EURURO Vol. 71 No. 5

816 844

of the evidence, and the authors concluded that there was
an urgent need to conduct a large, robust, multicenter RCT
to address these shortcomings. Pickard et al published
the results of such an RCT in 1167 patients and found no
evidence that either tamsulosin or nifedipine increased the
rate of spontaneous stone passage compared to placebo. The
results were consistent across subgroup and sensitivity
analyses.
We compare the RCT by Pickard et al to the MA with
the most studies, by Seitz et al , to explore and discuss
discordant findings. Most RCTs included in the Seitz MA
were small and recruited from a single center; only six of 35
(17%) recruited more than 100 patients. The majority had
low internal validity and only one RCT reported allocation
concealment. As small RCTs may report larger effect sizes
compared to larger RCTs, an MA of small RCTs can lead to
biased estimates of treatment effects . Seitz et al also
found evidence of publication bias, which can lead to
overestimation of treatment effects and compromise the
validity of the MA findings .
Seitz et al found evidence of clinical heterogeneity
concerning the patient inclusion criteria, stone character-
istics, intervention, treatment in the control group, and
outcome measurement. In the MA, the primary outcome of
being stone-free was inconsistently defined, assessed using
different imaging modalities, and measured at a variety of
time points. In the RCT of Pickard et al the primary
outcome was any need for further intervention within 4 wk
of randomization, which is compared here to being stone-
free. In the control group of the Pickard RCT, 80% of patients
were stone-free, whereas in the Seitz review, stone-free
rates ranged from 4% to 78%, which highlights the potential
impact of heterogeneity in the studies included.
With contrasting primary outcomes and different base-
line event rates in the control groups, it is not surprising that
the RCT and MA reported discordant findings. The choice of
primary outcome is clearly of paramount importance in any
trial. Heterogeneity in the conduct, design, and reporting of
trials in this MA makes pooled treatment effects difficult, if
not impossible, to interpret.
5.2. Partial versus radical nephrectomy
In a European Organization for Research and Treatment of
Cancer (EORTC) RCT involving 541 patients with a solitary
T1–2N0M0 renal tumor of 5 cm, 21 patients progressed,
nine after radical nephrectomy (RN) and 12 after partial
nephrectomy (PN). An intent-to-treat analysis found an
overall survival (OS) advantage in favor of RN (hazard ratio
[HR] 1.5, p = 0.03); however, only 12 of the 117 deaths were
due to kidney cancer, four after RN and eight after PN
Subsequently, Kim et al published an SR and MA
including some 41 000 patients and found statistically
significant improvements in both OS (HR 0.81, p < 0.001)
and disease-specific survival (HR 0.71, p < 0.001), but this
time in favor of PN How can this discordance be
explained?
The Kim MA has a number of limitations. First, the
38 trials included were mostly retrospective, single-center
studies. The only RCT was the EORTC study. No information
was provided about the distribution of follow-up or patient
characteristics by treatment group (T category when >T1,
tumor size, tumor grade, cell type, or renal function).
Consequently, the differences in survival observed may not
be directly due to differences in treatment efficacy. In
addition, it is not clear to which patients the results can be
generalized. Lastly, there was significant heterogeneity in
the size of the treatment effect across the studies, so the
overall estimate of the HR is not meaningful. Nevertheless,
the EORTC RCT also had limitations and should be
interpreted with caution: 55 patients crossed over to the
other randomized treatment, 140 patients were clinically or
pathologically ineligible, and there were few cancer-related
events.
The MA found that PNwas associated with a lower risk of
severe chronic kidney disease (CKD); however, the EORTC
study only found lower incidence of at least moderate renal
dysfunction, not of advanced kidney disease or renal failure,
and this was not associated with a corresponding difference
in survival . The studies in theMA did not always specify
the status of the contralateral kidney, whereas in the EORTC
study the contralateral kidney had to be normal.
Critical information regarding the biases of the studies
included in the SR were not made explicit, since a Grading of
Recommendations, Assessment, Development, and Evalua-
tion (GRADE) approach was not used to assess the
quality of the evidence. The quality of the studies in the SR
and the heterogeneity of the results call into question the
validity of the conclusions of the MA, which should thus be
viewed with skepticism. In the same year, another SR
suggested that localized renal cell carcinomas are best
managed by PN where technically feasible. However, the
evidence base had significant limitations owing to studies of
low methodological quality and high RoB .
Further nonrandomized studies have found improved
survival after PN and a reduction in the risk of
cardiovascular events relative to RN; however, patients
chosen for PN had a higher baseline likelihood of long-term
survival . In another study, only patients with stage II
CKD had a lower risk of developing significant renal
impairment after PN . More recently, an SR and MA
of 21 nonrandomized comparative studies in patients with
clinical T1b and T2 renal tumors found better tumor control
and survival for PN compared to RN , but the SR is
subject to the same biases as the Kim MA.
Taking into account all the available efficacy data and a
perceived advantage in renal function, the 2016 EAU
Guidelines recommend, with several exceptions, that
localized renal cancers are better managed by PN than RN.
6. Discussion
It is generally accepted that a high-quality SR of RCTs and an
associated MA can provide a higher LE than a single RCT
addressing the same question . This can be problematic,
however, when the results of the MA are in direct conflict
with the RCT, making it difficult for guideline organizations
to interpret the evidence and issue recommendations.
E U R O P E A N U R O L O G Y 7 1 ( 2 0 1 7 ) 8 1 1 – 8 1 9 816