Foot & Ankle International - September 2019 - 1072

1072
heuristic.30 This procedure is designed to reduce systemic
bias in treatment-control comparisons.34 PS subclasses were
identified by an independent statistician with no access to
outcome data. A PS design was identified such that within
each subclass, subjects in the 2 treatment groups had
approximately the same multivariate covariate distribution.
Consequently, analyses could proceed as if the patients
within each subclass had been randomly allocated to their
treatment. The analysis of the outcome variables was then
carried out in a stratified manner when estimating the difference between treatment groups.

Patients: Propensity Score Model
All 132 patients treated with rhPDGF-BB/β-TCP-collagen
were assigned to a PS subclass. In contrast, 167 of 189
patients treated with autograft were assigned to a PS subclass (Figure 1). The 22 excluded controls had combinations of baseline covariates not observed in the 132 patients
treated with rhPDGF-BB/β-TCP-collagen. After subclassification and controlling for PS subclass, there were no
important differences between the 2 groups for any of the
baseline covariates considered (all P ≥ .725).
Demographic data and outcome measure scores at baseline for the 132 rhPDGF-BB/β-TCP-collagen-treated
patients and 167 autograft-treated patients selected by the
PS model are summarized in Table 2. The mean age for the
rhPDGF-BB/β-TCP-collagen cohort was 53.5 ± 14.7 years
and for the autograft cohort was 56.2±14.0 years. The treatment and control groups consisted of 53% and 51.5% males,
respectively. The most common primary diagnoses were
posttraumatic osteoarthritis (126/299, 42.1%), primary
arthritis (125/299, 41.8%), and rheumatoid arthritis (20/299,
6.7%). BMI and relative risk factors for nonunion (ie, smoking and obesity) were comparable between the groups. All
preoperative clinical outcome measures were similar for
both groups.

Study Outcomes
Clinical, functional, and radiographic endpoints were
assessed to monitor safety, clinical healing, and progression
of fusion. Adverse events, complications, protocol deviations, and revision surgeries were recorded.
The primary outcome for the RCT was joint fusion,
assessed by CT using a robust definition of ≥50% osseous
bridging across the articulation for successful fusion. After
this protocol was finalized, a study demonstrated that clinical success is obtained with ≥25% osseous bridging.18 If
multiple joints were being fused, the full complement of
treated joints was assessed (ie, all treated joints had to be
fused for the procedure to be successful). The primary outcome for the PS model was pain at the fusion site with
weightbearing, assessed by 0- to 100-mm visual analog

Foot & Ankle International 40(9)
scale (VAS) pain scores; a difference <10 mm between the
groups indicated noninferiority.
Secondary outcomes incorporated patient-, surgeon-,
and independent radiologist-reported data. "Clinical healing status" was determined by the surgeon based on a global
assessment of the patient's progress at the level of the full
joint (if more than 1 joint was fused) and at the level of the
individual joints. Therapeutic failure was defined as any
nonunion or delayed union that required secondary therapeutic intervention. VAS pain was assessed at the surgical
site with weightbearing, at the fusion site, and at the graft
harvest site. "Clinical success" was defined as reduced VAS
pain on weightbearing and lack of revision surgery. Other
secondary outcomes included the SF-12 PCS score version
2,16,36 American Orthopaedic Foot & Ankle Society
(AOFAS) Ankle-Hindfoot Scale,27 and Foot Function Index
(FFI).3 Decreases in VAS pain and FFI scores represented
improvement, whereas increases in AOFAS and SF-12
scores represented improvement. Radiographic fusion was
determined by the presence of osseous bridging across at
least 3 of 4 predefined aspects (anterior, posterior, medial,
and lateral).
Safety-related outcomes analyzed included event frequency, severity and potential relationship to the study
device, surgical site complications including nonunion, and
patient dropout due to death or serious adverse events.

Statistical Analysis
A power analysis was performed before conducting the PS
subclassification analysis with the primary outcome of
weightbearing VAS pain at the fusion site. The original pivotal trial demonstrated a change of approximately 50 points
in weightbearing VAS pain from baseline to 1-year followup. Noninferiority was declared if the upper confidence
bound of a 2-sample confidence interval for the difference
between the treatments was less than the predetermined
10-point margin. Employing a type 1 error rate of 5% and
assuming equal responses for the 2 treatment arms, a sample size of 132 rhPDGF-BB/β-TCP-collagen subjects and
167 autograft subjects produced at least 88% power to
declare noninferiority of rhPDGF-BB/β-TCP-collagen relative to autograft.
Binary data are presented as counts and rates. The treatments were compared using a logistic regression model to
allow for PS quintile as a stratification variable. Odds ratios
(OR) were calculated as the odds of significant improvement for rhPDGF-BB/β-TCP-collagen divided by odds of
significant improvement for autograft, where OR >1.0
favors rhPDGF-BB/β-TCP-collagen and OR <1.0 favors
autograft. Noninferiority was declared when the 95% confidence bound of the OR exceeded 0.50.17 Continuous data
are presented as means and standard deviations.
Comparisons between treatment groups were carried out

Foot & Ankle International - September 2019

Table of Contents for the Digital Edition of Foot & Ankle International - September 2019