Transcript Document

OsteoporosisWRAP SlideCAST
Complexities and Challenges of
Measuring Fracture End Points
Interpreting the Results of Existing Trials
Steven T Harris MD FACP
Clinical Professor of Medicine
University of California, San Francisco
Fracture Endpoints
►
Regulatory approval of existing medications has
been based on fracture reduction, rather than
changes in surrogates such as bone mineral
density (BMD) alone
►
The most common fracture endpoint has been the
reduction in vertebral fracture
●
Vertebral fracture has most often been defined on the
basis of morphometric changes, including both “new” and
“worsening” fractures
●
“Clinical” vertebral fracture—defined as fracture
associated with back pain—has usually been identified as
a secondary endpoint
Fracture Endpoints, con’t
►
“Non-vertebral fracture” has most commonly been
identified as a secondary endpoint as well
● Varying definitions of “non-vertebral fracture”: all
fractures vs fragility fractures vs a particular
composite endpoint
● Hip fracture has sometimes—but not always—
been identified as a separate endpoint of
particular interest
Fracture Endpoints, con’t
►
Fracture reduction has most often been expressed —
both clinically and commercially—as relative risk
reduction, rather than absolute risk reduction
►
The rapidity with which fracture protection occurs has
been of clinical—and commercial—interest
►
There are no head-to-head randomized trials with
fracture as the primary endpoint
►
The pivotal fracture trials recruited patients with a
certain “sameness”—but there were notable differences
in demographics, baseline fracture status and baseline
BMDs nevertheless
●
There were obvious differences in observed fracture
rates—both vertebral and non-vertebral
Vertebral Fracture Effects In Prospective Trials
Direct Comparisons Among Trials Cannot Be Made
Calcium
30
% of
Patients
With New
Vertebral
Fractures
Over 3-5
Years
25
CT
RLX
ALN
RIS
IBN
ZOL
49%
40%
% = Relative Fracture Risk
Reduction
30%
20
47%
15
41%
70%
52%
10
50%
5
50%
0
PROOF MORE
FIT-VFA VERT-NA VERT-MN
With Prevalent Vertebral Fractures
BONE
HORIZON
± Vert Fx
MORE
FIT-CFA
Without Vert Fx
Chesnut CH, et al. Am J Med. 2000;109:267
Ettinger B, et al. JAMA. 1999;282:637
Black DM, et al. Lancet. 1996;348:1535
Cummings SR, et al. JAMA. 1998;280:2077
Harris ST, et al. JAMA. 1999;282:1344
Reginster J, et al. Osteoporos Int. 2000;11:83
Chesnut CH, et al. J Bone Miner Res. 2004;19:1241 Black DM, et al. N Engl J Med 2007:356:1809
Nonvertebral Fracture Effects
In Prospective Trials
Direct Comparisons Among Trials Cannot Be Made *
Calcium
CT
RLX
ALN
RIS
IBN
ZOL
30
% = Relative Fracture Risk Reduction
25
% of Patients
With New
Nonvertebral
Fractures
Over 3-5
Years
20
12%
20%
p = n.s.
15
p = 0.06
10%
12%
p = n.s.
25%
39%
p = n.s.
10
33%
p = 0.06
p < 0.001
p = n.s.
p < 0.05
5
0
PROOF
MORE
FIT-VFA
FIT-CFA VERT-NA
VERT-MN
BONE
HORIZON
* Patient demographics and the definition of nonvertebral fracture varied among studies
Chesnut CH, et al. Am J Med. 2000;109:267
Black DM, et al. Lancet. 1996;348:1535
Harris ST, et al. JAMA. 1999;282:1344
Chesnut CH, et al. J Bone Miner Res. 2004;19:1241
Ettinger B, et al. JAMA. 1999;282:637
Cummings SR, et al. JAMA. 1998;280:2077
Reginster J, et al. Osteoporos Int. 2000;11:83
Black DM, et al. N Engl J Med 2007:356:1809
Fracture Endpoints
►
If existing randomized clinical trials cannot
reasonably address all issues of clinical relevance,
then additional insights can be gleaned from:
● Additional—post hoc—analyses of existing trial
data
● Meta analyses of existing trial data
● Observational data
Nonvertebral Fracture Reduction
% patients with nonvertebral fracture
Post-hoc analyses
14
Control
12
Risedronate
12
59%
10
P=0.002
4
2
0
Alendronate
36%
10
8
6
Control
14
*
P=0.002
*
8
* *
*
*
* *
*
*
** *
6
12
18
24
30
36
Time (months)
n=1172; Patients with lumbar spine BMD T-score <-2.5. Nonvertebral
fractures based on a composite endpoint of the following: clavicle, hip,
humerus, leg, pelvis and wrist.
