analysis plan was
amended to include the copay variables above. The current document reports the results based on the
amended analysis plan. The results based on the original plan (without the copay variables) are
reported in Appendix A.
2.7.1 Patient Characteristics
Descriptive statistics were calculated for all pre-index characteristics for the unmatched and matched
populations. For continuous variables, means and standard deviations (SDs) were reported. Binary and
11
categorical variables were presented as numbers and percents. For the unmatched population,
comparisons between treatment groups were conducted using Chi-square test or Fischer's exact test for
categorical variables, and Student's t-test or Wilcoxon's rank sum test for continuous variables. For the
matched population, comparisons between treatment groups were conducted using McNemar test or
Bowker test for categorical variables, and paired t-test for continuous variables.
2.7.2 Study Outcomes
2.7.2.1 Time to Event Outcomes
For all time to event outcomes, Kaplan-Meier survival curves were derived by treatment group on
matched population. Patients were censored at the end of follow-up in these analyses. Median time to
event was compared across groups using log-rank statistics. Hazard ratios for FSC 100/50 versus MF
were calculated using Cox proportional hazards regression analysis [19]. Standard errors (SEs) and 95%
CIs were calculated for each outcome and treatment group and measure of treatment effect.
The validity of the proportional hazards assumption was ascertained by using interaction terms for
treatment and time, and using the supremum test for non-proportionality [19, 20].
2.7.2.2 Other Outcomes
Descriptive statistics were calculated for all other study outcomes. For continuous variables, means,
standard deviations (SDs), medians, and interquartile ranges, SEs, and 95% CIs were reported. Binary
and categorical variables were presented as numbers and percents. Comparisons between treatment
groups were conducted using McNemar test or Bowker test for categorical variables, and paired t-test
for continuous variables. MPR was also analyzed as both a continuous and categorical variable (<25%,
25%-<50%, 50%-<75%, and 75% to 100%).
12
2.7.3 Multivariate Analyses
Time-to-event outcomes were analyzed using multivariate Cox regression analysis. Covariates were
selected for inclusion and exclusion using entry and exit criteria of p=.10. Regression models were run
for both the unmatched and matched populations.
2.7.4 Statistical Significance
All tests of statistical significance employed two-tailed tests with an alpha level of .05. Ps were reported
out to three decimal places.
2.7.5 Statistical Software
All analyses were conducted using SAS® Proprietary Software, Release 9.2 (SAS Institute Inc., Cary, NC)
3 RESULTS
3.1 Overview
Results of patient selection are reported in Table 1. The parameter estimates of the logistic regression
modeling the probability of MF versus FSC 100/50 therapy that were used to estimate propensity scores
are reported in Table 2. The distributions of propensity scores for patient receiving MF and FSC 100/50
therapy groups are arrayed and compared graphically in Figure 1. Patient characteristics at the index
date are reported by treatment group in Table 3 for both un-matched and matched populations. All the
analyses of outcome measures are conducted on the matched population. Results of Cox regression
analyses of asthma-related exacerbation are reported in Table 4. Descriptive statistics on asthma-
related medical care utilization, asthma-related healthcare costs, and adherence to ICS therapy are 13
reported by treatment group in Table 5, 6 and 7, respectively. Figure 2 through 6 report time to first
asthma-related exacerbation by treatment groups graphically.
3.2 Study Sample
We identified 224,608 patients who had at least one claim with ICD-9 diagnosis code for asthma, and at
least two outpatient pharmacy claims for either FSC 100/50, MF 110, or MF 220 during study period
(Table 1). Of these, 206,325 patients were excluded (92%). The largest sources of attrition (not
mutually exclusive) were less than 12 month of continuous enrollment prior to the initial exacerbation
date (51.5%), no claims with asthma diagnosis on index date or during 365-day pre-index period (35.8%),
index date prior to 1/1/2005 (34.8%), receipt of LTRA, IMCS, MTHL, IgEB, or non-study ICS or LABA
during treatment identification period (28.4%), and receipt of ICS prescriptions during 365-day pre-index
period (26.8%). A total of 18,283 patients met all inclusion and exclusion criteria, including 14,044
patients who received FSC 100/50 and 4,239 patients who received MF 110 or MF 220.
