Abstract

BACKGROUND

Peripheral natural killer (pNK) and uterine NK (uNK) cells have been associated with reproductive failure. We systematically reviewed the literature to assess whether numbers or activity of pNK or uNK cells predicted subsequent pregnancy and outcome.

METHODS

We searched the electronic MEDLINE database from 1950 to April 2010 for relevant publications by using MeSH terms ‘natural killer cells’, ‘reproduction’ and ‘pregnancy complications’. We included studies that measured pre-pregnancy pNK and uNK cell numbers or activity in women with recurrent miscarriage (RM) or infertility, and reported subsequent pregnancy outcomes of miscarriage or failure to conceive after assisted reproductive technology (ART).

RESULTS

The search identified 783 publications and 12 fulfilled the inclusion criteria. There were too few women entered into the observational studies to assess whether high pNK cell percentages or activity predicted subsequent miscarriage in women with idiopathic RM (numbers: n = 32, OR 17, 95% CI 0.82–350.6, activity: n = 92, OR 2.51, 95% CI 0.16–40.29), or implantation failure (n = 203, OR 1.35, 95% CI 0.28–6.46), or miscarriage in infertile women after ART (n = 79, OR 2.48, 95% CI 0.50–12.32). Similarly, the studies of uNK cells were not large enough to assess whether abnormal uNK cell density predicted subsequent miscarriage in women with idiopathic RM (n = 72, OR 1.33, 95% CI 0.16–11.11). None of the uNK cell studies in women with infertility reported pregnancy outcomes dichotomized for uNK cell numbers.

CONCLUSIONS

The prognostic value of measuring pNK or uNK cell parameters remains uncertain. More studies are needed to confirm or refute the role of NK cell assessments as a predictive test for screening women who may benefit from immunotherapy.

Introduction

Immunological mechanisms have been thought to play a role in reproductive problems such as recurrent miscarriage (RM) (defined as three or more consecutive miscarriages), infertility and implantation failure. This implies that a successful pregnancy involves maternal adaptation of the immune response to the semi-allogenic developing embryo.

Natural killer (NK) cells are part of the innate immune system, and are found in both peripheral blood and endometrium. Although both peripheral NK (pNK) and uterine NK (uNK) cells express the surface antigen CD56, pNK cells are phenotypically and functionally different from uNK cells and <10% of pNK cells resemble uNK cells (Moffett-King, 2002). Furthermore 90% of pNK cells are CD56dim and CD16+ whereas 80% of uNK cells are CD56bright and CD16 (Nagler et al., 1989; King et al., 1991).

Peripheral NK cells (CD56dim) have been demonstrated to show significant cytotoxic activity with well-established antiviral and anti-neoplastic functions, while uNK cells have little cytotoxic activity, but are a rich source of cytokines, particularly angiogenic ones, with possible roles in regulation of trophoblast invasion and angiogenesis (Bulmer and Lash, 2005; Dosiou and Giudice, 2005). A high proportion of peripheral NK cells may not reflect the condition of the endometrium where implantation occurs and the mechanism of how these cells can be associated with miscarriage is unclear. In contrast large numbers of uNK cells appear in the mid-secretory phase although the mechanism is still not known. There are two theories: recruitment from pNK cells which subsequently differentiate in the uterine microenvironment into uNK cell phenotype through a series of organized processes, or uNK cells come from the in utero proliferation and differentiation of stem cells or endogenous NK cells in the endometrium (Bulmer and Lash, 2005; Kitaya et al., 2007). The former theory is the rationale for testing pNK cells although the latter is the more widely held view. In contrast to pNK cells, uNK cells are resident in the endometrium and constitute 70% of endometrial leucocytes, the most predominant leucocyte population during the time of implantation and early pregnancy (Bulmer et al., 1991). They are adjacent to fetal trophoblast cells in the maternal–fetal interface and express receptors such as killer-cell immunoglobulin-like receptor (KIR), immunoglobulin-like transcript-2 (ILT2) and NKG2, for fetal trophoblast antigens, human leucocyte antigen-C (HLA-C), HLA-E and HLA-G (Moffett-King, 2002). Thus, pNK and uNK cells have different surface antigens, functions and receptors and hence, should be considered separate entities.

There have been a series of case–control studies reporting an association between pNK cell numbers (Kwak et al., 1995; Ntrivalas et al., 2001; Yamada et al., 2003) or activity (Aoki et al., 1995; Shakhar et al., 2003) with RM, as well as some studies that have shown no difference in pNK cells parameters between RM and controls (Emmer et al., 2000; Souza et al., 2002; Wang et al., 2008). Similarly, there have been case–control studies reporting an association between uNK cells and RM (Clifford et al., 1999; Quenby et al., 1999, 2005; Tuckerman et al., 2007), but others, that have included women with only two miscarriages, have failed to find this association (Michimata et al., 2002; Shimada et al., 2004).

There is also inconsistency in the association of pNK cells and uNK cells with infertility. Some groups have found an association of pNK cells and infertility (Beer et al., 1996; Matsubayashi et al., 2001; Ntrivalas et al., 2001) while some have not (Vujisic et al., 2004). Likewise for uNK cells, one group has reported an association with infertility (Ledee-Bataille et al., 2005) and another has found no difference (Matteo et al., 2007).

Evidence for a causative role for NK cells test in reproductive problems would be considerably improved if the tests of pre-pregnancy NK cell numbers or activity predicted subsequent pregnancy outcome. Thus, our aim was to perform a systematic review of the current literature to ascertain the relationship between pre-pregnancy NK cell tests results and outcomes of miscarriage, live births or implantation failure in women with RM or infertility requiring assisted reproductive technology (ART).

Materials and Methods

Data sources

We searched the electronic MEDLINE database through OvidSP from 1950 to April 2010 for published literature in all languages. The MeSH terms ‘natural killer cells’, ‘reproduction' and ‘pregnancy complications’ were exploded. To identify relevant citations about NK cells, we used terms ‘CD56’, ‘uterus’, ‘uterine’, ‘endometrial’, ‘decidual’ and ‘peripheral’, in addition to ‘NK cells’. Search terms such as ‘abortion/miscarriage’, ‘ectopic pregnancy’, ‘fetal death’, ‘fertilization’, ‘insemination’, ‘live birth’, ‘pregnancy’, ‘pregnancy outcome’ and ‘stillbirth’, in relation to pregnancy outcomes were under the MeSH tree of ‘reproduction’ and ‘pregnancy complications’, to retrieve all papers relevant to NK cells and reproductive outcomes. The search was then limited to humans and females. Advice was sought from the Trials search coordinator of the Cochrane Collaboration Pregnancy and Childbirth Group with regards to the development of the search strategy protocol, who advised on the terminology and methods of searching. The Trials search coordinator was not involved in the study selection of the articles, or data extraction.

