This review aimed to evaluate the potential benefits of dual triggering versus hCG-only triggering in poor responders undergoing controlled ovarian stimulation with a GnRH antagonist protocol and selected according to the Bologna or POSEIDON criteria.

Our analysis indicates that, among patients classified according to the Bologna criteria, dual triggering is associated with a higher number of retrieved and mature oocytes. However, this advantage was not confirmed in patients classified according to the POSEIDON criteria.

It is reasonable to hypothesize that the conflicting results observed in our study could be attributed, at least in part, to the persistent heterogeneity in the definition of poor responders, despite the seemingly strict Bologna and POSEIDON inclusion criteria.

As highlighted by Papathanasiou [27], five different phenotypes of poor ovarian response (POR) can be derived from various combinations of the Bologna criteria: (I) one previous poor response and age ≥ 40 years; (II) one previous poor response and abnormal ovarian reserve markers; (III) age ≥ 40 years and abnormal markers (the so-called "expected poor response"); (IV) a previous poor response in women aged ≥ 40 years with abnormal markers; and (V) two previous POR cycles after maximal stimulation. Thus, grouping such heterogeneous patients under a single label may introduce some degree of bias when comparing different treatment strategies, as this intrinsic clinical heterogeneity can limit the generalizability of pooled results and potentially mask differential responses to triggering strategies among specific subgroups.

Similarly, despite the improvements introduced by the POSEIDON criteria, recent critical analyses have revealed inconsistencies and incomplete reporting in studies that use these criteria.

This led to the development of the POSORT guidelines (POSEIDON Statement Of Reporting Trials) [28], a 20-item checklist aimed at improving reporting and ensuring more homogeneous patient characterization. POSORT recommends specifying which ovarian reserve marker (AFC, AMH, or both) is used and detailing the methods for their assessment to promote valid and reproducible research.

An important confounding factor in our analysis is patient age. Even within groups classified under the same criteria, age distributions varied widely. For instance, even when restricting the analysis to patients meeting the Bologna criteria, age distributions varied notably between studies: Eftekhar et al. reported a median age of 32–33 years, whereas Maged et al. [20] included women with a median age of 38 years. Similarly, within POSEIDON-based population, Chern et al. [18] focused on a cohort with a median age of 40 years (POSEIDON group 4), whereas Ozer et al. [19] examined a younger population, with a median age of 32 years. Interestingly, older patients appeared to benefit more from dual triggering [18, 20, 21] while younger patients [19, 22, 24] showed no significant advantage.These findings suggest that age-related oocyte competence may influence the effectiveness of dual triggering. A subgroup analysis of POSEIDON groups 3 (< 35 years) and 4 (≥ 35 years) would have been informative, but only Tulek et al. stratified patients accordingly. Such an analysis was therefore not possible and further research with extensive, well-designed studies is needed.

In this meta-analysis, the number of mature (MII) oocytes was selected as primary outcome. This parameter, together with the number of retrieved oocytes, represent the most direct and biologically coherent measure of trigger efficacy. These upstream outcomes allow minimizing the confounding influence of downstream factors such as sperm quality, embryo competence, laboratory performance, embryo-selection policies, luteal-phase support, and endometrial receptivity.

This approach aligns with the Vienna consensus [29], which identifies MII and oocyte yeld rate as key performance indicators of ovarian response.

Notably, the only previous meta-analysis on this topic, conducted by Sloth et al. [11], primarily focused on implantation, clinical pregnancy, and live birth rates, without evaluating oocyte yield as a primary efficacy endpoint. By prioritizing oocyte yield, our analysis provides a more proximal and biologically sound assessment of the impact of dual triggering.

From a clinical perspective, it is nevertheless essential to acknowledge the translational gap between increased oocyte yield and improved live birth rates.In particular, an increase in oocyte number does not automatically translate into a higher probability of live birth.

However, in poor responders, even modest improvements in oocyte yield may have clinically meaningful implications. In poor responders, where the reproductive margin is extremely narrow, maximizing the number of mature oocytes retrieved remains pivotal, as the retrieval of even a single additional oocyte may substantially influence the probability of achieving a live birth. Robust evidence from large population-based cohorts supports this concept: for istance, Oudendijk et al. [33] reported a pregnancy probability of only 0–7% with one retrieved oocyte, compared with 11.5–18.6% when four oocytes were obtained. Similarly, Sunkara et al. [34], analyzing more than 400,000 ART cycles, showed that even a single additional retrieved oocyte was associated with a significant increase in live birth rates among low responders.

Consistent with these observations, in our Bologna-defined subgroup, patients receiving dual triggering retrieved approximately one additional oocyte compared with those receiving a single trigger. Specifically, the mean number of MII oocytes was 3.6 ± 1.84 in the dual-trigger group versus 2.7 ± 1.66 in the single-trigger group, while the total number of retrieved oocytes was 4.3 ± 2.15 versus 3.3 ± 1.97, respectively. These findings provide biological and clinical plausibility to the observed benefit of dual triggering in this specific population.