3.3 Patient Characteristics of Un-Matched Patients
Pre-index characteristics of un-matched patients are reported by treatment group in Table 3. FSC
100/50 group was more likely to have index date in 2005 (44% versus 4%, p< .001) and during winter
(30% versus 25%, p< .001). The mean ages among FSC 100/50 and MF groups were 36 years and 38
years (p< .001), respectively. Compared with patients in receiving MF, those treated with FSC 100/50
were more likely to be located in New England (40% versus 36%, p<.001) and less likely to be located in
South (28% versus 34%), less likely to be enrolled in POS plans (46% vs. 56%, p<.001), more likely to be
enrolled in HMOs (25% versus 16%), and had lower Charlson index than patients treated (0.95 versus
1.03, p<.001).
14
In terms of medical care utilization rates and costs during pre-treatment period, MF patients had higher
rates and costs than FSC 100/50 patients in both asthma-related and non-asthma-related medical care
except for the proportion of patients with asthma-related ED visits which was 6% for FSC 100/50 group
versus 4% for MF group (p<.001). There are a few other measures where FSC 100/50 group had a higher
rate or cost, but differences across treatment groups were not statistically significant. FSC 100/50
patients paid a higher copayment for study medications on the index date: 40% paid $21-$30 and 31%
paid $31-$50 (the largest category), compared with 28% and 14% among MF patients (p<.001).
However, mean copayment for all medications during pre-index period was lower among FSC 100/50
group than MF group (14 versus 15, p<.001).
Mean duration of follow-up was longer among FSC 100/50 group by more than 200 days (730 vs. 525
days, p<.001), likely reflecting earlier index dates.
3.4 Propensity Score Matching
The parameter estimates from the logistic model of MF versus FSC 100/50 groups are reported in
Table 2. Figure 1 presents the distributions of propensity scores by treatment group. The C-statistic of
the logistic regression was 0.866. Patients with index date in year 2006 or later (vs. year 2005), or in
summer or fall (vs. winter) were more likely to receive MF therapy. Also patients with age≥18 years (vs.
age<18 years), Charlson index equaled 1 (vs. 0), specialist visits (vs. none), airflow tests (vs. none), three
or more asthma-related medications other than study medications (vs. 2 or less), one or more asthma-
related physician office or hospital outpatient visit (vs. none), asthma-related outpatient prescriptions
(vs. none), or mean copay on all prescriptions greater than $8 (vs. $8 or less) were more likely to receive
15
MF therapy. On the other hand, patients located in West (vs. New England), with HMO (vs. POS), plans
other than POS, PPO or HMO (vs. POS), three or more SABA prescription (vs. none), or costs of non-
asthma related physician office or hospital outpatient visits less than $200 (vs. none) were more likely to
receive FSC 100/50 therapy.
Using propensity scores alone was not sufficient to match MF versus FSC 100/50 therapy on all baseline
characteristics. When categorical variables representing the number of asthma related physician office
or hospital outpatient visits and the copay paid on study medications on index date were included as
along with propensity scores, a successfully matched sample of 2,590 pairs of FSC 100/50 and MF
patients was obtained with none of the pre-index characteristics included in the logistic regression
differing significantly by treatment group (Table 3).
3.5 Study Outcomes
3.5.1 Time to Event Outcomes
Cox regression analyses of time-to-event outcomes are summarized in Table 4. Time to first asthma-
related exacerbation by treatment groups are reported graphically in Figures 2 through 6. The hazard
ratio (HR) for asthma-related hospitalizations was 0.20 for FSC 100/50 vs. MF therapy groups (p
=0.0838). The HR for asthma-related composite outcome of emergence department visit, inpatient
hospitalization, or receipt of SCS was 0.94 for FSC 100/50 vs. MF therapy groups (p=0.2436). There was
no significant difference between matched samples in the risk of the composite outcome of ED visit or
inpatient hospitalization. The risk of asthma-related emergence department visits was lower in the MF
group vs. the FSC 100/50 group (HR=1.04 for FSC 100/50 vs. MF) but this difference was not statistically
significant (p=0.7929).