The abstracts for all the citations were retrieved and assessed for their suitability for inclusion. Papers that were published in other languages all had abstracts in English. Original articles of abstracts where relevance could not be judged from the abstract alone were obtained for detailed analysis. Additionally, the reference lists of the publications identified were examined for possible studies not included in the initial search.

Study selection

Review articles, letters and studies with no pregnancy outcomes reported were excluded after reading the abstracts. This review also excluded studies that reported on treating women with immunotherapy such as prednisolone or intravenous immunoglobulin (IVIG) as prednisolone reduces both uNK (Quenby et al., 2005) and pNK (Thum et al., 2008) cells and IVIG alters pNK cell parameters (Morikawa et al., 2001b), which may either positively or negatively affect the pregnancy outcome, and distort the results of this review. Furthermore, these treatments are still experimental without evidence from methodologically sound randomized controlled trials (RCTs).

Inclusion criteria were studies that identified NK cells using the CD56 marker, either CD56+, CD56bright or CD56dim, the CD69 activation marker or NK cells activity measured by Chromium51 release cytotoxicity assay, and investigated women with RM (defined as two or more consecutive miscarriages) or women with infertility seeking ART.

Two reviewers (A.T. and S.Q.) read through all the papers selected for detailed evaluation. Study quality was assessed using The Guidelines Manual 2009 published by National Institute for Health and Clinical Excellence (NICE). Information was obtained for the inclusion and exclusion criteria of women in the study, the source and method of analysing NK cells, the percentage/number and activity levels of NK cells, the level of normality of NK cells in the unit, and pregnancy outcomes.

Publications were divided into four groups according to the source of NK cells and type of reproductive failure: pNK cells test in RM, pNK cells test in infertility, uNK cells test in RM and uNK cells test in infertility. Pregnancy outcomes of implantation failure (defined as no positive pregnancy test after ART), miscarriage <24 weeks gestation, live births or implantation success (positive pregnancy test) leading to either miscarriage of <24 weeks gestation or live births were collected for women in all these groups. Data were extracted from texts, tables and graphs of each of the included studies. Original data in our unit were re-examined by classifying the pregnancy outcomes according to the cut-off of 5%, published in a later article (Quenby et al., 2005). When appropriate, meta-analyses were performed using Review Manager (RevMan) Version 5.0 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration 2008) on studies reporting pregnancy outcomes according to predetermined cut-offs for normality of NK cell parameters in similar patient groups. In the presence of significant heterogeneity, random effects have been used to pool the results. The results were reported according to the Meta-analysis of Observational Studies in Epidemiology (MOOSE) Guidelines (Stroup et al., 2000).

Results

There were altogether 783 citations identified. Seven hundred and eighty were from the search terms mentioned above and three were from the assessment of reference lists of these publications. There were no non-English language publications that were found to be relevant to this review. The selection process leading to the included publications in the review are presented in Fig. 1. After reading the abstracts, 756 articles were excluded and 27 full text articles regarding pNK and uNK cells were retrieved for evaluation. After detailed analysis, 15 publications were excluded: three were letters, 6 did not report relevant outcomes (Kwak et al., 1995; Ntrivalas et al., 2001; Michou et al., 2003; Putowski et al., 2004; Matteo et al., 2007; Thum et al., 2007), 4 reported on studies using the same group of women for slightly different aspects of NK cells (Emmer et al., 1999; Yamada et al., 2001; Morikawa et al., 2001a; Thum et al., 2004) and 2 reported immunotherapy use in some women (Coulam et al., 1995; Beer et al., 1996). A total of 12 publications met the criteria for analysis. There were seven studies reporting on pNK cells (Table I), and six studies reporting on uNK cells (Table II), as one study investigated both pNK and uNK cells. Five publications reporting on pNK cells and two publications on uNK cells presented pregnancy outcomes, dichotomized into groups of high and normal levels of NK cell parameters, established from controls, as described in Tables I and II.

Table I

Included studies investigating peripheral NK cells (seven studies).