Harrington Calcif Tissue Int 2004;74:129
*
6
4
2
0
0
6
12
18
24
30
36
Time (months)
n= 3658; Patients with a baseline vertebral fracture or femoral
neck T score of -2.5 or less. Nonvertebral osteoporotic fractures
measured as a composite endpoint (sites are not defined).
Black JCEM 2000;85:4118
Ibandronate and the risk of non-vertebral
and clinical fractures in women with
postmenopausal osteoporosis (PMO):
results of a meta-analysis of
Phase III studies
Harris ST, Blumentals WA and Miller PD. Curr Med Res Opin. 2008;24:237-45
Study Objectives
 To evaluate the nonvertebral and clinical
fracture efficacy of different doses of
ibandronate using a pooled analysis of data
from 4 pivotal phase III clinical trials: BONE,1
IV Dose Fracture,2 MOBILE,3 and DIVA4
 To estimate the fracture rate over time using
the Kaplan-Meier method
1oral
iBandronate Osteoporosis vertebral fracture trial in North America and Europe. Chesnut CH, et al.
Curr Med Res Opin. 2005;21:11-18.
2Recker
R, et al. Bone. 2004;34:890-899.
3Monthly
4Dosing
Oral iBandronate In LadiEs. Reginster JY, et al. Ann Rheum Dis. 2006;65:654-661.
IntraVenous Administration. Delmas PD, et al. Arthritis Rheum. 2006;54:1838-1846.
Pooled Fracture Analysis
Methods
►
Cox regression analyses were adjusted for variables
to allow for meaningful comparisons among trials
● Age
● Previous clinical fracture
● Baseline BMD
• Lumbar spine BMD for models with all clinical
fractures
• Hip BMD for models with nonvertebral fractures
(NVFs) or key NVFs
►
Kaplan-Meier methods were used to:
● Plot survival curves for time to fracture
● Compare rates using a log-rank test
Harris ST, Blumentals WA and Miller PD. Curr Med Res Opin. 2008;24:237-45
Pooled Fracture Analysis
Methods (cont.)
►
Fracture rates were determined by tabulating the
proportion of all ITT patients with at least 1 fracture
►
Doses were pooled based on annual cumulative
exposure (ACE)
● ACE was calculated by multiplying the drug
strength by the number of annual doses and by an
absorption factor (0.6% for oral dosing and 100%
for IV dosing)
● For example, the ACE for oral BONIVA 150 mg
once-monthly would be 10.8 mg = 150 x 12 x 0.006
Harris ST, Blumentals WA and Miller PD. Curr Med Res Opin. 2008;24:237-45
Annual cumulative exposure (mg)
IV and oral ibandronate annual cumulative
exposure across studies
12 mg
12
10.8 mg
8 mg
8
5.5 mg
4 mg
4
0
Quarterly IV
Oral daily
MF43801 BONE2
Quarterly IV
Oral monthly
Quarterly IV
IRIS3
MOBILE4
DIVA5
1. Recker R, et al. Bone. 2004;34:890-899. 2. Chesnut CH, et al. J Bone Miner Res. 2004;19:1241-1249. 3. Adami S,
et al. Bone. 2004;34:881-889. 4. Miller PD, et al. J Bone Miner Res. 2005;20:1315-1322. 5. Data on file (Reference
#161-153), Hoffman-La Roche Inc, Nutley, NJ 07110
Fracture risk reductions for high dose of
ibandronate (≥10.8 mg/yr), by fracture type
All clinical fractures
Nonvertebral fractures
28.8%*
29.9%*
Key nonvertebral sites
0
Relative risk reduction,
†
%
-10
-20
-30
-40
-50
-60
*P<0.05
†Relative
risk reduction adjusted for age, baseline BMD, and fracture history.
