Biosimilars versus the originator of follitropin alfa for ovarian stimulation in ART: a systematic review and meta-analysis.

Ovarian stimulation represents an integral part of ART, as it has been shown that obtaining multiple oocytes is associated with a higher number of good-quality embryos, live birth, and cumulative live birth rates (LBRs) ( Law et al. , 2019 ; Venetis et al. , 2019 ; Vermey et al. , 2019 ; Law et al. , 2021 ; Zhang et al. , 2021 ). The first recombinant human FSH molecule (reference product recombinant human follicle-stimulating hormone, rhFSH, follitropin alfa) received marketing approval in Europe in 1995 (GONAL-f ® , Merck KGaA, Darmstadt, Germany) ( European Medicines Agency, 2010 ) and in the USA in 1997 (GONAL-f ® RFF, EMD Serono, Inc., Rockland, MA) ( Food and Drug Administration, 2013 ). Biosimilar preparations are defined as biological medicinal products that contain a version of the active substance of an already authorized original biological medicinal product (reference medicinal product). Similarity to the reference medicinal product in terms of quality characteristics, biological activity, safety, and efficacy based on a comprehensive comparability exercise needs to be established ( European Medicines Agency, 2014 ). Two biosimilars of follitropin alfa have been approved in the EU, Ovaleap ( Strowitzki et al. , 2016 ) (Theramex, Ireland; launched in 2013) and Bemfola ® ( Rettenbacher et al. , 2015 ) (Gedeon Richter PLC, Hungary; launched in 2014; also known as Afolia in the USA ( Fertility Biotech AG, 2017 )), based on Phase III clinical studies and by demonstrating non-inferiority to the reference product Gonal-F in the number of oocytes retrieved. Similarly, Primapur ( Barakhoeva et al. , 2019 ) in Russia, Follitrope ( Hu et al. , 2020 ) and QL1012 ( Hu et al. , 2023a ) in China, Cinnal-f ( Rashidi et al. , 2021 ) in Iran, and Folitime ( Pasqualini et al. , 2021 ) in Argentina have all received approval from the local authorities and were launched in the respective countries, by demonstrating comparability regarding the number of oocytes retrieved. Nevertheless, the most clinically relevant outcome in ART for both clinicians and patients is live birth ( Min et al. , 2004 ; Malizia et al. , 2009 ; Braam et al. , 2018 ), which is supported to be well correlated with clinical and ongoing pregnancy as possible secondary clinical outcomes in randomized clinical trials (RCTs) ( Clarke et al. , 2010 ; Braakhekke et al. , 2014 ). The Phase III clinical trials of the aforementioned biosimilar preparations were powered to detect a difference in the number of oocytes retrieved as the primary endpoint, but also reported outcomes for live birth ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ; Pasqualini et al. , 2021 ; Hu et al. , 2023a ), clinical ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Rashidi et al. , 2021 ; Hu et al. , 2023a ), and ongoing pregnancy ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Barakhoeva et al. , 2019 ; Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Rashidi et al. , 2021 ; Hu et al. , 2023a ). However, these studies were not powered to detect differences in these outcomes. Systematic reviews and meta-analyses are widely regarded as the optimal approach to synthesize outcomes from various trials, particularly when reduced power of individual studies is suspected. A meta-analysis published in 2021 ( Chua et al. , 2021 ) suggested that biosimilars of follitropin alfa are associated with lower LBRs compared to the originator. Since then, more relevant RCTs have been published, and thus an updated critical synthesis of the available evidence is urgently warranted.