Study n Inclusion criteria Method of analysis Results 
Recurrent miscarriage 
Aoki et al. (1995) 68 ≥2 miscarriages; idiopathic RM 51Cr release assay Higher % NK activity in patients who miscarried 
    Normal range—<41.8% NK activity determined by the mean +1 SD of 47 healthy controls with no history of miscarriages with pregnancy outcomes reported according to normal range 
Emmer et al. (2000) 142 ≥2 miscarriages; idiopathic RM Flow cytometry, 51Cr release assay Normal range—<12% NK cells taken from publications by other groups and <322 lytic units for NK cells activity (no mention of how this was obtained) with pregnancy outcomes reported according to normal range 
Yamada et al. (2003) 113 ≥2 miscarriages; all RM Flow cytometry, 51Cr release assay Higher % NK cells in miscarriage of normal karyotype and biochemical pregnancy compared with live birth 
    Normal range—<16.4% CD56+ cells and <46% NK activity determined by ROC curve for optimal discrimination between miscarriage (normal karyotype or biochemical pregnancy) and live birth 
Infertility 
Fukui et al. (1999) 85 NK cells Patients undergoing IVF Flow cytometry % NK cells 
 297 NK activity  51Cr release assay Pre-IVF cycle—No difference in % NK cell and subpopulation in women with infertility who failed to get pregnant and those who became pregnant after ART 
    % NK cell activity 
    Pre-IVF cycle—No difference in % NK cell activity in women with infertility who failed to get pregnant and those who became pregnant after ART, and between miscarriage and live birth in those who were pregnant 
    No normal range reported 
Thum et al. (2005) 138 Patients undergoing IVF Flow cytometry No difference in % NK cell and NK cell subpopulation in women with infertility who failed to get pregnant and those who became pregnant after ART, and between miscarriage and live birth in those who were pregnant 
    Normal range—<12% taken from publications by other groups with pregnancy outcome reported according to this normal range 
Matsubayashi et al. (2005) 94 Patients undergoing IVF 51Cr release assay Higher % NK cell activity in women with infertility who failed to get pregnant and those who became pregnant after ART 
    No difference in % NK cell activity between miscarriage and live birth in those who were pregnant 
    Normal range—<44% as determined by the mean +1 SD of 94 healthy, age-matched controls with pregnancy outcomes reported according to normal range 
Baczkowski and Kurzawa (2007) 58 Patients undergoing IVF Flow cytometry No difference in % NK cell in women with infertility who failed to get pregnant and those who became pregnant after ART 
    No normal range reported 
Study n Inclusion criteria Method of analysis Results 
Recurrent miscarriage 
Aoki et al. (1995) 68 ≥2 miscarriages; idiopathic RM 51Cr release assay Higher % NK activity in patients who miscarried 
    Normal range—<41.8% NK activity determined by the mean +1 SD of 47 healthy controls with no history of miscarriages with pregnancy outcomes reported according to normal range 
Emmer et al. (2000) 142 ≥2 miscarriages; idiopathic RM Flow cytometry, 51Cr release assay Normal range—<12% NK cells taken from publications by other groups and <322 lytic units for NK cells activity (no mention of how this was obtained) with pregnancy outcomes reported according to normal range 
Yamada et al. (2003) 113 ≥2 miscarriages; all RM Flow cytometry, 51Cr release assay Higher % NK cells in miscarriage of normal karyotype and biochemical pregnancy compared with live birth 
    Normal range—<16.4% CD56+ cells and <46% NK activity determined by ROC curve for optimal discrimination between miscarriage (normal karyotype or biochemical pregnancy) and live birth 
Infertility 
Fukui et al. (1999) 85 NK cells Patients undergoing IVF Flow cytometry % NK cells 
 297 NK activity  51Cr release assay Pre-IVF cycle—No difference in % NK cell and subpopulation in women with infertility who failed to get pregnant and those who became pregnant after ART 
    % NK cell activity 
    Pre-IVF cycle—No difference in % NK cell activity in women with infertility who failed to get pregnant and those who became pregnant after ART, and between miscarriage and live birth in those who were pregnant 
    No normal range reported 
Thum et al. (2005) 138 Patients undergoing IVF Flow cytometry No difference in % NK cell and NK cell subpopulation in women with infertility who failed to get pregnant and those who became pregnant after ART, and between miscarriage and live birth in those who were pregnant 
    Normal range—<12% taken from publications by other groups with pregnancy outcome reported according to this normal range 
Matsubayashi et al. (2005) 94 Patients undergoing IVF 51Cr release assay Higher % NK cell activity in women with infertility who failed to get pregnant and those who became pregnant after ART 
    No difference in % NK cell activity between miscarriage and live birth in those who were pregnant 
    Normal range—<44% as determined by the mean +1 SD of 94 healthy, age-matched controls with pregnancy outcomes reported according to normal range 
Baczkowski and Kurzawa (2007) 58 Patients undergoing IVF Flow cytometry No difference in % NK cell in women with infertility who failed to get pregnant and those who became pregnant after ART 
    No normal range reported 
Table II

Included studies investigating uNK cells (six studies).

Study n Inclusion criteria Method of analysis Results 
Recurrent miscarriage 
LaChapelle et al. (1996) 20 ≥3 miscarriages; idiopathic RM Flow cytometry No difference in % total NK cells in women with miscarriage and ongoing pregnancy 
    No normal range reported 
Quenby et al. (1999) 22 ≥3 miscarriages; idiopathic RM Immunohistochemistry Higher % NK cells in women with miscarriage compared with LB 
    Normal range—<5% determined by 75th percentile of 18 control women who were attending for sterilization with two or more pregnancies and no miscarriages 
Michimata et al. (2002) 17 ≥2 miscarriages; idiopathic RM Immunohistochemistry No difference in number of NK cells/10 high power fields in women with miscarriage and LB 
    No normal range reported 
Tuckerman et al. (2007) 87 ≥3 miscarriages; idiopathic RM Immunohistochemistry No difference in % NK cells in women with miscarriage and LB 
    Normal range—<13.8% determined by the 90th percentile of 10 control women with regular menstrual cycle not on hormonal contraception 
Infertility 
Fukui et al. (1999) 76 Patients undergoing IVF Flow cytometry No difference in % NK cells in women with infertility who failed to get pregnant and those who became pregnant after ART; and between miscarriage and LB in those who were pregnant 
    Higher subpopulation of %CD56bright NK cells in women who had LB compared with miscarriage 
    No normal range reported 
Ledee-Bataille et al. (2004) 15 ≥3 IVF cycles failures Immunohistochemistry No difference in number of NK cells/100x fields in natural IVF cycle in women with infertility who failed to get pregnant and those who became pregnant after ART 
    No normal range reported 
Study n Inclusion criteria Method of analysis Results 
Recurrent miscarriage 
LaChapelle et al. (1996) 20 ≥3 miscarriages; idiopathic RM Flow cytometry No difference in % total NK cells in women with miscarriage and ongoing pregnancy 
    No normal range reported 
Quenby et al. (1999) 22 ≥3 miscarriages; idiopathic RM Immunohistochemistry Higher % NK cells in women with miscarriage compared with LB 
    Normal range—<5% determined by 75th percentile of 18 control women who were attending for sterilization with two or more pregnancies and no miscarriages 
Michimata et al. (2002) 17 ≥2 miscarriages; idiopathic RM Immunohistochemistry No difference in number of NK cells/10 high power fields in women with miscarriage and LB 
    No normal range reported 
Tuckerman et al. (2007) 87 ≥3 miscarriages; idiopathic RM Immunohistochemistry No difference in % NK cells in women with miscarriage and LB 
    Normal range—<13.8% determined by the 90th percentile of 10 control women with regular menstrual cycle not on hormonal contraception 
Infertility 
Fukui et al. (1999) 76 Patients undergoing IVF Flow cytometry No difference in % NK cells in women with infertility who failed to get pregnant and those who became pregnant after ART; and between miscarriage and LB in those who were pregnant 
    Higher subpopulation of %CD56bright NK cells in women who had LB compared with miscarriage 
    No normal range reported 
Ledee-Bataille et al. (2004) 15 ≥3 IVF cycles failures Immunohistochemistry No difference in number of NK cells/100x fields in natural IVF cycle in women with infertility who failed to get pregnant and those who became pregnant after ART 
    No normal range reported 
Figure 1

The selection process of the included studies.

Figure 1

The selection process of the included studies.

Only three studies (Lachapelle et al., 1996; Quenby et al., 1999; Tuckerman et al., 2007), all investigating uNK cells, fit the definition of idiopathic RM where women were included after three consecutive miscarriages with no causes for the miscarriages found after routine investigations in their hospital. Three studies on pNK cells (Aoki et al., 1995; Emmer et al., 2000; Yamada et al., 2003) and one study on uNK cells (Michimata et al., 2002) included women after only two miscarriages, and one (Yamada et al., 2003) included women with known associations of RM such as endocrine disorders, antiphospholipid syndrome (APS) and thrombophilia. The group for infertility included all women who underwent IVF treatment, regardless of the cause for infertility, apart from one study (Ledee-Bataille et al., 2005), which included women after three cycles of implantation failure.