16
When an interaction term between treatment group and time was added, it was statistically significant
in the model of asthma-related hospitalizations (p =.0228), suggesting that the assumption of
proportionality was not valid. However, the p-value for supremum test for non-proportionality was
0.269, suggesting the proportionality assumption was valid. For the other time to event outcomes, both
tests supported the proportionality assumption.
3.5.2 Asthma-Related Healthcare Utilization
Descriptive statistics on asthma-related medical care utilization are reported by treatment group in
Table 5. Mean numbers of asthma-related physician office and outpatient visits were significantly
greater amongst patients receiving MF vs. FSC 100/50 (1.72 vs. 1.14, p <.001). Although the mean
numbers of SABA prescriptions during follow-up were less for FSC 100/50 vs. MF (1.68 vs. 1.54
respectively), these differences were not statistically significant (p=0.063). A similar result was observed
for the mean number of ICS prescriptions (0.36 vs. 0.32, p= 0.090).
3.5.3 Asthma-Related Healthcare Costs
Descriptive statistics on asthma-related healthcare costs are reported by treatment group in Table 6.
Mean costs of asthma-related outpatient pharmacy and mean total asthma-related costs (both
excluding the costs of MF and FSC 100/50), were higher for MF group ($64 vs. $53, p<.001 and $356 vs.
$270, p=0.003 respectively. Mean cost of other asthma-related healthcare services also was higher for
MF group ($259 vs. $183, p=0.005). Mean cost of asthma-related inpatient care was also higher for MF,
although this difference was not statistically significant ($9 vs. $5, p =0.485). On the other hand, mean
cost of asthma-related outpatient pharmacy including the cost of MF and FSC 100/50 was higher for FSC
17
100/50 group ($742 vs. $602, p<0.001). Mean cost of asthma-related ED visits and mean total asthma-
related costs were also higher for FSC 100/50 group, although this difference was not statistically
significant ($27 vs. $24, p =0.571 for mean ED visits and $958 vs. $894, p = 0.064 for mean total costs).
3.5.4 Adherence to Controller Therapy
Descriptive statistics on adherence to ICS therapy are reported by treatment group in Table 7. Both
mean MPR and mean refill rate per year were higher for the MF group (57.9 vs. 55.6, p=0.003 for mean
MPR and 4.08 vs. 4.02, p=0.461 for mean refill rate per year).
3.5.5 Multivariate Analyses
Results of multivariate Cox regression analyses on time to event outcomes for the un-matched and
matched populations were similar to those for the unadjusted analyses using the matched population.
4 SUMMARY AND CONCLUSIONS
4.1 Summary
This was a retrospective observation study comparing healthcare utilization and costs in asthma patients
who receive FSC 100/50 or MF in typical clinical practice. Data were obtained from a large health
insurance claims database. Healthcare utilization and costs after the first prescription for FSC 100/50 or
MF were compared using propensity score matching to control for possible differences in baseline
characteristics. Measures of interest included the utilization of rescue medications (short-acting beta
agonists [SABAs], systemic corticosteroids [SCS]); asthma-related emergency department (ED) visits,
hospitalizations, and physician office visits; costs of asthma-related care; and adherence with study
therapy (FSC 100/50 or MF).
18
In this sample of patients, those who received FSC 100/50 had fewer asthma-related physician office
and hospital outpatient visits (1.14 vs. 1.72, p<0.001), lower asthma-related outpatient pharmacy costs
(excluding study medications) ($53 vs. $64, p<0.001), lower asthma-related outpatient costs excluding
ED visits ($183 vs. $259, p=0.005), and lower total costs of asthma-related healthcare (excluding study
medications) ($270 vs. $356, p= 0.003). Although not statistically significant, patients who received FSC
100/50 had 80% lower risk of asthma-related hospitalizations, 6% lower risk of the composite outcome
of asthma-related emergency department visits or hospitalizations or receipt of SCS, fewer numbers of
SABA prescriptions (1.54 vs. 1.68), fewer number claims for SCS (0.32 vs. 0.36), and lower mean cost of
asthma-related hospitalizations ($5 vs. $9).