The literature search yielded 504 potentially relevant reports ( Fig. 1 ). The titles of these manuscripts were examined, resulting in 124 potentially eligible publications. Subsequently, the abstracts of these studies were assessed, and 36 were identified as potentially suitable to provide data to answer the research question. The full text of these studies was then examined, resulting in the exclusion of 27 published reports (reasons for exclusion are presented in Fig. 1 ), and the final inclusion of 9 reports representing 8 RCTs ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ; Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Rashidi et al. , 2021 ; Hu et al. , 2023a ) published between 2015 and 2023. Seven biosimilar preparations were investigated, Bemfola/Afolia (two RCTs; NCT01121666 , NCT01687712 ), Ovaleap (ISRCTN74772901), Cinnal-f (IRCT201011155181N1), Folitime ( NCT02454556 ), QL1012 ( NCT05149924 ), Primapur ( NCT03088137 ), and Follitrope ( NCT03506243 ). For the RCT NCT01687712 , the data were obtained from ClinicalTrials.gov as no publication presenting the outcomes of this RCT was available in the databases screened. The Bemfola ( NCT01121666 ) and Afolia ( NCT01687712 ) trials had two and three stimulation cycles, respectively. However, only data from the first cycle were included, as per this systematic review and meta-analysis protocol. With regard to the Follitrope study ( NCT03506243 ) ( Hu et al. , 2020 ), a correction was published in 2023 ( Hu et al. , 2023a ) and was taken into consideration during the data extraction and synthesis. PRISMA flow chart. For more information, visit: http://www.prisma-statement.org/ . PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-analysis; RCT, randomized clinical trials. Characteristics of the eight studies are presented in Table 1 and Supplementary Tables S2 and S3 . Randomization method was reported in all the included studies, while allocation concealment was partially reported in the Afolia study ( Fertility Biotech AG, 2017 ). Most of the studies (7/8) were single blinded for the outcome assessor, and only the Cinnal-f study ( Rashidi et al. , 2021 ) was triple blinded (investigators, clinicians, patients). All studies were sponsored. The largest study was the Afolia by Fertility Biotech AG, published only on ClinicalTrials.gov (n = 1101). The risk of bias assessment of the eligible studies is presented in Supplementary Figs S1 and S2 . Overall, three studies ( Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Hu et al. , 2023a ) were deemed to be at high risk of bias, either due to deviations from the intended intervention (only per protocol data were presented regarding live birth, the primary outcome of this meta-analysis) ( Pasqualini et al. , 2021 ; Hu et al. , 2023a ) or due to deviations from the intended intervention and missing outcome data ( Hu et al. , 2020 ) (only per protocol data were presented and >30% of randomized patients did not have an embryo transfer). Four out of eight studies excluded patients who previously had severe OHSS ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ) and the remaining studies excluded patients who had a previous history of any type of OHSS ( Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Rashidi et al. , 2021 ; Hu et al. , 2023a ) ( Table 2 ). All RCTs excluded women with a history of poor ovarian response except from the Cinnal-f trial ( Rashidi et al. , 2021 ). Based on the information provided, five out of eight studies ( Rettenbacher et al. , 2015 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ; Hu et al. , 2020 ; Pasqualini et al. , 2021 ) were deemed to have been performed in normal responders, a single study ( Strowitzki et al. , 2016 ) was performed in infertile patients excluding poor responders and the provided data were not sufficient to confidently ascertain the type of population for the remaining two studies ( Rashidi et al. , 2021 ; Hu et al. , 2023a ). None of the studies was explicitly performed on poor or high responders. Additional information were obtained after contacting the authors of the Folitime ( Pasqualini et al. , 2021 ) and the Primapur ( Barakhoeva et al. , 2019 ) study. Any other attempts to contact the authors from the remaining studies were unsuccessful (either no reply was received or corresponding authors’ contact details were no longer valid). A significantly lower LBR was observed using the biosimilars of follitropin alfa ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ; Pasqualini et al. , 2021 ; Hu et al. , 2023a ) compared to the originator in women undergoing ovarian stimulation for ART (RR: 0.83, 95% CI: 0.72–0.96; I 2 = 0%, six studies, n = 2335; moderate certainty of evidence) ( Fig. 2A ). Publication bias did not seem to be present ( P  = 0.21). Forest plots, presenting the results of the meta-analysis on pregnancy outcomes. ( A ): Forest plot comparing live birth rates between patients stimulated with biosimilars of follitropin alfa and those who were stimulated with the originator; ( B ): Forest plot comparing ongoing pregnancy rates between patients stimulated with biosimilars of follitropin alfa and those who were stimulated with the originator; ( C ): Forest plot comparing clinical pregnancy rates between patients stimulated with biosimilars of follitropin alfa and