Peripheral NK cells

All studies on pNK cells used flow cytometry to investigate the numbers of pNK cells as a percentage of lymphocytes or leucocytes depending on the panel of monoclonal antibodies used to identify these cells, or 51Cr release cytotoxicity assay to assess NK cells activity.

There were three studies (Aoki et al., 1995; Emmer et al., 2000; Yamada et al., 2003) investigating pNK cells (number or activity) in women with RM, but only two studies included women with idiopathic RM (Aoki et al., 1995; Emmer et al., 2000). Emmer et al. (2000) reported pregnancy outcomes in relation to both number and activity of NK cells. It is likely that the same women were tested for both parameters and, therefore, these results were not pooled to avoid double counting. When the results of the different studies were pooled, there was significant heterogeneity between the studies with I2 = 85%. Therefore, random effects were used. High pNK cell numbers or activity did not predict miscarriage in a subsequent pregnancy in women with idiopathic RM as the meta-analysis of these studies did not reach statistical significance, although positive OR were found (n = 22, OR 17, 95% CI 0.82–350.60; n = 92, OR 2.51 95% CI 0.16–40.29; Fig. 2).

Figure 2

Odds of miscarriage with high pre-pregnancy peripheral NK cell parameters in women with idiopathic RM.

Figure 2

Odds of miscarriage with high pre-pregnancy peripheral NK cell parameters in women with idiopathic RM.

Of the four studies (Fukui et al., 1999; Matsubayashi et al., 2005; Thum et al., 2005; Baczkowski and Kurzawa, 2007) that investigated pNK cells in women with infertility, two (Matsubayashi et al., 2005; Thum et al., 2005) reported outcomes of implantation failure after ART, dichotomized into high and normal levels of pNK cells, but used different parameters of pNK cell numbers and activity. High pNK cell parameters did not predict subsequent implantation failure (n = 203, OR 1.35, 95% CI 0.28–6.46; Fig. 3). The same two studies (Matsubayashi et al., 2005; Thum et al., 2005) also reported outcomes of miscarriage after implantation success, and the effect of abnormal pNK cell test results on the risk of subsequent miscarriage after implantation success was also uncertain as the meta-analysis was not statistically significant (n = 79, OR 2.48, 95% CI 0.50–12.32; Fig. 4). There was again significant heterogeneity with I2 of 84% between studies reporting implantation failure and I2 of 46% for studies reporting miscarriage after implantation success, in women with infertility.

Figure 3

Odds of implantation failure after ART with high levels of pre-pregnancy peripheral NK cell parameters in women with infertility.

Figure 3

Odds of implantation failure after ART with high levels of pre-pregnancy peripheral NK cell parameters in women with infertility.

Figure 4

Odds of miscarriage (after implantation success from ART) with high levels of pre-pregnancy peripheral NK cell parameters in women with infertility.

Figure 4

Odds of miscarriage (after implantation success from ART) with high levels of pre-pregnancy peripheral NK cell parameters in women with infertility.

The other two studies (Fukui et al., 1999; Baczkowski and Kurzawa, 2007) that investigated women with infertility, did not report pregnancy outcomes according to a predetermined cut-off of normality. Fukui et al. (1999) reported no difference in pre-pregnancy pNK cells number and activity between women who failed to get pregnant and those who became pregnant after ART, and between miscarriage and live birth in those who were pregnant after ART. Baczkowski and Kurzawa (2007) also found no difference in the percentage of pNK cells in these two groups of women.

uNK cells

All the samples in the studies were timed from the LH surge as uNK cells population vary throughout the menstrual cycle: scanty in the proliferative phase then increasing in numbers after ovulation and through the secretary phase (Bulmer and Lash, 2005). Thus, the mid-luteal biopsy examines the endometrium in the implantation window of a non-conception cycle.

Four studies (Quenby et al., 1999; Michimata et al., 2002; Ledee-Bataille et al., 2004; Tuckerman et al., 2007) investigating uNK cells used immunohistochemistry of frozen or paraffin fixed sections with antibodies to CD56 to identify NK cells staining, and two studies (Lachapelle et al., 1996; Fukui et al., 1999) used enzymatic digestion of endometrium and then flow cytometry to identify C56bright or dim cells. uNK cells results were presented either as an absolute count of NK cells per 3 (Ledee-Bataille et al., 2004) and 10 high power fields (Michimata et al., 2002) or as a percentage of total stromal cells (Quenby et al., 1999; Tuckerman et al., 2007). There were no studies that reported pregnancy outcomes according to uNK cells activity.

There were four studies investigating uNK cells in women with idiopathic RM but only two (Quenby et al., 1999; Tuckerman et al., 2007) reported pregnancy outcomes according to high and normal levels of uNK cells. A meta-analysis, with high heterogeneity (I2 = 69%), found that uNK cell density did not predict pregnancy outcome (n = 72, OR 1.33, 95% CI 0.16–11.11; Fig. 5). The two studies that did not report dichotomized pregnancy outcomes found no differences in mean percentage of uNK cells in women who subsequently miscarried and those who had live births.

Figure 5

Odds of miscarriage with high levels of pre-pregnancy uNK cell numbers in women with idiopathic RM.

Figure 5

Odds of miscarriage with high levels of pre-pregnancy uNK cell numbers in women with idiopathic RM.

Two studies (Fukui et al., 1999; Ledee-Bataille et al., 2004), differing in study design to each other, investigated uNK cells in women with infertility, and both did not report pregnancy outcomes according to a predetermined cut-off of normality. One included women who underwent IVF treatment and reported NK cells as a percentage of lymphocytes. The other included women with implantation failure after three cycles of IVF, and reported NK cells as mean absolute numbers in three 100x fields. Both report no differences in uNK cells between women who failed to get pregnant and who became pregnant after ART. However, the study by Fukui et al. (1999) which analysed NK cells by flow cytometry, showed a higher percentage of the subpopulation CD56bright NK cells in women who had live births compared with women who miscarried.

