To date, this meta-analysis is the most comprehensive compilation of data to assess the outcomes of r-hFSH plus r-hLH or r-hFSH alone for ovarian stimulation during ART. Our findings indicate that there was no significant difference in the number of oocytes retrieved with r-hFSH plus r-hLH versus r-hFSH alone in the overall population studied. However, we also found that significantly more oocytes were retrieved in women treated with r-hFSH plus r-hLH versus r-hFSH alone in the subgroup of poor responders (+0.75; 95% CI 0.14–1.36). In addition, a significant benefit of r-hFSH plus r-hLH versus r-hFSH alone on clinical pregnancy rate was demonstrated in the poor responders subgroup (RR 1.30; 95% CI 1.01–1.67), which suggests a 30% relative increase in clinical pregnancy rate among poor responders who received r-hLH supplementation. In the overall pooled population analysed here, a significant difference in clinical pregnancy rate was also found in favour of r-hFSH plus r-hLH versus r-hFSH alone (RR 1.09; 95% CI 1.01–1.18). This suggests that there was a smaller (estimate of 9%) relative increase in clinical pregnancy rate among all patients who received r-hLH supplementation compared with that seen in poor responders.
Our findings in the subgroup of poor responders are particularly relevant because many patients undergoing ART are poor responders to ovarian stimulation (although prevalence estimates vary because of differences in the definitions of POR used). Ferraretti and colleagues reported that approximately 33% of patients aged <30–39 years undergoing ovarian stimulation were poor responders (patients with <4 oocytes retrieved) . In addition, evidence in the literature to identify interventions that could improve treatment outcomes in women with POR is limited [28–30] and the traditional clinical approach of increasing the FSH dose to improve follicular response appears to be ineffective [31–33]. Some physicians have attempted to exploit the potential benefit of r-hLH supplementation in women with POR. However, currently, women with POR may undergo multiple unsuccessful ART treatment cycles because of inadequate follicular response, repeated cycle cancellation or a negative pregnancy test.
The use of r-hLH supplementation during ovarian stimulation is a subject of debate in the literature and this lack of clarity has led to the publication of a number of earlier meta-analyses. The first showed a beneficial effect of ‘LH activity’ versus r-hFSH on clinical pregnancy rates (RR 1.18; 95% CI 1.02–1.36) in normogonadotropic women who underwent GnRH agonist downregulation . In contrast to those results and the findings of our meta-analysis, three subsequent meta-analyses reported no clinical benefit of LH supplementation: no statistically significant differences were observed with r-hFSH plus r-hLH versus r-hFSH alone in pregnancy , live birth  or clinical pregnancy  rates. Another meta-analysis conducted in women of advanced reproductive age (≥35 years) found that the clinical pregnancy rate was higher in the r-hLH supplementation group than in the r-hFSH alone group (seven studies, n = 902; OR 1.37; 95% CI 1.03–1.83) . In agreement with the findings of our meta-analysis, an additional meta-analysis found a statistically significant higher pregnancy rate in favour of r-hFSH plus r-hLH compared with r-hFSH alone in a subgroup of patients with poor ovarian response (POR) (three studies; n = 310; OR 1.85; 95% CI 1.10–3.11) . It should be noted that until the development of the ESHRE consensus POR definition in 2011 , heterogeneous definitions of POR were used.
In the current meta-analysis, the chosen co-primary endpoints were number of retrieved oocytes and clinical pregnancy. Of the 45 studies analysed, 41 reported data for number of oocytes, and 39 reported data for clinical pregnancy. Although live birth rate is the outcome measure that patients are most interested in, data for this endpoint were reported sporadically in the RCTs. Oocytes are the direct physiological result of ovarian stimulation by FSH and so reflect the pharmacological effect of FSH, therefore, the number of oocytes retrieved is an appropriate endpoint. Furthermore, the number of oocytes retrieved is not influenced by events occurring after oocyte retrieval, such as fertilization, implantation, or embryo/foetal development, whereas other endpoints of cycle success, such as pregnancy outcomes and live birth rates, may be impacted by these other events.
In addition, number of oocytes retrieved has been widely used as a measure of ovarian response to FSH stimulation and is commonly referenced as a predictor of successful ART outcomes. Increase in pregnancy rates associated with an increased number of oocytes retrieved has been reported by numerous authors through large and smaller retrospective analyses of IVF, ICSI and oocyte donation cycles [36–43]. Sunkara and co-workers (using data from 400,135 ART cycles) found a strong association between live birth rate and number of oocytes retrieved, with live birth rate increasing as the number of oocytes retrieved increased (up to 15 oocytes retrieved) . For patients with a low number (three or fewer) of oocytes retrieved, an increase in live birth rate was observed when as few as one additional oocyte was retrieved . Also, for patients aged ≥40 years with a low number of oocytes retrieved, an increase of just one oocyte had a marked increase in the predicted live birth rate. In addition, an evaluation of nearly 8000 ART cycles found the ongoing pregnancy rate to be highly correlated with the number of oocytes retrieved . These findings suggest that in patients who may be expected to have lower numbers of oocytes retrieved, for example patients with POR, an increment of one additional oocyte might have a significant effect on pregnancy outcomes, such as ongoing pregnancy rate and live birth rate. Supporting this, a recent systematic review reported that the likelihood of pregnancy is reduced in women with POR when fewer oocytes are retrieved (pregnancy rate per started cycle of 0–7% with one oocyte, compared with 11.5–18.6% with four oocytes) .
The meta-analysis reported here found an increase in clinical pregnancy rate with r-hLH supplementation in the overall study population of women undergoing ovarian stimulation, and this finding has been reported in one previous meta-analysis . However, this was not observed in two other meta-analyses [15, 18] that involved fewer studies and smaller numbers of patients than the analysis reported here.
Our analysis attempted to utilize all available data, by imputing missing secondary endpoints, so that the sample size for each endpoint was the best possible for all endpoints. Data imputation for uncommon endpoints, such as live birth rate, may be viewed with caution by some, however, we felt that this practice was justified to allow a greater sample size to be analysed for live birth rate. In contrast to the statistically significant difference in clinical pregnancy rates between r-hFSH plus r-hLH versus r-hFSH alone in poor responders in the ITT population, the difference between the two treatments was not significant in poor responders in the PP population, although the effect size was similar. Data were imputed for the ITT population but not for the PP population.
It is interesting to note that a non-significant negative bias was identified for trials sponsored by the pharmaceutical industry, with a smaller effect of r-hFSH plus r-hLH versus r-hFSH alone in industry-sponsored trials compared with non-sponsored trials. This contrasts with previous findings of a significant positive outcome bias related to industry sponsorship . In addition, the decision to include unpublished data was made to enable evaluation of as much data as possible in the analysis to provide a complete picture of the use of r-hFSH plus r-hLH compared with r-hFSH alone. Furthermore, we conducted analyses of various covariates, including ones for publication status (unpublished data versus congress abstract and unpublished data versus peer-reviewed paper).
The key strength of this meta-analysis is that it comprised the largest number of studies (to the best of our knowledge, all studies) on this subject. In addition, no a priori selection was admitted and bias control was systematically conducted through meta-regression.
A possible limitation of the current meta-analysis is that the 14 studies of women with POR that were included had been conducted prior to the publication of the ESHRE consensus POR definition in 2011 . Accordingly, heterogeneous definitions of POR were used in these studies. When comparing the study authors’ definitions of POR with the ESHRE consensus criteria , each of the studies in the POR analysis were aligned with at least one of the ESHRE criteria and the ESHRE definition of POR was reflected (through alignment with at least two ESHRE criteria) in 10 of these studies.