INTRODUCTION Embryo implantation is a complex process involving maternal hormonal changes, immune responses and maturational events in the embryo. A pregnancy could fail when these events are not synchronized. It is speculated that in women, an elevation of natural killer (NK) cells may have an effect on reproductive performance, and NK cell levels in blood are currently being used as a diagnostic test to guide the initiation of therapies in patients with infertility. METHODS We conducted a systematic review to evaluate the (i) levels of NK cells in blood and endometrium in infertile versus fertile women, (ii) association between NK cells and IVF outcome, (iii) levels of NK cells in blood and endometrium in women with recurrent miscarriage (RM) versus controls. The following electronic databases were searched: Medline, EMBASE, Cochrane Library, Web of Science and National Research Register. RESULTS A total of 22 studies were included. Meta-analysis of studies that evaluated peripheral and uterine NK (uNK) cell percentages in infertile versus fertile women showed no significant difference between the two groups [standardized mean difference (SMD) -0.33; 95% confidence intervals (CI) -1.06, 0.4; P = 0.37; SMD -1.82; 95% CI -4.80, 1.17; P = 0.23 respectively]. Pooling of studies that reported peripheral NK cells as numbers showed significantly higher NK cell numbers in infertile women compared with fertile controls (SMD 3.16; 95% CI 1.07, 5.24; P = 0.003). Meta-analysis of studies that evaluated the role of NK cells in IVF outcome showed no significant difference in live birth rates in women with elevated NK cells or NK cell activity compared with women without elevated peripheral NK cells or NK cell activity (NK activity assessed using a cytotoxicity assay) (relative risk 0.57; 95% CI 0.06, 5.22; P = 0.62). Meta-analysis of studies that evaluated peripheral NK cell percentages in women with RM versus controls showed significantly higher NK cell percentages in women with RM (SMD 1.36; 95% CI 0.04, 2.69; P = 0.04). Meta-analysis of studies that evaluated peripheral NK cell numbers showed significantly higher NK cell numbers in women with RM compared with controls (SMD 0.81; 95% CI 0.47, 1.16; P < 0.00001). Meta-analysis of studies that evaluated uNK cells showed no significant difference in women with RM compared with controls (SMD 0.40; 95% CI -1.24, 2.04; P = 0.63). CONCLUSIONS Further research is needed before NK cell assessment can be recommended as a diagnostic tool in the context of female infertility or RM. There is no clear explanation as to why the results differ when data for NK cells are expressed as numbers or a percentage. On the basis of current evidence, NK cell analysis and immune therapy should be offered only in the context of clinical research.