Discussion

This systematic review did not demonstrate that abnormal pNK and uNK cell parameters predicted adverse pregnancy outcomes of miscarriage or implantation failure, in women with RM or infertility. The observational studies were individually and collectively underpowered to answer this important question. Assuming that, irrespective of the reason for NK testing, women with a normal NK cell count have miscarriage rate of around 30%, one would expect an increase in miscarriage rate by at least 10% (33% increase) when NK cells are high. At least 376 women per group would have to be followed until delivery (752 in total) to test this hypothesis with 80% power (α = 5%). This is considerably more than the largest individual study available, which investigated 126 women (Thum et al., 2005). There was also significant heterogeneity between studies in terms on inclusion criteria, methodology of NK cells analysis and outcome measurements.

Peripheral NK cells

It has been postulated that uNK cells originate from pNK cells, which subsequently differentiate in the uterine microenvironment into the uNK cell phenotype (Kitaya et al., 2007). Testing for pNK cells involves venous blood sampling at any time during the menstrual cycle as they have not been shown to fluctuate throughout the cycle (Pantazi et al., 2010), or with sampling in pregnancy. All studies used flow cytometry, the best method to analyse and quantify lymphocyte subsets (Dosiou and Giudice, 2005). Investigations have been reported in the pre-pregnancy period, during pregnancy or on the day of embryo transfer in infertility patients undergoing ART. However, only results from investigations done prior to pregnancy were analysed in this study as the prespecified aim of the study was to assess the association of pre-pregnancy NK testing and subsequent pregnancy outcomes.

All three studies of pNK cells and RM included women after only two miscarriages which does not fit the ESHRE (European Society of Human Reproduction and Embryology) definition of RM (Farquharson et al., 2005). Yamada et al. (2003) and Emmer et al. (2000) investigated both pNK cells number and activity, while Aoki et al. (1995) only studied pNK activity. Emmer et al. (2000) and Aoki et al. (1995) both studied women with idiopathic RM whereas Yamada et al. (2003), the study with the largest sample size in these group, included women with known associations of RM such as endocrine disorders, APS and thrombophilia. Although they reported that high pNK cells number and activity predicted subsequent biochemical miscarriage and miscarriage of normal karyotype, more than half of the women investigated had another possible contributing factor to their miscarriage, creating potential bias in the results.

There is also significant heterogeneity between studies which is not surprising given potentially important differences in the analysis and interpretation. The determination for the cut-off of normality in different studies was through different methods with different control groups. Emmer et al. (2000) used <12% pNK cell numbers as normal range set by Beer et al. (1996) with no explanation of how this level was calculated. Yamada et al. (2003) used <16.4% pNK cell numbers and <46% pNK cells activity as the normal range determined by receiver operating characteristic (ROC) curve for optimal discrimination between miscarriage (normal karyotype or biochemical pregnancy) and live birth. However, Aoki et al. (1995) set <41.8% pNK cells activity measurement as normal range determined by the mean +1 SD of 47 healthy controls with no history of miscarriages and Emmer et al. (2000) set <322 lytic units as normal with no mention of how this was obtained. The inconsistency was also seen in studies of women with infertility where Matsubayashi et al. (2005) determined <44% pNK activity as normal range with controls, although age-matched, not all of proven fertility, whilst Thum et al. (2005) used <12% pNK cell numbers as set by Beer et al. (1996). Therefore, it is clear that there is a lack of a commonly accepted normal range for pNK cells number and activity, or generally accepted type of pNK cells testing. In addition, there were also differences in IVF protocols in different units and the definition for implantation failure or success was not mentioned in some studies (Matsubayashi et al., 2005; Baczkowski and Kurzawa, 2007).

It is also known that pNK cells increase significantly with stress and exercise and this was not taken into account when blood was taken for investigation (Benschop et al., 1998; Timmons and Cieslak, 2008). Furthermore, the value of an abnormal test for pNK cells activity is also unknown as it may be a reflection of a transient stress response at the time of blood withdrawal, or a representation of the response to other stresses in daily life (Shakhar et al., 2006). It is unclear if these phenomena exist for uNK cells.

uNK cells

Testing of uNK cells involves an endometrial biopsy that can only be carried out in the pre-pregnancy period. Immunohistochemistry was the method used in most of the studies of uNK cells. This is more time consuming than flow cytometry but it reveals the location of the uNK cells (Bulmer et al., 1991). Analysis using flow cytometry involves digesting the tissue, and thereby potentially losing cells and antigens. Furthermore, it needs a large sample of endometrium that may be difficult to obtain in some women.

The evidence for the association between preimplantation uNK cell density and miscarriage in a subsequent pregnancy is limited. Two studies reported no difference in uNK cell density between women who subsequently miscarried and those who had live births while two other studies reported pregnancy outcomes according to high and normal uNK cell density had contradictory results (Quenby et al., 1999; Tuckerman et al., 2007). The studies that reported no difference in uNK cell density were different to each other and are not comparable. Michimata et al. (2002) included women after two miscarriages and used immunohistochemistry for analysis while LaChapelle et al. (1996) included women with three miscarriages and analysed NK cells using flow cytometry.

There was again considerable heterogeneity between the two studies that reported dichotomized pregnancy outcomes. Similarly to pNK cell parameters, the normal ranges were obtained with different control women. Quenby et al. (1999, 2005) used <5% uNK cells as a normal range based on the upper quartile of 18 control women while Tuckerman et al. (2007) defined <13.8% as normal determined by the 90th percentile of 10 control women. For analysis, Quenby et al. (1999) used frozen sections and pressure cooker for antigen retrieval while Tuckerman et al. (2007) used waxed embedded specimens and microwave for antigen retrieval. Furthermore, uNK cells are not evenly distributed through the tissues and their density varies depending on where the cells are counted. One study counted cells near the epithelial edge (Quenby et al., 1999) and the other at random, including deeper into the section (Tuckerman et al., 2007).

Both studies investigating uNK cells in women with infertility reported no difference in uNK cell density and percentage between women who failed to get pregnant and those who became pregnant after ART. However, the population of women in both studies was different as Fukui et al. (1999) included all women undergoing ART while Ledee-Bataille et al. (2004) included women without pregnancies after three cycles of ART. In addition, the method of analysis was different where one employed flow cytometry and the other immunohistochemistry. Thus, the studies are not comparable enough to draw conclusions about the implications of uNK cells tests in women with infertility.

Despite the lack of clinical evidence, there is a biological plausibility for a role for uNK cells in reproductive failure. uNK cells are most numerous in the implantation window and in early pregnancy (Bulmer and Lash, 2005), and they are adjacent to and interact with extravillous trophoblast cells (Moffett-King, 2002). Different uNK cell populations have been found in the deciduas of normal and miscarried early pregnancy (Quack et al., 2001). Furthermore, uNK cells have been shown to regulate angiogenesis (Hanna et al., 2006; Kalkunte et al., 2009; Quenby et al., 2009), an important factor in implantation, and trophoblast cells express antigens that are recognized by the receptors on uNK cells, resulting in changes during the implantation process which may affect pregnancy outcome (Hiby et al., 2010).