On the other hand, patients who received FSC 100/50 had lower ICS MPR compared to MF patients
(55.6 vs. 57.9, p=0.003). Although not statistically significant, patients who received FSC 100/50 had 4%
higher risk of asthma-related ED visits, higher cost of asthma-related ED visits ($27 vs. $24), higher costs
of study medications ($742 vs. $602), higher total cost of asthma-related healthcare ($958 vs. $894), and
lower ICS refill rate (4.02 vs. 4.08).
4.2 Comparisons with Prior Studies
Navaratnam et al. used data from a large administrative claims dataset to examine asthma-related
healthcare utilization and costs and adherence to controller therapy in patients with mild asthma who
initiated treatment with MF versus FSC [14]. Patients were matched for pre-index characteristics using
propensity score matching and compared using generalized linear regression models. The authors
reported that, compared with FSC patients (n =4094), MF patients (n =4094) had significantly lower
19
post-index asthma-related total charges ($2136 vs. $2315, respectively; p=0.0003), lower
pharmaceutical charges ($727 vs. $925, respectively; p<0.0001), fewer exacerbations (0.14 vs. 0.16,
respectively; p=0.0306), fewer SABA canister claims (0.9 vs. 1.0, respectively; p<0.0001), and greater
adherence measured by prescription fills (3.0 vs. 2.8, respectively; p<0.0001). Asthma-related inpatient
charges, outpatient charges, and adherence measured by percent of days covered were not significantly
different between cohorts [14]. Results reported here are similar to those of Navaratnam in terms of
total asthma-related costs and outpatient pharmaceutical costs. Unlike Navaratnam, however, in the
study reported here there was no difference between groups in exacerbations or refill rate, and patients
receiving MF had higher SABA use than those receiving FSC (although this difference was not statistically
significant). The reasons for these differences in findings are difficult to ascertain but may relate to
differences in the patients included in the study. The Navaratnam et al. study focused exclusive on
patients with mild asthma (defined as patients with ICD-9-CM code of 493.0X, 493.1X, or 493.9X and
who did not have claims that indicated use of >2 SABA canisters or experienced an asthma exacerbation
during the pre-index period). The study reported here had no such restrictions. Also, while the study by
Navaratnam controlled for a variety of patient pre-index characteristics, including age, sex,
comorbidities, pre-index asthma charges, SABA use, total number of asthma-related records, and
asthma related days, the study reported here controlled for a variety of additional potentially
confounding factors including plan type, region, year and season of index prescription, numbers of chest
x-rays and airflow tests, specialist visits, oxygen therapy, numbers of SCS claims, receipt of antibiotics,
and copayment amounts.
20
4.3 Limitations
Limitations of our study should be noted. Because treatment was not randomly assigned, it is possible
that differences between MF and FSC 100/50 treatment groups in study outcomes were due to
differences in patient characteristics that were not observed and hence, not controlled for by propensity
score matching (i.e., “selection bias” or “residual confounding”). In particular, because our study relied
on health-insurance claims, we lacked information on some clinical parameters that may be
independently predictive of outcomes in asthma patients (e.g., frequency and severity of symptoms).
The possibility of residual confounding must therefore be recognized. It should be noted that such
confounding might also explain the failure to find differences between groups in some outcomes.
4.4 Conclusions
Compared with MF, FSC 100/50 may reduce number of physician office and hospital outpatient visits for
the treatment of asthma, and may reduce the total cost of asthma treatment excluding the cost of
controller medications.
21
5 REFERENCES
1. National Asthma Education and Prevention Program. Expert Panel Report 2: Guidelines for the
Diagnosis and Management of Asthma. Bethesda, MD: National Institutes of Health, 1997;
publication No. 97–4051.
2. Global Initiative for Asthma. Global Strategy for Asthma Management and Prevention. Updated
2005. National Institutes of Health and National Heart, Lung and Blood Institute. NIH Publication
No. 02-3659. Bethesda, MD, Public Health Service. National Institute of Health. National, Heart,
Lung and Blood Institute. Available at URL: www.ginasthma.org.