34.4%*
+/- 95% Confidence Intervals
Pooled Ibandronate Fracture Analysis
Estimated Clinical
Fracture Rate, %
10
8
6
4
2
†Cox
ACE ≥10.8
mg
Placebo
Relative Risk
Reduction
28.8%
P<0.05†
0
50 100 150 200 250 300 350 400 450 500 550 600 650 700 750
Days
0
regression analyses for difference in relative risk of fracture with ibandronate vs placebo.
Kaplan-Meier plot (2-year data).
Harris ST, Blumentals WA and Miller PD. Curr Med Res Opin. 2008;24:237-45
Pooled Ibandronate Fracture Analysis, con’t
Estimated Nonvertebral
Fracture Rate, %
10
8
Relative Risk
Reduction
29.9%
P<0.05†
6
4
2
†Cox
ACE ≥10.8
mg
Placebo
0
50 100 150 200 250 300 350 400 450 500 550 600 650 700 750
Days
regression analyses for difference in relative risk of fracture with ibandronate vs placebo.
Kaplan-Meier plot (2-year data).
0
Harris ST, Blumentals WA and Miller PD. Curr Med Res Opin. 2008;24:237-45
Limitations
►
Not all studies were placebo-controlled
►
A limited number of baseline patient characteristics
(age, previous fracture, baseline BMD) were
available for use in multivariable models
►
This was an exploratory, post hoc analysis with no
adjustment for multiple comparisons
Harris ST, Blumentals WA and Miller PD. Curr Med Res Opin. 2008;24:237-45
Conclusions
►
In this pooled analysis, doses of ibandronate
resulting in an ACE of ≥10.8 mg—which include the
marketed doses of 150 mg monthly oral and 3 mg
IV quarterly—significantly reduced the risk of clinical
fractures and non-vertebral fractures compared to
placebo
►
The high ACE group (≥10.8 mg) had a longer time
to fracture compared to placebo for all clinical
fractures, non-vertebral fractures and key nonvertebral fractures
Harris ST, Blumentals WA and Miller PD. Curr Med Res Opin. 2008;24:237-45
Observational Studies
Strengths and Limitations of Claims Data
►
Strengths
● Treatment patterns are those in “real world” clinical practice
● A large number of patients can be followed over time
►
Limitations
● No chart review available to validate the ICD-9 codes
● Selection bias is possible—if not probable
● Cannot assess the use of samples or OTC products
● Extraneous variables are not controlled
Protocare Sciences Database
Nonvertebral Fracture Incidence
Fracture incidence (%)
4.0
ALN vs. CAL
RR=0.74
(p=NS)
RIS vs. CAL
RR=0.31
(p<0.05)
ALN vs. CAL
RR=0.75
(p=NS)
RIS vs. ALN
RR=0.46
(p=0.067)
3.0
RIS vs. CAL
RR=0.25
(p<0.01)
RIS vs. ALN
RR=0.41
(p<0.05)
25%
2.0
75%
26%
59%
69%
1.0
54%
0.0
6-months
Nasal calcitonin
12-months
Alendronate
Risedronate
*Adjusted for age, sex, HRT use, number of concomitant
medications and fragility fractures in pre-treatment period
Watts N, et al. J Manag Care Pharm 2004;10:142-151
► Effectiveness
of Bisphosphonates
on Nonvertebral and Hip Fractures
in the First Year of Therapy: The
Risedronate and Alendronate
(REAL) Cohort Study
Silverman SL, Watts NB, Delmas PD, Lange JL, Lindsay R
Silverman L et al. Osteoporos Int. 2007;18:25-34
Study Objective
► To
evaluate the onset of fracture
reduction by bisphosphonate
therapies in clinical medical practice
by measuring the incidence of hip
and nonvertebral fractures among
women 65+ in the year following
initiation of either once-a-week doing
of risedronate or alendronate
Fracture Outcomes
Fracture Sites and Analysis Methods
Study Population
Inclusion Criteria
►
All women
►
> 65 years of age
►
Once-weekly bisphosphonate (BP) use between July 2002
and September 2004
Exclusion criteria
►
< 6 months of health plan enrollment before first BP use
►
< 3 months of health plan enrollment after first BP use
►
Any BP use during 6-month history period
►
Diagnosis of malignancy or Paget’s disease
►
Discontinuation of BP therapy with first 3 months of
exposure period
Baseline Characteristics
Determinants of Fracture Risk
