Sixty-eight studies (number of participants not reported) were included in the review; three studies reported data for two different subgroups which were counted as separate studies. Sixty-eight studies reported data on detection rate and 18 reported data on false negative rate. Approximately 25% of the included studies reported consecutive recruitment (most of the others appeared to be retrospective). Approximately 60% of the included studies recruited an appropriate spectrum of patients (most of the others excluded T3 stage patients). Most (>85%) studies provided a complete description of the index test. Approximately 40% of studies used an appropriate reference standard regardless of the sentinel node biopsy result.
Detection rate: Pooled detection rates for patients with and without previous surgical biopsy were 91.3 (95% CI 89.2% to 93.1%) and 92.8% (95% CI 91.1% to 94.3%). Odds ratio and risk difference indicated no statistically significant difference in detection failure between the surgical and non-surgical biopsy groups; Ι² values indicated low to moderate between-study heterogeneity for both values.
Subgroup analyses indicated that superficial injection or use of radiotracer could result in a statistically significant difference in detection rates (OR 1.9, 95% CI 1.2 to 3.16; 16 studies) and (OR 2.8, 95% CI 1.7 to 4.5; 10 studies). Publication year had no significant effect on odds ratio. Publication bias was considered to be a concern for this analysis.
False negative rate: Pooled false negative rates for patients with and without previous surgical biopsy were 12.3% (95% CI 9.2% to 16.2%) and 9.9% (95% CI 7.7% to 12.7%). The odds ratio and risk difference of having false negative results in patients with a history of previous surgical biopsy compared to those without were 1.4 (95% CI 1.1 to 1.8, p=0.006) and 0.02 (95% CI 0.003 to 0.04, p=0.023) which meant that false negative results were statistically significantly less likely to occur in patients without a history of previous surgical biopsy. Ι² values indicated no between-study heterogeneity for both values.
Subgroup analyses indicated that biopsy method and depth of injection may affect the difference in false negative rates between the two groups. Publication year had no significant effect on odds ratio. Publication bias was not considered to be a concern for this analysis.