3. Campbell LM. Once-daily inhaled corticosteroids in mild to moderate asthma: improving acceptance
of treatment. Drugs 1999; 58(suppl 4):25-33.
4. Kruse W, Rampmaier J, Ullrich G, et al. Patterns of drug compliance with medications to be taken
once and twice daily assessed by continuous electronic monitoring in primary care. Int J Clin
Pharmacol Ther 1994; 13:914-920.
5. Diette GB, Wu AW, Skinner EA, et al. Treatment patterns among adult patients with asthma: factors
associated with overuse of inhaled beta-agonists and underuse of inhaled corticosteroids. Arch
Intern Med 1999; 159:2679-2704.
6. Eisen SA, Miller DK, Woodward RS, et al. The effect of prescribed daily dose frequency on patient
medication compliance. Arch Intern Med 1990; 150:1881-1884.
7. Mann M, Eliasson O, Patel K, et al. A comparison of the effects of b.i.d. and q.i.d. dosing on
compliance with inhaled flunisolide. Chest 1992; 101:496-499.
8. Suissa S, Ernst P, Benayoun S, et al. Low-dose inhaled corticosteroids and the prevention of death
from asthma. N Engl J Med 2000; 343:332-336.
22
9. Suissa S, Ernst P, Kezouh A. Regular use of inhaled corticosteroids for the long term prevention of
hospitalization for asthma. Thorax 2002; 57:880-884.
10. Schatz M, Cook EF, Nakahiro R, Petitti D. Inhaled corticosteroids and allergy specialty care reduce
emergency hospital use for asthma. J Allergy Clin Immunol 2003; 111:503-508.
11. Product Information ASMANEX® TWISTHALER® 220mcg (mometasone furoate inhalation powder).
Available at URL: http://www.spfiles.com/piasmanex.pdf. Accessed March 12, 2009.
12. Boulet L-P. Once-Daily Inhaled Corticosteroids for the Treatment of Asthma. Curr Opin Pulm Med
2004; 10(1):15-21.
13. Tan RA, Corren J. Mometasone furoate in the management of asthma: a review. Therapeutics and
Clinical Risk Management 2008; 4(6): 1201-1208.
14. Navaratnam P, Friedman HS, Urdaneta E. Mometasone furoate vs fluticasone propionate with
salmeterol: multivariate analysis of resource use and asthma-related charges. Current Medical
Research & Opinion 2009; 25(12): 2895-2901.
15. Complete Prescribing Information and Medication Guide for ADVAIR DISKUS® (fluticasone
propionate and salmeterol inhalation powder). Available at URL: http://www.advair.com/.
Accessed June 8, 2007.
16. Deyo RA. Adapting a comorbidity index for use with ICD-9-CM administrative data: A response. J Clin
Epidemiol 1993; 46:1085-90.
17. Rosenbaum P, Rubin D. The central role of the propensity score in observational studies for causal
effects. Biometrika 1983; 70:41-55.
18. Rubin DB. Estimating causal effects from large datasets using propensity scores. Ann Intern Med
1997;127:757-63.
23
19. Allison, P. A. 1995. Survival analysis using the SAS system: a practical guide. SAS Institute, Inc., Cary,
N.C.
20. Lin D, Wei LJ, Ying Z. Checking the Cox Model with Cumulative Sums of Martingale-Based Residuals,”
Biometrika 1993; 80:557-572.
24
6 TABLES AND FIGURES
25
7 APPENDIX A
43
Study Protocol
Comparison of Healthcare Utilization and Costs in Patients with Asthma who Receive Advair Diskus® (Fluticasone/Salmeterol Inhalation Powder) versus Asmanex® Twisthaler® (Mometasone Furoate Inhalation
Powder) in Typical Clinical Practice Using Health Insurance Claims Data.