Cumulative Hip Fracture Incidence
Cumulative Incidence of Hip Fractures During Therapy
Cumulative Nonvertebral Fracture Incidence
Cumulative Incidence of Nonvertebral
Fractures During Therapy
Conclusion
► In
an observational study of medical
practice, patients using risedronate
had a lower incidence of
nonvertebral and of hip fractures in
the first year of therapy than those
using alendronate
The VIBE Study:
eValuation of IBandronate Efficacy
►A
retrospective cohort study
comparing fracture rates for women
receiving monthly ibandronate
versus weekly bisphosphonates
Objective
►
To investigate the efficacy of monthly
ibandronate versus weekly BPs by
comparing rates of incident clinical fractures
over 12 months in a retrospective cohort
study
BP = bisphosphonate
Methodology
►
Longitudinal patient data from health plans: eligibility, medical and
pharmaceutical claims (anonymous, HIPAA-compliant)
● i3 Research database (single, large health plan with 14 million
annual enrolment)
● i3 Innovus IMPACT database (multiple, unaffiliated health plans
with over 70 million enrolled 1997–2007)
►
Endpoints: new fractures
● hip
● non-vertebral (inclusive of hip)
● vertebral (clinical)
● any (inclusive of all the above)
►
Analysis
● time to first incident fracture (Kaplan-Meier methods)
● relative risk (hazard ratio) from Cox regression, controlled for
baseline covariates
Study population
►
Women aged ≥45 years newly started on an oral BP
●
●
►
Continuous enrolment in health plan for ≥9 months
●
●
►
received ≥1 prescription between 1 April and 31 December
2005 (index date) for ibandronate (Boniva/Bonviva) 150mg
once monthly, alendronate (Fosamax) 35mg or 70mg
weekly or risedronate (Actonel) 35mg weekly
no BP in baseline period (6 months prior to index date)
6 months before starting BP therapy
≥3 months after the first BP prescription
Excluded women with Paget’s disease of bone or
malignant neoplasm
Baseline characteristics
►
Patient
●
●
●
►
●
age at index date
co-pay of index treatment
GI diagnosis
●
►
►
Osteoporosis
●
●
●
●
fracture history
bone densitometry
procedure
osteoporosis diagnosis
osteopenia diagnosis
Healthcare utilisation
number of outpatient visits
hospitalisation
Medication use
●
●
●
●
●
GI = gastrointestinal; SERM = selective estrogen receptor modulator
PPI = proton pump inhibitor; H2B = H2B=H2 blocker; CTP = cytoprotective
number of therapeutic
classes
Estrogen
Non-estrogen antiosteoporotics (calcitonin,
SERMs)
GI medication (PPI, H2B,
CTP)
corticosteroid
Period of observation
►
Each subject required to have a 6 month baseline period
● to examine medication use and medical history
►
After starting BP therapy (index date) each subject was
observed for fracture for up to 12 months, or until
● loss to follow-up (end of health plan enrolment)
● discontinue therapy (for primary analysis only)
● change in BP therapy
• switched to a different BP
• switched dosing regimen (e.g. weekly to daily)
►
Further observations were then stopped (censored)
Study design: retrospective cohort study
Patient intake period: start BP therapy (index date)
1 April 2005
31 December 2005
Baseline period: 6 months
1 October 2004
30 June 2005
Follow-up period: up to 12 months
1 April 2005
31 December 2006
Overview of Analyses
Primary analysis: adherent patients
►
“Head-to-Head” comparison of drug efficacy
●
●
patients received minimum of 90 days therapy
observed for fracture after 90 days, only while remaining on therapy
Secondary analysis: all patients starting BP therapy
►
patients required to have received ‘≥1 BP prescription’
►
observed for fracture after index prescription
Sensitivity analyses
►
exclude patients with clinically important baseline differences: use of
oestrogen or other anti-osteoporotics, corticosteroids, GI
medications, etc.
