This review concluded that soluble human leukocyte antigen-G in embryo cultures had moderate ability to predict pregnancy in women undergoing infertility treatment, and had better performance with good quality embryos. The review and analysis methods seemed appropriate, but were limited by lack of quality assessment. Given the variation in diagnostic accuracy, the clinical relevance of this conclusion is in doubt.

To evaluate the diagnostic accuracy of soluble human leukocyte antigen-G for predicting the achievement of clinical pregnancy in women undergoing infertility treatment.

EMBASE and MEDLINE were searched to September 2007, search terms were reported. A manual search of references was also performed and there were no language restrictions.

Cohort studies of soluble human leukocyte antigen-G in embryo supernatants in women undergoing in vitro fertilisation (IVF) or intracytoplasmic sperm injection (ICSI) were eligible for inclusion. The outcome of interest was clinical pregnancy, which was defined as at least one gestational sac observed on ultrasound. Studies not reporting clinical pregnancy as a separate outcome measure, or providing insufficient data to create a 2x2 table, were excluded.

The included studies were all prospective cohort studies; around half were of ICSI treatment only, the rest included both IVF and ICSI. All except two studies used multiple embryo transfer (one did not report this and one was both single and multiple), with a median transfer of three days after ovum pick-up (range two to five days). Most assays used were ELISA sandwich (one was Luminex). The cut point for dichotomising soluble human leukocyte antigen-G results varied between studies, but the most common was a detection limit of 1ng/mL; some studies used optical density.

The authors did not state how studies were selected for the review or how many reviewers performed the selection.

The authors did not state that they assessed validity.

Data on women's age, clinical pregnancy rates, embryo transfer characteristics, culture conditions, assay protocols, and true/false positive and negative results were extracted by two reviewers independently, and rechecked by reviewing the data. Sensitivity, specificity, positive likelihood ratios, negative likelihood ratios and diagnostic odds ratios (DOR), with 95% confidence intervals (CI), were calculated for each study for the ability of soluble human leukocyte antigen-G to predict clinical pregnancy. If one value was zero for a study, then 0.5 was added to each cell of the 2x2 table.

The sensitivity and specificity for each study (with 95% confidence intervals) were plotted in a receiver operating characteristic plot. Heterogeneity was assessed through visual inspection of the receiver operating characteristic plots and by using Cochran's Q test (with p<0.10 indicating significance) and the I² statistic. If heterogeneity was low or moderate (I² less than 50%), weighted estimates of diagnostic accuracy measures were calculated using inverse variance weights.

Diagnostic odds ratios were pooled using a random-effects model (DerSimonian and Laird). Meta-regression was used to explore the effect of study design (prospective or retrospective), embryo transfer (single or multiple), culture conditions (single or group) and method of dichotomising soluble human leukocyte antigen-G test results (detection limit or optical density).

Pre-specified subgroup analyses were also performed by creating separate 2x2 tables based on the following characteristics: ICSI treatment; IVF treatment; age (less than 38 or 38 or more); and good quality embryos. Positive and negative predictive values were calculated and plotted against the baseline prevalence of clinical pregnancy.

Publication bias was assessed using a funnel plot, Begg and Mazumdar's rank correlation test, and Egger's regression method.

Eleven studies were included (n=1,813 participants). The mean clinical pregnancy rate was 45% (range 18 to 70%).

Sensitivity, specificity, positive and negative likelihood ratios showed high heterogeneity (I²≥75%) and were not combined in a meta-analysis. Sensitivity ranged from 0.01 to 0.97, specificity ranged from 0.18 to 0.98, positive likelihood ratios ranged from 0.34 to 3.21 and negative likelihood ratios ranged from 0.08 to 1.01.

Diagnostic odds ratios ranged from 0.92 to 24.82 and had moderate heterogeneity (I²=49%); the pooled estimate from the random-effects model was 4.38 (95% CI 2.93 to 6.55). None of the factors explored in the meta-regression showed statistically significant relationships with the diagnostic odds ratio. Six studies with good quality embryos showed better diagnostic performance with a pooled diagnostic odds ratio of 12.67 (95% CI 3.66 to 43.80). There was no evidence of publication bias.

This review indicated that soluble human leukocyte antigen-G in embryo culture supernatants was moderately helpful to predict the achievement of pregnancy in women undergoing fertility treatment. It had a better diagnostic performance if the embryos were of good quality.

This review of the prognostic value of a non-invasive marker for embryo selection in fertility treatment had clearly defined inclusion criteria and an adequate literature search, although there could have been more attempts to locate unpublished research. The authors only included cohort studies, which are the stronger study design, and excluded case-control studies, which are at a greater risk of bias, although these might have added additional evidence to the review. There was no assessment of study quality which is a major drawback of this review.

The authors calculated appropriate measures of diagnostic performance and presented ROC plots and forest plots. The statistical methods were appropriate, but basing the conclusion on the diagnostic odds ratio limits the interpretation of the results, as diagnostic odds ratios are not easily interpreted by clinicians. The authors presented a plot of predictive values, but did not give any of these results; reporting the predictive values for each study may have been helpful.

The authors' conclusions appear to be reliable, but are limited by the lack of consideration of study quality, and their clinical relevance is in doubt given the wide variation in diagnostic accuracy.

Practice: The authors stated that the current evidence suggests that clinical embryologists should continue to select embryos on the basis of currently widely accepted morphological criteria, and that the presence or absence of soluble human leukocyte antigen-G using current methods is not a valid selection criteria.

Research: The authors stated that further research into single-embryo transfer, single culture condition and soluble human leukocyte antigen-G detection thresholds is needed.

Supported by grants from the Willy Gepts Foundation, UZ Brussel.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.