I. STUDY OBJECTIVES
The objective of this study is to compare healthcare utilization and costs in asthma patients who receive Advair Diskus 100mcg/50mcg (“FSC 100/50”) vs. Asmanex Twisthaler (mometasone furoate inhalation power “MF” either 110 mcg or 220 mcg) in typical clinical practice II. DATA SOURCE
Data for this study will be obtained from the PharMetrics Patient-Centric Database.
a. Database Description: The database is comprised of information from enrollment files and medical and outpatient pharmacy claims from a variety of private insurers providing healthcare coverage to over 40 million patients enrolled in 70+ health plans across the US. Health plans provide data to the database on a continuous basis, and the number of plans contributing to the dataset has increased over time.
b. Study Period The database employed in this study will span the period from January 2004 through December 2008 (i.e., the most recent five years for which claims data are available) (“study period”).
III. SAMPLE-SELECTION CRITERIA
a. Inclusion Criteria
• One or more medical claims with a diagnosis (primary or secondary) of asthma (ICD-
9-CM 493.XX) during study period;
• Two or more outpatient pharmacy claims for FSC 100/50 or two or more outpatient pharmacy claims for MF (either 110 mcg or 220 mcg);
b. Definition of Index Date, Pre-Index, and Follow-up Periods
• The date of the first prescription for FSC 100/50 or MF will be designated the “index
date”;
• The 12-month (365-day) period prior to the index date will be defined as the “pre-index period”;
• The period beginning with the index date and ending with the last date for which
complete claims data are available will be designated the “follow-up period”;
c. Exclusion Criteria
• Less than 12 months of complete and uninterrupted claims history prior to index date;
• Less than 3 months of follow-up; • Any medical claims during study period with a diagnosis of:
o Chronic obstructive pulmonary disease (COPD) (ICD-9-CM 491, 492, or 496); or
o Respiratory tract cancer (ICD-9-CM 160-164, or 231);
• Any pharmacy claims during the study period for; o Ipratropium; or o Tiotropium
• One or more prescriptions within three months of index date (pre or post) for: o Leukotriene receptor antagonist (LTRA); o Inhaled mast-cell stabilizer (IMCS); o Methylxanthine (MTHL); o IgE blocker (IgEB); or o Non-study ICS or LABA (i.e., other than FSC 100/50 or MF);
• Less than 12 years or greater than 65 years of age as of index date.
• Greater than 65 years of age as of index date. Although the above-described proposed sample-selection algorithm presumes that the pre-
index period will be 12 months and that the follow-up period will be a minimum of 3 months, alternative inclusion criteria may be considered to maximize sample size while also ensuring that the pre-index period is long enough to reliably assess patient characteristics, and that the follow-up period is sufficiently long to permit robust evaluation of study outcomes.
IV. PATIENT CHARACTERISTICS
a. Demographic Characteristics
For each patient in the study sample, demographic and clinical information will be
assessed at index date (“baseline characteristics”). Demographic characteristics will be obtained from enrollment files and include: age (years) at index date, gender, plan type (Health Maintenance Organization [HMO], Preferred Provider Organization [PPO], Other) and US Census region (Northeast, Mid-Atlantic, and other).
b. Pre-Index Comobidities
Diagnosis codes on medical claims during the pre-index period will be scanned to ascertain
comorbidities of potential interest. The Deyo’s version of the Charlson Comordidity Index, a weighted index of 19 chronic medical conditions that is predictive of mortality, post-
2
operative complications, and length of hospital stay, will be calculated for each patient based on diagnoses reported during the pre-treatment period (1).
c. Pre-Index Utilization and Costs of Asthma-Related Care
Procedure codes on professional service claims during the pre-index period will be scanned to identify persons receiving asthma and other respiratory-related services (e.g., chest x-rays, airflow tests, oxygen therapy), and pharmacy claims will be scanned to ascertain the number of prescriptions for rescue medications (SABAs, SCS), antibiotics, and other medications received during this period. The numbers of asthma-related and non-asthma-related ED visits, hospitalizations, and outpatient visits during the pre-index period also will be calculated, as will costs of asthma-related and non-asthma-related treatment (inpatient, pharmacy, and other). Costs will be estimated using estimated costs paid amounts. Measures of adherence (e.g., medication possession ratio [MPR], numbers of prescriptions) for ICS during the pre-index period also will be examined.