►
definitions of adherence: refill gap for ibandronate* (30 days vs 45
days), requiring 90-day minimum adherence
*Gaps for monthly ibandronate (always 30 days for weekly BPs)
Statistical Analysis
►
Time-to-event analysis of fracture incidence with Kaplan-Meier methods
►
Cox regression used to estimate RR (hazard ratio) of fracture comparing monthly
versus weekly cohorts, adjusted for covariates
● results shown are from stepwise regression models with all baseline
characteristics entered as candidate variables, retaining statistically significant
variables
• age, osteoporosis diagnosis, use of DXA, fracture history
• number of medication classes used, oestrogen use
• number of outpatient visits
►
Conclusions were unchanged in models that included other clinically important
variables
●
●
use of other anti-osteoporotics (calcitonin, SERMs)
use of glucocorticosteroids
RR = relative risk; DXA = dual energy X-ray absorptiometry
Patient disposition:
primary and secondary analyses
Candidate patients
≥1 prescription for BP*
(n=445,430)
Females aged ≥45 years
(n=338,872)
Eligible patients
First treatment, no cancer or Paget’s disease
(n=91,630)
1
• Aged <45 years or male (n=13,217)
• >1 study drug (n=19)
2
• BP pre-index (n=237,480)
• Cancer pre-index (n=9,681)
• Paget’s disease (n=81)
3
• Invalid data: negative costs (n=32)
Secondary analysis population:
all patients starting BPs
(n=91,598)
Primary analysis population:
adherent patients
(n=64,182)
*Alendronate, ibandronate or risedronate
4
• Discontinue BP <90 days (n=27,416)
Baseline Characteristics: Primary Analysis
Weekly
BP therapy
(n=56,837)
Monthly
ibandronate
(n=7,345)
p value
Duration of observation in days, mean (SD)
218 (98.3)
223 (94.1)
0.002
Age, mean (SD)
60.5 (8.8)
60.1 (8.6)
0.002
Osteoporosis diagnosis, %
35.5
40.2
<0.001
Osteopenia diagnosis, %
0.3
0.4
0.076
Bone densitometry procedure, %
55.3
56.2
0.116
Fracture history, %
3.7
3.6
0.520
GI diagnosis, %
16.6
21.9
<0.001
GI medication use, %
16.5
23.6
<0.001
Oestrogen use, %
19.2
24.9
<0.001
Other anti-osteoporotic, %
5.9
10.8
<0.001
Glucocorticosteroid use, %
9.8
12.4
<0.001
Number of therapeutic classes, mean (SD)
5.0 (5.30)
6.1 (4.93)
<0.001
Outpatient visits, mean (SD)
15.0 (17.2)
15.9 (16.8)
<0.001
5.5
5.4
0.559
Hospitalisation, %
Primary analysis: monthly ibandronate
vs weekly bisphosphonates at 12 months
1.6
RR=0.88
p=0.26
Fracture incidence (%)
1.30 1.29
Monthly oral ibandronate
(n=7,345)
Weekly oral BPs
(n=56,837)
RR=0.82
p=0.052
1.51
1.40
1.2
0.8
RR=1.06
p=0.84
0.4
0.19 0.20
0
Fx=738
Fx=95
Non-vertebral
Fx=106
Fx=15
RR=0.36
p<0.01
0.24
0.11
Fx=135
Fx=8
Vertebral
Hip
Fractures
RR = adjusted RR (hazard ratio) using Cox regression controlling for potential
confounding variables; Persistent patient cohort with no refill gap >45 days
(monthly) or 30 days (weekly); Fx = absolute number of fractures
Fx=858 Fx=103
Any
Primary analysis: monthly ibandronate
versus weekly alendronate at 12 months
Fracture incidence (%)
1.6
Monthly oral ibandronate
(n=7,345)
Weekly oral alendronate
(n=35,865)
RR=0.88
p=0.26
1.29 1.29
RR=0.80
p<0.05
1.51
1.40
1.2
0.8
RR=1.00
p=1.00
RR=0.34
p<0.01
0.4
0.19 0.20
Fx=464
Fx=95
Fx=68
Fx=15
0.24
0.11
Fx=86
Fx=8
Fx=542 Fx=103
0
Non-vertebral
Hip
Vertebral
Fractures
