Foot & Ankle International - September 2019 - 1076

1076	

Foot & Ankle International 40(9)

Table 6.  Safety-Related Outcomes for the Propensity Score Subclassification Cohorts.
rhPDGF-BB/β-TCP-Collagen
(n = 132)
 

No. of Subjects (%)

Serious TEAEs
Device-related TEAEs
Complications associated with surgical procedure
Serious operative complications
Infection-surgical or treatment emergent
Chronic pain at autograft donor site (≥20 mm on VAS)
  At 6 months
  At 12 months

17 (12.7)
3 (2.3)
47 (35.6)
8 (6.1)
27 (20.5)
0 (0)
0 (0)

No. of Events

Autograft
(n =167)
No. of Subjects (%) No. of Events P Value

20
3
79
8
36

25 (15.0)
6 (3.6)
52 (31.1)
10 (6.0)
26 (15.6)

35
10
76
11
32

1.00
.741
.141
.808
.127

0
0

21/161 (13.0)
16/158 (10.1)

21
16

<.01
<.001

Abbreviations: rhPDGF-BB/β-TCP, recombinant human platelet-derived growth factor BB homodimer; TEAE, treatment-emergent adverse event;
VAS, visual analog scale.

Fusion in the ankle and hindfoot was evaluated by CT at
24 weeks and radiographs at 24 and 52 weeks using rigorous
benchmarks of ≥50% osseous bridging on CT and at least 3
of 4 aspects fused on radiographs.11 Using these criteria,
fusion rates for rhPDGF-BB/β-TCP-collagen were statistically noninferior to autograft. The clinical healing status
achieved in both groups was much higher than fusion by CT
scan; Glazebrook et al recently demonstrated that clinical
success is achieved if ≥25% osseous bridging is obtained.18
Patient-reported outcomes, including FFI, AOFAS AnkleHindfoot Scale and SF-12 PCS scores, and VAS pain, were
equivalent in both groups. Additionally, key safety-related
outcomes were equivalent in both groups, with the benefit of
no donor site pain in the rhPDGF-BB/β-TCP-collagen
cohort. These results are consistent with an RCT of rhPDGFBB/β-TCP-collagen conducted at 5 Canadian sites.7
The current study builds on earlier work by conducting
an additional RCT across both the United States and Canada,
and then using a PS subclassification technique for analysis
to provide a robust observational design. Inferences regarding treatment group differences from combined studies may
proceed as if patients within each cohort had been randomly
allocated to a treatment arm within 5 subclasses, designed so
treated and control subjects within each subclass share
approximately the same multivariate covariate distribution.
Control subjects with covariate combinations not observed
in the treated group are trimmed, promoting covariate balance. The PS subclasses and trimming of controls was performed by an independent statistician without access to
outcome data to avoid bias. PS modeling is preferred over a
simple pooling of patient cohorts, as such pooling fails to
account for potential differences in patient enrollment characteristics across studies. This methodology is especially
valuable in orthopedic surgery, particularly ankle and hindfoot surgery, where it is difficult to conduct RCTs with large
sample sizes. PS model cohort analyses were recently

utilized to compare patient-reported functional outcomes,
quality of life, and satisfaction following unicompartmental
versus total knee arthroplasty.20,25
Patients included in this study required supplemental
bone graft, as determined by the surgeons on the basis of
accepted clinical risk factors for nonunion, such as smoking, diabetes, soft tissue injury, and interfragmentary
gaps.35 With the efficacy and safety of rhPDGF-BB/βTCP-collagen now demonstrated in these patients, future
studies could evaluate the use of rhPDGF-BB/β-TCPcollagen in patients at very high risk for nonunion, including patients with deformity, neuropathy, or metabolic
disease; in more complicated cases requiring plate fixation
or intramedullary rods or nails; and in patients requiring
revision surgery.
Strengths of this study include the PS subclassification
design, which allowed a valid analysis to compare, as if under
randomization, the 2 treatment groups from multiple RCTs.
Patients were enrolled at 37 sites in the United States and
Canada for the 3 RCTs, allowing generalizability of the data.
Another strength is the high rate of patient follow-up, with
121 of 132 (91.7%) patients treated with rhPDGF-BB/β-TCPcollagen having outcome data available at 1 year.
A limitation of this study is that the PS design could only
control for differences on the observed covariates included
in the model. A rich set of covariates were included to
reduce this risk. Nonetheless, bias may result from covariates for which no data were collected in these cohorts, such
as diabetes.
In conclusion, a PS subclassification model analysis of 3
RCTs demonstrated that rhPDGF-BB/β-TCP-collagen was
at least as effective as autograft in achieving fusion in
patients who underwent ankle and hindfoot arthrodesis for
end-stage ankle arthritis, with less pain and fewer side
effects compared with treatment with autograft.
Foot & Ankle International - September 2019

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