Conclusions

This review suggests that the prognostic value of measuring pNK and uNK cell numbers or activity remains uncertain as these parameters have not been shown to be associated with subsequent pregnancy outcome. This finding is similar to that of the many conditions that have been associated with RM such as thrombophilia and structural uterine anomalies, none of which have been shown to predict pregnancy outcome.

This may be because of the disappointingly small number of studies reporting clinical outcomes on small numbers of women. The inclusion criteria were also inappropriate in studies that investigated women after only two miscarriages. Furthermore, there is still no consensus on what an abnormal NK cell test result is, as the normal ranges in different studies are derived from different controls. There is a need for more studies investigating pNK and uNK cells population and function in women with reproductive problems. Ideally, future studies should be prospective, with appropriate inclusion criteria and have a standardized methodology of analysing and reporting pNK and uNK test results. Sample size should also be calculated to avoid lack of power in the study.

Before the availability of results from these larger, more methodologically sound evaluations of prognostic value or RCTs of therapeutic modalities on specifically selected women with potential immunological pathology, women with reproductive problems should not be offered NK testing in routine clinical practice, and prescribed empirical immunotherapy, without clear evidence of benefit. An alternative would be to counsel them about the lack of available evidence, and encourage them to participate in well-designed studies, to confirm or refute the role of NK cells as a clinically useful marker for screening. We echo the Cochrane review on immunotherapy for RM that quotes ‘a specific assay to diagnose immune-mediated early pregnancy loss and a reliable method to determine which women might benefit from manipulation of the maternal immune system are urgently needed’ (Porter et al., 2006).

Authors' roles

A.T., S.Q. and Z.A. conceived the idea, developed the methodology and analysed and interpreted the data. A.T. and S.Q. reviewed all the original papers for inclusion and exclusion. A.T. performed the searches, read the abstracts and wrote the initial draft. S.Q. and Z.A. critically revised the manuscript. All the authors approved the final version of the manuscript.

Acknowledgements

We would like to thank Lynn Hampson, the Trial Search Co-ordinator for the Cochrane Pregnancy and Childbirth Group for her advice on the search methodology.

References

Aoki
K
Kajiura
S
Matsumoto
Y
Ogasawara
M
Okada
S
Yagami
Y
Gleicher
N
Preconceptional natural-killer-cell activity as a predictor of miscarriage
Lancet
 , 
1995
, vol. 
345
 (pg. 
1340
-
1342
)
Baczkowski
T
Kurzawa
R
Immunophenotypic profiles of peripheral blood lymphocytes on the day of embryo transfer in women undergoing in vitro fertilization
Folia Histochem Cytobiol
 , 
2007
, vol. 
45
 
