EURURO Vol. 72 No. 4

599 656

considered for quality assessment and meta-analysis.
When critical information needed (eg, surgical procedure
or number of animals/patients) was incomplete, studies
were excluded. As only two studies used rats these were
also excluded at this stage.
2.5. Quality assessment
Due to the nonrandomized, noncontrolled nature of most
preclinical and clinical studies, no standard risk of bias
analysis could be performed as validated tools are unavail-
able for these types of studies. Instead, overall quality was
independently scored by PdJ and LV based on the reporting
of specific key information . Discrepancies were
discussed until agreement was reached.
2.6. Meta-analysis
The following main research question was considered:
‘‘What is the evidence for the efficacy of urethral tissue
engineering in preclinical and clinical studies?’’ Subques-
tions included the effects of the addition of (stem) cells to the
template, the type of biomaterial, as well as potential
differences between animal species on the separate outcome
measures. Analyses for preclinical and clinical studies were
conducted separately, as were full circumferential and inlay
procedures. Statistical analyses were performed with SAS/
STAT software version 9.2 for Windows, copyright 2002–
2008 by SAS Institute Inc., Cary, NC, USA.
2.6.1. Preclinical studies
The following preclinical data were extracted for all
available time points per study: the total number of animals
as well as the number of animals without side effects, with
functionality, and alive at the study endpoint. Time points
were categorized in three periods: 0–4 wk, 5–11 wk, and
12 wk or longer.
Per study, the probability of response (eg, having no side
effects) with a corresponding 95% exact (Clopper-Pearson)
confidence interval (CI) was estimated per outcome. An
additive random-effects logistic meta-regression model
was fitted by means of a generalized linear mixed model
approach. The number of responding animals out of the
total was used as outcome parameter. In addition, the
following independent parameters were used: treatment
(combining the addition of cells and the type of biomaterial)
and animal species. Random effects for study and for
treatment grouped by study, were added. The Akaike
Information Criterion showed that models based on
combined study data were preferable to models based on
the period data (period as factor), therefore all time points
per study were combined. When possible, the maximum
likelihood approach with adaptive quadrature was used as
estimation method. If this did not converge, the maximum
likelihood with the Laplace approximation was applied. The
resulting estimated odds were backtransformed into
percentages and corresponding 95% CIs. In addition, the
marginal effects of the treatments were estimated by
combining the estimated percentages for rabbits and dogs,
including 95% logit-based CIs, as described by Zou . All p
values were based on these CIs.
2.6.2. Clinical studies
For the analyses of the clinical outcomes, the following data
per study were extracted: total number of patients, and
numbers of patients without side effects, with functionality,
and completing the study. No separate time points were
analyzed in the human studies. For each study, the
probability of response with corresponding 95% exact CIs
was estimated per outcome. Due to limited study diversity,
meta-regression models similar to preclinical studies were
only fit for inlay repair and biomaterial type decellularized.
A compound symmetry random effect was added for the
addition of cells, grouped by study. Estimated odds from
meta-regression were backtransformed into probabilities
and corresponding 95% CIs.
3. Evidence synthesis
3.1. Literature search and screening
A and B show the results of the literature search and
screening of collected studies. After the search, 1524 unique
preclinical and 5361 unique clinical studies were identified.
During title and abstract screening of these studies,
1349 and 5282 were excluded, respectively. After full text
screening, 80 preclinical studies and 23 clinical studies were
included in the study characteristics table (see Section .
Only 63 preclinical and 13 clinical studies were eligible for
the quality assessment (Section and meta-analysis
(Section ).
3.2. Study characteristics
3.2.1. Preclinical studies
Preclinical study characteristics are summarized in
(see Supplementary data 3 for references of listed studies).
Only three animal species, rabbits (59/80), dogs (19/80), and
rats (2/80) were used, which were predominantly males
(72/80). Full circumferential repair was investigated in
41 studies, inlay repair in 30 studies, both methods in three,
while the procedure was unclear in the remaining studies
(6/80). In dogs, primarily full defect repairs were performed
(14 full vs 4 inlay), while in rabbits both inlay (25) and full
repairs (26) were employed.
Due to the wide variety of materials used, they were
categorized into three categories: decellularized templates
(46/80), de novo prepared templates from natural materials
(18/80), and de novo prepared templates from synthetic
materials (12/80). Four (4/80) studies used multiple
material types in different groups and these were assessed
separately in the meta-analysis. Synthetic materials were
almost exclusively used for full repair (10 full vs 3 inlay).
Cells were incorporated into templates in 34 studies, of
which bladder smooth muscle cells and urothelial cells
were mostly used (13/34), followed by keratinocytes and
fibroblasts from oral tissue (6/34) or a combination thereof
(2/34), foreskin epidermal cells (2/34) and omental
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 5 9 4 – 6 0 6 599