Early versus deferred androgen suppression therapy for patients with lymph node-positive prostate cancer after local therapy with curative intent: a systematic review

Table 4 Risk of bias

	RTOG-85-31[9, 10]	Granfors et al.[11, 12]	EST-3886[13–15]	EPC program[16, 17]
random sequence generation	random number generator	not described	random number generator	random number generator
allocation concealment	central allocation	not described	central allocation	central allocation
blinding of participants/personnel	no	no	no (only pathologists were blinded)	double-blinded (placebo-controlled)
blinding of outcome assessment	unclear	unclear	unclear	unclear
incomplete outcome data	low risk (a)	low risk (a)	low risk (a)	low risk (a)
selective reporting	low risk (b)	high risk (c)	low risk (b)	low risk (b, d)
note/other bias	randomization of 977 patients but only 173 (18%) presented with lymph node-positive disease.	staging was retrospectively regraded to ensure comparable groups; initially planned for 400 patients but stopped after inclusion of 91 of which only 39 patients (43%) presented with lymph node-positive disease.	staging was retrospectively regraded to ensure comparable groups; initially planned for 220 lymph node-positive patients but stopped after inclusion of 100 of which only 98 were randomized	randomization of 8113 patients but only 150 (2%) presented lymph node-positive disease (radical prostatectomy: 74 patients, radiotherapy: 14 patients, watchful waiting: 62 patients).

(a). We found no evidence for missing outcome data for patients with node-positive prostate cancer. Additionally, survival/progression outcome data were presented by intention-to-treat.
(b). The study protocol is not available but we suggest that the published reports include all expected outcomes.
(c). One or more outcomes of interest are reported incompletely so that they cannot be entered in a meta-analysis.
(d). Authors reported data for adverse events in the subgroup of patients with node-positive prostate cancer inconsistently. However, adverse events were reported sufficiently for all patients included in the study in other reports, which were not eligible for this review.

ISSN: 1471-2407