To ensure the completeness of our analysis, clinical pregnancy and miscarriage rates were also assessed, although considered to be influenced by variables unrelated to the trigger intervention, particularly in poor responder populations.

When considering outcomes such as clinical pregnancy rates, defined by the Bologna criteria continued to show a significant benefit with double triggering. In contrast, the overall POSEIDON population did not appear to derive a benefit from dual triggering. It should be noted, however, that one of the included studies — the randomized controlled trial by Keskin et al. [16]—compared two populations with a notable age imbalance, with a mean age of 36.75 years in the double-trigger group versus 33.91 years in the single-trigger group. As embryo transfers were performed without preimplantation genetic testing for aneuploidy (PGT-A), this age difference may represent a substantial source of bias.

Regarding miscarriage rates, the only study supporting a benefit of double triggering was that by Tulek et al. [17], which demonstrated a reduced rate in both POSEIDON group 3 and group 4. The authors attributed this finding to the positive effects of GnRH agonists on both endometrial receptivity and embryo quality, supported by previous evidence of increased HOXA gene expression and improved embryonic development. Although these findings suggest possible biological mechanisms, they remain preliminary and cannot be generalized.

Regarding miscarriage rates, the only study reporting a benefit of dual triggering was that by Tulek et al. [17], which showed a reduced rate in both POSEIDON groups 3 and 4. The authors attributed this finding to potential biological effects of GnRH agonists on endometrial receptivity and embryo quality, supported by evidence of increased HOXA gene expression and improved embryonic development.However, this result appears to be largely driven by a single large retrospective study contributing most of the statistical weight, while the remaining studies did not demonstrate a consistent effect. Moreover, despite relatively standardized luteal-phase support, significant differences in embryo transfer practices were observed between groups, with a higher proportion of day-5 transfers and good-quality embryos in the dual-trigger arm.

These factors may confound the observed association making it difficult to determine whether the observed reduction reflects a true biological effect of dual triggering on embryo competence or is primarily driven by embryo selection and center-specific clinical practices.. Therefore, this finding should be considered exploratory and warrants confirmation in prospective studies with standardized post-retrieval management.

The strength of our article lies in the subgroup analysis conducted according to the Bologna and POSEIDON criteria, which aims to capture potential differences among distinct poor-responder populations. This strategy was intended to minimize heterogeneity and better identify clinical scenarios in which dual triggering could offer a real advantage. Nonetheless, some limitations must be acknowledged. The currently available literature on poor responders is heterogeneous, particularly with regard to study design and control of confounding—an established challenge in this field. We were fully aware of these constraints and addressed them by applying a rigorous methodological approach, including ROB-I/ROB-II assessments, a random-effects model, and evaluation of publication bias. Notably, despite the methodological variability, the direction and magnitude of the effect were consistent across studies, supporting the robustness of our findings.

In addition, only 10 studies were eligible for inclusion, and most were retrospective cohort studies, which are inherently limited by potential confounding factors. Such designs may introduce selection bias due to non-random patient inclusion, as well as information bias arising from incomplete or inconsistently recorded data. Unmeasured confounding related to ovarian reserve, stimulation protocols, and laboratory practices is also possible, as is selective reporting of key outcomes. The lack of blinding is another concern, as it may influence subjective outcomes, notably the assessment of embryo quality. In addition, measurement bias cannot be excluded, as most included studies—except that of Maged et al. [20]—did not specify the AMH assay used, potentially leading to misclassification of ovarian reserve.

Because one included study [25] was rated as having serious confounding bias, we performed a sensitivity analysis excluding it from the meta-analysis of oocyte yield. The pooled estimate remained essentially unchanged, supporting the stability of our findings despite the methodological limitations of some primary studies (Supplemental Fig. 2).

Beyond these intrinsic limitations of retrospective designs, additional methodological concerns may further reduce the strength of our conclusions. Selection bias warrants particular attention: although the Bologna and POSEIDON criteria are widely accepted, they may still cluster women with distinct prognostic profiles—such as differences in age or previous ovarian response. Without refined subgroup stratification, cross-study comparisons may therefore become misleading, potentially diluting any actual benefit of dual triggering in narrowly defined subsets of poor responders.

Although all studies followed the same COH regimen and the standard GnRH antagonist protocol, the type and dosage of gonadotrophins varied across studies, and it remains unclear whether such differences may have influenced oocyte-related outcomes. Nonetheless, available evidence suggests that moderate variations in hCG or GnRH-agonist dosage, as well as differences between recombinant and urinary hCG preparations, are unlikely to substantially affect oocyte maturation or related outcomes [30,31,32].

Additional heterogeneity arises from variations in luteal-phase support and embryo-transfer strategies (single versus double embryo transfer, day-3 versus day-5 transfers), as well as from incomplete reporting of PGT-A procedures. These factors call for caution when interpreting downstream reproductive outcomes such as clinical pregnancy and miscarriage rates.

This study underscores the critical need for precise characterization of poor responders in future trials. Stratification based on patient age, ovarian reserve markers, and standardized diagnostic criteria is essential to generate reliable and clinically meaningful evidence.