Suppl 1
(pg. 
S73
-
S77
)
Beer
AE
Kwak
JY
Ruiz
JE
Immunophenotypic profiles of peripheral blood lymphocytes in women with recurrent pregnancy losses and in infertile women with multiple failed in vitro fertilization cycles
Am J Reprod Immunol
 , 
1996
, vol. 
35
 (pg. 
376
-
382
)
Benschop
R
Geenen
R
Mills
P
Naliboff
B
Kiecolt-Glaser
J
Herbert
T
Van Der Pompe
G
Miller
G
Matthews
K
Godaert
G
, et al.  . 
Cardiovascular and immune responses to acute psychological stress in young and old women: a meta-analysis
Psychosom Med
 , 
1998
, vol. 
60
 (pg. 
290
-
296
)
Bulmer
JN
Lash
GE
Human uterine natural killer cells: a reappraisal. [Review] [101 refs]
Mol Immunol
 , 
2005
, vol. 
42
 (pg. 
511
-
521
)
Bulmer
JN
Morrison
L
Longfellow
M
Ritson
A
Pace
D
Granulated lymphocytes in human endometrium: histochemical and immunohistochemical studies
Hum Reprod
 , 
1991
, vol. 
6
 (pg. 
791
-
798
)
Clifford
K
Flanagan
AM
Regan
L
Endometrial CD56+ natural killer cells in women with recurrent miscarriage: a histomorphometric study
Hum Reprod
 , 
1999
, vol. 
14
 (pg. 
2727
-
2730
)
Coulam
CB
Goodman
C
Roussev
RG
Thomason
EJ
Beaman
KD
Systemic CD56+ cells can predict pregnancy outcome
Am J Reprod Immunol
 , 
1995
, vol. 
33
 (pg. 
40
-
46
)
Dosiou
C
Giudice
LC
Natural killer cells in pregnancy and recurrent pregnancy loss: endocrine and immunologic perspectives
Endocr Rev
 , 
2005
, vol. 
26
 (pg. 
44
-
62
)
Emmer
PM
Veerhoek
M
Nelen
WL
Steegers
EA
Joosten
I
Natural killer cell reactivity and HLA-G in recurrent spontaneous abortion
Transplant Proc
 , 
1999
, vol. 
31
 (pg. 
1838
-
1840
)
Emmer
PM
Nelen
WL
Steegers
EA
Hendriks
JC
Veerhoek
M
Joosten
I
Peripheral natural killer cytotoxicity and CD56(pos)CD16(pos) cells increase during early pregnancy in women with a history of recurrent spontaneous abortion
Hum Reprod
 , 
2000
, vol. 
15
 (pg. 
1163
-
1169
)
Farquharson
RG
Jauniaux
E
Exalto
N
Updated and revised nomenclature for description of early pregnancy events
Hum Reprod
 , 
2005
, vol. 
20
 (pg. 
3008
-
3011
)
Fukui
A
Fujii
S
Yamaguchi
E
Kimura
H
Sato
S
Saito
Y
Natural killer cell subpopulations and cytotoxicity for infertile patients undergoing in vitro fertilization
Am J Reprod Immunol
 , 
1999
, vol. 
41
 (pg. 
413
-
422
)
Hanna
J
Goldman-Wohl
D
Hamani
Y
Avraham
I
Greenfield
C
Natanson-Yaron
S
Prus
D
Cohen-Daniel
L
Arnon
TI
Manaster
I
, et al.  . 
Decidual NK cells regulate key developmental processes at the human fetal-maternal interface
Nat Med
 , 
2006
, vol. 
12
 (pg. 
1065
-
1074
)
Hiby
SE
Apps
R
Sharkey
AM
Farrell
LE
Gardner
L
Mulder
A
Claas
FH
Walker
JJ
Redman
CW
Morgan
L
, et al.  . 
Maternal activating KIRs protect against human reproductive failure mediated by fetal HLA-C2
J Clin Invest
 , 
2010
, vol. 
120
 (pg. 
4102
-
4110
)
Kalkunte
SS
Mselle
TF
Norris
WE
Wira
CR
Sentman
CL
Sharma
S
Vascular endothelial growth factor C facilitates immune tolerance and endovascular activity of human uterine NK cells at the maternal-fetal interface
J Immunol
 , 
2009
, vol. 
182
 (pg. 
4085
-
4092
)
King
A
Balendran
N
Wooding
P
Carter
NP
Loke
YW
CD3− leukocytes present in the human uterus during early placentation: phenotypic and morphologic characterization of the CD56++ population
Dev Immunol
 , 
1991
, vol. 
1
 (pg. 
169
-
190
)
Kitaya
K
Yamaguchi
T
Yasuo
T
Okubo
T
Honjo
H
Post-ovulatory rise of endometrial CD16(−) natural killer cells: in situ proliferation of residual cells or selective recruitment from circulating peripheral blood?
J Reprod Immunol
 , 
2007
, vol. 
76
 (pg. 
45
-
53
)
Kwak
JY
Beaman
KD
Gilman-Sachs
A
Ruiz
JE
Schewitz
D
Beer
AE
Up-regulated expression of CD56+, CD56+/CD16+, and CD19+ cells in peripheral blood lymphocytes in pregnant women with recurrent pregnancy losses
Am J Reprod Immunol
 , 
1995
, vol. 
34
 (pg. 
93
-
99
)
Lachapelle
MH
Miron
P
Hemmings
R
Roy
DC
Endometrial T, B, and NK cells in patients with recurrent spontaneous abortion. Altered profile and pregnancy outcome
J Immunol
 , 
1996
, vol. 
156
 (pg. 
4027
-
4034
)
Ledee-Bataille
N
Dubanchet
S
Kadoch
J
Castelo-Branco
A
Frydman
R
Chaouat
G
Controlled natural in vitro fertilization may be an alternative for patients with repeated unexplained implantation failure and a high uterine natural killer cell count
Fertil Steril
 , 
2004
, vol. 
82
 (pg. 
234
-
236
)
Ledee-Bataille
N
Bonnet-Chea
K
Hosny
G
Dubanchet
S
Frydman
R
Chaouat
G
Role of the endometrial tripod interleukin-18, -15, and -12 in inadequate uterine receptivity in patients with a history of repeated in vitro fertilization-embryo transfer failure
Fertil Steril
 , 
2005
, vol. 
83
 (pg. 
598
-
605
)
Matsubayashi
H
Hosaka
T
Sugiyama
Y
Suzuki
T
Arai
T
Kondo
A
Sugi
T
Izumi
S
Makino
T
Increased natural killer-cell activity is associated with infertile women
Am J Reprod Immunol
 , 
2001
, vol. 
46
 (pg. 
318
-
322
)
Matsubayashi
H
Shida
M
Kondo
A
Suzuki
T
Sugi
T
Izumi
S
Hosaka
T
Makino
T
Preconception peripheral natural killer cell activity as a predictor of pregnancy outcome in patients with unexplained infertility
Am J Reprod Immunol
 , 
2005
, vol. 
53
 (pg. 
126
-
131
)
Matteo
MG
Greco
P
Rosenberg
P
Mestice
A
Baldini
D
Falagario
T
Martino
V
Santodirocco
M
Massenzio
F
Castellana
L
, et al.  . 
Normal percentage of CD56bright natural killer cells in young patients with a history of repeated unexplained implantation failure after in vitro fertilization cycles
Fertil Steril
 , 
2007
, vol. 
88
 (pg. 
990
-
993
)
Michimata
T
Ogasawara
MS
Tsuda
H
Suzumori
K
Aoki
K
Sakai
M
Fujimura
M
Nagata
K
Nakamura
M
Saito
S
Distributions of endometrial NK cells, B cells, T cells, and Th2/Tc2 cells fail to predict pregnancy outcome following recurrent abortion
Am J Reprod Immunol
 , 
2002
, vol. 
47
 (pg. 
196
-
202
)
Michou
VI
Kanavaros
P
Athanassiou
V
Chronis
GB
Stabamas
S
Tsilivakos
V
Fraction of the peripheral blood concentration of CD56+/CD16−/CD3− cells in total natural killer cells as an indication of fertility and infertility
Fertil Steril
 , 
2003
, vol. 
80
 (pg. 
691
-
697
)
Moffett-King
A
Natural killer cells and pregnancy
Nat Rev Immunol
 , 
2002
, vol. 
2
 (pg. 
656
-
663
)
Morikawa
M
Yamada
H
Kato
EH
Shimada
S
Ebina
Y
Yamada
T
Sagawa
T
Kobashi
G
Fujimoto
S
NK cell activity and subsets in women with a history of spontaneous abortion. Cause, number of abortions, and subsequent pregnancy outcome
Gynecol Obstet Invest
 , 
2001
, vol. 
52
 (pg. 
163
-
167
)
Morikawa
M
Yamada
H
Kato
EH
Shimada
S
Kishi
T
Yamada
T
Kobashi
G
Fujimoto
S
Massive intravenous immunoglobulin treatment in women with four or more recurrent spontaneous abortions of unexplained etiology: down-regulation of NK cell activity and subsets
Am J Reprod Immunol
 , 
2001
, vol. 
46
 (pg. 
399
-
404
)
Nagler
A
Lanier
LL
Cwirla
S
Phillips
JH
Comparative studies of human FcRIII-positive and negative natural killer cells
J Immunol
 , 
1989
, vol. 
143
 (pg. 
3183
-
3191
)
Ntrivalas
EI
Kwak-Kim
JY
Gilman-Sachs
A
Chung-Bang
H
Ng
SC
Beaman
KD
Mantouvalos
HP
Beer
AE
Status of peripheral blood natural killer cells in women with recurrent spontaneous abortions and infertility of unknown aetiology
Hum Reprod
 , 
2001
, vol. 
16
 (pg. 
855
-
861
)
Pantazi
A
Tzonis
P
Perros
G
Graphou
O
Keramitsoglou
T
Koussoulakos
S
Margaritis
L
Varla-Leftherioti
M
Comparative analysis of peripheral natural killer cells in the two phases of the ovarian cycle
Am J Reprod Immunol
 , 
2010
, vol. 
63
 (pg. 
46
-
53
)
Porter
TF
Lacoursiere
Y
Scott
JR
Immunotherapy for recurrent miscarriage
Cochrane Database Syst Rev
 , 
2006
pg. 
CD000112
 