RR = adjusted RR (hazard ratio) using Cox regression controlling for potential
confounding variables; Persistent patient cohort with no refill gap >45 days
(monthly) or 30 days (weekly); Fx = absolute number of fractures
Any
Primary analysis: monthly ibandronate
versus weekly risedronate at 12 months
1.6
Fracture incidence (%)
Monthly oral ibandronate
(n=7,345)
Weekly oral risedronate
(n=20,972)
RR=0.91
p=0.42
1.31
1.29
RR=0.85
p=0.14
1.51
1.40
1.2
0.8
RR=1.19
p=0.57
RR=0.39
p<0.01
0.4
0.18
Fx=274 Fx=95
Fx=38
0.20
0.23
0.11
Fx=15
Fx=49
0
Non-vertebral
Hip
Fractures
Fx=8
Vertebral
RR = adjusted RR (hazard ratio) using Cox regression controlling for potential
confounding variables; Persistent patient cohort with no refill gap >45 days
(monthly) or 30 days (weekly); Fx = absolute number of fractures
Fx=316 Fx=103
Any
Sensitivity analyses: exclude patients with
clinically important baseline differences
►
In order to understand whether the results were due
to baseline differences in the populations, sensitivity
analyses were performed which excluded patients
with the following during baseline
1. Estrogen or other anti-osteoporotic medication use (SERMS,
calcitonin)
2. corticosteroid use
3. fracture
4. GI medication use
5. corticosteroid use and/or osteopenia
►
The conclusions regarding fracture endpoints
remained unchanged in the analyses
Limitations
►
Observational studies are subject to potential confounding (not randomised)
●
●
►
Diagnosis data was collected for billing purposes, not for research
●
●
●
►
there may be residual confounding that was not adequately controlled for
but this was also evaluated with multiple sensitivity analyses
a diagnosis code on a medical claim is not necessarily confirmation of a fracture
vertebral fracture may be coded on basis of patient symptoms, rather than X-ray
it is unlikely that fracture misclassification is differential and clinical fractures are the
most important
There is no perfect measure of patient compliance
●
●
a filled prescription does not mean the medication was taken, and medication
samples are not recorded in prescription claims data
the above would underestimate antifracture benefit (and unlikely to be differential)
►
Duration of follow-up period was limited to 12 months by the available data
►
P values were not adjusted for multiple comparisons
Strengths
►
Large observational study in a real-world population
● findings more appropriate for generalisation and
representative than clinical trials
● learnings from other observational studies considered
in study design
►
All potential confounders were entered into the Coxregression model and results were unchanged
►
Results consistent across a broad range of sensitivity
analyses
Discussion
►
Risk of hip and non-vertebral fractures was the same
with monthly ibandronate and weekly bisphosphonates
►
Suggests efficacy of ibandronate on non vertebral
fractures, consistent with data obtained in a recent metaanalysis1
►
The rates of vertebral fractures were significantly lower
with monthly ibandronate versus weekly
bisphosphonates in the primary analysis, but the number
of patients with vertebral fracture with monthly
ibandronate was limited (n=8)
1Adachi
R, et al. J Bone Miner Res 2007;S221:S210–11 (Abstract M428)