V. PRIMARY MEASURES OF INTEREST
Primary measures of interest will include: a. Time-to-Event Analyses:
• Time to first asthma-related ED visit; • Time to first asthma-related hospitalization; • Time to first asthma-related ED visit or hospitalization; and • Time to first asthma-related exacerbation, defined as an asthma-related ED
visit, asthma-related hospitalization, or receipt of SCS for asthma treatment;
b. Utilization of Asthma-Related Healthcare:
• Number of prescriptions for short-acting beta agonists [SABAs] during follow-up; • Number of prescriptions for SCS and ICS during follow-up; and • Number of asthma-related physician’s office or outpatient visits;
c. Costs of Asthma-Related Care:
• Costs of asthma-related medications • Costs of asthma-related ED visits or hospitalizations; • Other asthma-related costs ; • Total costs of asthma-related care during follow-up; • Costs of asthma-related medications excluding FSC 100/50 or MF; and • Total costs of asthma-related care excluding FSC 100/50 or MF;
d. Adherence with Study Therapy (FSC 100/50 or MF):
• MPR for study therapy (FSC 100/50, MF); and • Refill rate for study therapy;
3
VI. ANALYSES OF DATA
a. Pre-Index Characteristics:
Descriptive statistics will be calculated for all pre-index characteristics. For continuous variables, means and standard deviations (SDs) will be reported. Binary and categorical variables will be presented as frequency counts and percents. For the unmatched population, comparisons between the treatment groups are conducted using Chi-square tests or Fisher’s exact test for categorical variables and Student’s t-test or Wilcoxon’s rank sum test for continuous variables. For the matched population, comparisons between treatment groups will be conducted using the McNemar test or Bowker test for categorical variables, and a paired t-test for continuous variables.
b. Outcome Measures: Propensity scoring techniques (2, 3) will be used to mach each person in the FSC 100/50 treatment
group to one person in the MF treatment group. Propensity scores for each subject is defined as the estimated probability (range: 0 - 1) of receiving FSC 100/50 from a logistic regression model, conditional upon the observed values of the other baseline characteristics as well as duration of follow-up. Matched pairs of subjects receiving FSC and MF will then be identified using nearest-neighbor matching, based on the propensity score.
The coefficients of the parameters of the logistic model used to estimate propensity scores will be
reported. The distribution of propensity scores across treatment groups will be arrayed and compared graphically for all subjects (i.e., via histogram). The C-statistic for the logistic regression analysis on treatment groups will be reported.
For all time-to-event outcomes, Kaplan-Meier survival curves will be estimated by treatment group.
Patients will be censored at the end of follow-up in these analyses. Median time to event will be compared between the two treatment groups using log-rank statistics. Hazard ratios for FSC versus MF will be calculated using Cox proportional hazards regression analysis (4). The validity of the proportional hazards assumption will be ascertained using interaction terms for treatment and time, by examining Shoenfeld residuals, and using the supremum test for nonprorortionality (5).
Descriptive statistics will be calculated for all other study outcomes. For continuous variables,
means, standard deviations (SDs), medians, and interquartile ranges will be reported. Binary and categorical variables will be presented as frequency counts and percents. Comparisons between treatment groups will be conducted using the McNemar test or Bowker test for categorical variables, and a paired t-test for continuous variables. Subgroup analyses will be conducted for selected baseline characteristics, as appropriate.
4
REFERENCES 1. Deyo RA. Adapting a comorbidity index for use with ICD-9-CM administrative data: A response. J Clin
Epidemiol. 1993;46:1085-90.
2. Rosenbaum P, Rubin D. The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70:41-55.
3. Rubin DB. Estimating causal effects from large datasets using propensity scores. Ann Intern Med 1997;127:757-63.
4. Allison, P. A. 1995. Survival analysis using the SAS system: a practical guide. SAS Institute, Inc., Cary, N.C.
5. Lin, D., Wei, L. J., and Ying, Z. (1993), “Checking the Cox Model with Cumulative Sums of Martingale ‐Based Residuals,” Biometrika 1993;80:557-572.
5