Putowski
L
Darmochwal-Kolarz
D
Rolinski
J
Oleszczuk
J
Jakowicki
J
The immunological profile of infertile women after repeated IVF failure (preliminary study)
Eur J Obstet Gynecol Reprod Biol
 , 
2004
, vol. 
112
 (pg. 
192
-
196
)
Quack
KC
Vassiliadou
N
Pudney
J
Anderson
DJ
Hill
JA
Leukocyte activation in the decidua of chromosomally normal and abnormal fetuses from women with recurrent abortion
Hum Reprod
 , 
2001
, vol. 
16
 (pg. 
949
-
955
)
Quenby
S
Bates
M
Doig
T
Brewster
J
Lewis-Jones
DI
Johnson
PM
Vince
G
Pre-implantation endometrial leukocytes in women with recurrent miscarriage
Hum Reprod
 , 
1999
, vol. 
14
 (pg. 
2386
-
2391
)
Quenby
S
Kalumbi
C
Bates
M
Farquharson
R
Vince
G
Prednisolone reduces preconceptual endometrial natural killer cells in women with recurrent miscarriage
Fertil Steril
 , 
2005
, vol. 
84
 (pg. 
980
-
984
)
Quenby
S
Nik
H
Innes
B
Lash
G
Turner
M
Drury
J
Bulmer
J
Uterine natural killer cells and angiogenesis in recurrent reproductive failure
Hum Reprod
 , 
2009
, vol. 
24
 (pg. 
45
-
54
)
Shakhar
K
Ben-Eliyahu
S
Loewenthal
R
Rosenne
E
Carp
H
Differences in number and activity of peripheral natural killer cells in primary versus secondary recurrent miscarriage
Fertil Steril
 , 
2003
, vol. 
80
 (pg. 
368
-
375
)
Shakhar
K
Rosenne
E
Loewenthal
R
Shakhar
G
Carp
H
Shamgar
B.-E
High NK cell activity in recurrent miscarriage: what are we really measuring?
Hum Reprod
 , 
2006
, vol. 
21
 (pg. 
2421
-
2425
)
Shimada
S
Kato
EH
Morikawa
M
Iwabuchi
K
Nishida
R
Kishi
R
Onoe
K
Minakami
H
Yamada
H
No difference in natural killer or natural killer T-cell population, but aberrant T-helper cell population in the endometrium of women with repeated miscarriage
Hum Reprod
 , 
2004
, vol. 
19
 (pg. 
1018
-
1024
)
Souza
SS
Ferriani
RA
Santos
CM
Voltarelli
JC
Immunological evaluation of patients with recurrent abortion
J Reprod Immunol
 , 
2002
, vol. 
56
 (pg. 
111
-
121
)
Stroup
DF
Berlin
JA
Morton
SC
Olkin
I
Williamson
GD
Rennie
D
Moher
D
Becker
BJ
Sipe
TA
Thacker
SB
Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group
J Am Med Assoc
 , 
2000
, vol. 
283
 (pg. 
2008
-
2012
)
Thum
MY
Bhaskaran
S
Abdalla
HI
Ford
B
Sumar
N
Shehata
H
Bansal
AS
An increase in the absolute count of CD56dimCD16+CD69+ NK cells in the peripheral blood is associated with a poorer IVF treatment and pregnancy outcome
Hum Reprod
 , 
2004
, vol. 
19
 (pg. 
2395
-
2400
)
Thum
MY
Bhaskaran
S
Bansal
AS
Shehata
H
Ford
B
Sumar
N
Abdalla
HI
Simple enumerations of peripheral blood natural killer (CD56+ NK) cells, B cells and T cells have no predictive value in IVF treatment outcome
Hum Reprod
 , 
2005
, vol. 
20
 (pg. 
1272
-
1276
)
Thum
MY
Abdalla
HI
Bhaskaran
S
Harden
EL
Ford
B
Sumar
N
Shehata
H
Bansal
A
The relationship of systemic TNF-alpha and IFN-gamma with IVF treatment outcome and peripheral blood NK cells
Am J Reprod Immunol
 , 
2007
, vol. 
57
 (pg. 
210
-
217
)
Thum
MY
Bhaskaran
S
Abdalla
HI
Ford
B
Sumar
N
Bansal
A
Prednisolone suppresses NK cell cytotoxicity in vitro in women with a history of infertility and elevated NK cell cytotoxicity
Am J Reprod Immunol
 , 
2008
, vol. 
59
 (pg. 
259
-
265
)
Timmons
BW
Cieslak
T
Human natural killer cell subsets and acute exercise: a brief review
Exerc Immunol Rev
 , 
2008
, vol. 
14
 (pg. 
8
-
23
)
Tuckerman
E
Laird
SM
Prakash
A
Li
TC
Prognostic value of the measurement of uterine natural killer cells in the endometrium of women with recurrent miscarriage
Hum Reprod
 , 
2007
, vol. 
22
 (pg. 
2208
-
2213
)
Vujisic
S
Lepej
SZ
Aksamija
A
Jerkovic
L
Sokolic
B
Kupesic
S
Vince
A
B- and T-cells in the follicular fluid and peripheral blood of patients undergoing IVF/ET procedures
Am J Reprod Immunol
 , 
2004
, vol. 
52
 (pg. 
379
-
385
)
Wang
Q
Li
TC
Wu
YP
Cocksedge
KA
Fu
YS
Kong
QY
Yao
SZ
Reappraisal of peripheral NK cells in women with recurrent miscarriage
Reprod Biomed Online
 , 
2008
, vol. 
17
 (pg. 
814
-
819
)
Yamada
H
Kato
EH
Kobashi
G
Ebina
Y
Shimada
S
Morikawa
M
Sakuragi
N
Fujimoto
S
High NK cell activity in early pregnancy correlates with subsequent abortion with normal chromosomes in women with recurrent abortion
Am J Reprod Immunol
 , 
2001
, vol. 
46
 (pg. 
132
-
136
)
Yamada
H
Morikawa
M
Kato
EH
Shimada
S
Kobashi
G
Minakami
H
Pre-conceptional natural killer cell activity and percentage as predictors of biochemical pregnancy and spontaneous abortion with normal chromosome karyotype
Am J Reprod Immunol
 , 
2003
, vol. 
50
 (pg. 
351
-
354
)