those who were stimulated with the originator. The sensitivity analysis, which excluded two studies ( Pasqualini et al. , 2021 ; Hu et al. , 2023a ) due to high risk of bias, did not alter the effect size (RR: 0.83, 95% CI: 0.71–0.97; I 2 = 0%, four studies, n = 1881; moderate certainty of evidence) ( Supplementary Fig. S3 ). The subgroup analyses performed did not detect a statistically significant difference between subgroups when studies were grouped by types of biosimilars (Afolia/Bemfola/Ovaleap, Primapur, QL1012, Cinnal-f, Folitime, Follitrope) ( P  = 0.518), by whether the biosimilar had received FDA and/or EMA approval (Afolia/Bemfola/Ovaleap) or not (Primapur, QL1012, Cinnal-f, Folitime, Follitrope) ( P  = 0.668), and the type of population (normal responders: Bemfola, Afolia, Primapur, Follitrope and Folitime versus unclassified: Ovaleap, Cinnal-f, QL1012) ( P  = 0.648). A significantly lower ongoing pregnancy rate was observed using the biosimilars of follitropin alfa ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Barakhoeva et al. , 2019 ; Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Rashidi et al. , 2021 ; Hu et al. , 2023a ) compared to the originator in women undergoing ovarian stimulation for ART (RR: 0.81, 95% CI: 0.70–0.94; I 2 = 0%, seven studies, n = 1886; low certainty of evidence) ( Fig. 2B ). Publication bias did not seem to be present ( P  = 0.43). The sensitivity analysis, which excluded three studies ( Hu et al. , 2020 ; Pasqualini et al. , 2021;  Hu et al. , 2023a ) due to high risk of bias, did not alter the effect size (RR: 0.80, 95% CI: 0.67–0.97; I 2 = 0%, four studies, n = 981; moderate certainty of evidence) ( Supplementary Fig. S4 ). A significantly lower clinical pregnancy rate was observed using the biosimilars of follitropin alfa ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Rashidi et al. , 2021 ; Hu et al. , 2023a ) compared to the originator in women undergoing ovarian stimulation for ART (RR: 0.82, 95% CI: 0.73–0.92; I 2 = 0%, seven studies, n = 2876; low certainty of evidence) ( Fig. 2C ). Publication bias did not seem to be present ( P  = 0.46). The sensitivity analysis, which excluded three studies ( Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Hu et al. , 2023a ) due to high risk of bias, did not alter the effect size (RR: 0.81, 95% CI: 0.70–0.93; I 2 = 0%, four studies, n = 1971; moderate certainty of evidence) ( Supplementary Fig. S5 ). In view of very limited data for moderate and severe OHSS ( Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Pasqualini et al. , 2021 ), OHSS rate was calculated encompassing all the cases reported, regardless of their severity. No difference was observed in OHSS rates (RR: 1.17, 95% CI: 0.90–1.52; I 2 = 0%, eight studies, n = 2986) ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ; Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Rashidi et al. , 2021 ; Hu et al. , 2023a ) ( Fig. 3A ) between the two groups. Publication bias did not seem to be present ( P  = 0.74). Forest plots, presenting the results of the meta-analysis on secondary, non-pregnancy outcomes. ( A ): Forest plot comparing ovarian hyperstimulation syndrome rates between patients stimulated with biosimilars of follitropin alfa and those who were stimulated with the originator; ( B ): Forest plot comparing the duration of gonadotrophin administration (in days) between patients stimulated with biosimilars of follitropin alfa and those who were stimulated with the originator; ( C ): Forest plot comparing the total dose of FSH between patients stimulated with biosimilars of follitropin alfa and those who were stimulated with the originator; ( D ): Forest plot comparing the number of oocytes retrieved between patients stimulated with biosimilars of follitropin alfa and those who were stimulated with the originator. The sensitivity analysis, which excluded three studies ( Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Hu et al. , 2023a ) due to high risk of bias, did not alter the effect size (RR: 1.14, 95% CI: 0.72–1.83; I 2 = 0%, five studies, n = 2081) ( Supplementary Fig. S6 ). A shorter duration of gonadotrophin administration was observed using the biosimilars of follitropin alfa (Ovaleap, Bemfola, Cinnal-f, Primapur, Afolia) ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ; Rashidi et al. , 2021 ) compared to the originator in women undergoing ovarian stimulation for ART (weighted-mean-difference [WMD]: –0.19 days, 95% CI: –0.34 to –0.05; I 2 = 0%, five studies, n = 2081) ( Fig. 3B ). Publication bias did not seem to be present ( P  = 0.99). No significant difference was observed in the total dose of FSH when using the biosimilars of follitropin alfa ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ; Rashidi et al. , 2021 ) as compared to the originator in women undergoing ovarian stimulation for ART (WMD: –34.69 IUs, 95% CI: –74.54 to 5.16; I 2 = 15.53%, five studies, n = 2081) ( Fig. 3C ). Publication bias did not seem to be present ( P  = 0.16). No difference was observed in the number of COCs retrieved using the biosimilars of follitropin alfa ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ; Barakhoeva et al. , 2019 ; Hu et al. , 2020 ; Pasqualini et al. , 2021 ; Rashidi et al. , 2021 ) compared to the originator in women undergoing ovarian stimulation for ART (WMD: 0.27, 95% CI: –0.43 to 0.96; I 2 = 10.71%, six studies, n = 1527) ( Fig. 3D ). Publication bias did not seem to be present ( P  = 0.42). Only one study provided data on fertilization rates per intention to treat ( Rettenbacher et al. , 2015 ) and therefore no meta-analysis was performed. Based on the results of that single study, no significant differences in fertilization rates between the two groups were detected (Biosimilar mean: 66.1%, SD: 24.84 versus Originator mean: 64%, SD: 24.76). No difference in multiple pregnancy rates were observed, based on pooling of data from three studies ( Strowitzki et al. , 2016 ; Barakhoeva et al. , 2019 ; Pasqualini et al. , 2021 ) (RR: 1.31, 95% CI: 0.62–2.78; I 2 = 0%, three studies, n = 509). Publication bias did not seem to be present ( P  = 0.76). No differences in ectopic pregnancy rates were observed based on pooling of data from three studies ( Strowitzki et al. , 2016 ; Fertility Biotech AG, 2017 ; Barakhoeva et al. , 2019 ) (RR: 1.15, 95% CI: 0.40–3.26; I 2 = 0%, three studies, n = 1509). Publication bias did not seem to be present ( P  = 0.82). No differences in miscarriage rates were observed based on pooling of data from two studies ( Rettenbacher et al. , 2015 ; Strowitzki et al. , 2016 ) (RR: 0.91, 95% CI: 0.45–1.82; I 2 = 0%, two studies, n = 671). Publication bias did not seem to be present ( P  = 0.83). Data per randomized patient on the mean number of embryos and mean number of good and/or top-quality embryos were not provided in the eligible studies, and, therefore, a meta-analysis on these outcomes was not feasible. A meta-analysis on cumulative live birth, ongoing pregnancy, and clinical pregnancy rates per oocyte collection (encompassing all embryo transfers originating from a single attempted/planned oocyte retrieval, required to reach a live birth, ongoing pregnancy, or clinical pregnancy, respectively) was also not feasible due to the lack of relevant data from the eligible studies.

The Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) was used during the conduct of this study ( http://www.prisma-statement.org ). A protocol for the systematic review was registered in The International Prospective Register of Systematic reviews (PROSPERO; CRD42024498237) prior to the initiation of the literature search. Two independent reviewers performed a computerized literature search in MEDLINE, Embase, CENTRAL (The Cochrane Library), Scopus, Web of Science, Clinicaltrials.gov, US Food and Drug Administration (FDA), European Medicines Agency (EMA), and the World Health Organization (WHO) registries, published up to January 2024, in order to identify all the available RCTs comparing biosimilars versus the originator of follitropin alfa in women undergoing ovarian stimulation for ART. Relevant keywords/terms, synonyms, related variants, and database-specific indexing terminology were used to describe the biosimilar preparations of rhFSH and the reference product, the intervention, the study type, and the population ( Supplementary Table S1 ). Additionally, the citation lists of relevant publications and systematic reviews were hand-searched. The studies retrieved by the literature search were screened independently by two reviewers (K.I.K. and C.A.V.) based on titles, abstracts, and then by full text. Any disagreement was resolved by discussion. The inclusion and exclusion criteria were set prior to the literature search as follows: the studies had to be RCTs comparing biosimilars of follitropin alfa with the reference product (Gonal-f) in couples with infertility of any type and duration, undergoing controlled ovarian stimulation for IVF/ICSI in fresh and/or frozen–thawed cycles. Studies involving asymmetric co-interventions (i.e. trials in which the protocols for both treatment arms were not the same, except for the intervention assessed) were excluded. Furthermore, pseudo RCTs were also not considered eligible for the present systematic review and meta-analysis. Demographic data were extracted from each of the eligible studies, including the year and the country of publication, as well as the size of the study population. Information regarding the study design was also reported, such as the methodological steps for the randomization and the allocation concealment, the blinding, whether power analysis was performed, whether there was financial support for the trial, and whether there was a protocol registration ( Table 1 ). With regard to the actual trial and protocols details, information was extracted in a predefined standardized data extraction form assessing population characteristics and details of ART treatment protocol used for the randomized controlled trials included in the meta-analysis ( Table 2 ). The data collection was performed independently by two authors, and any disagreement was resolved by a third party (A.S.). In accordance with the Cochrane Handbook for Systematic Reviews of Interventions ( Higgins et al. , 2024 ), in cases where the data relevant to the analysis were not available in the published report, attempts were made to contact the authors of the individual studies, or data from other credible sources (e.g. trial registries) were used to extract the complete dataset. Main methodological characteristics of the randomized controlled trials included in the meta-analysis. NCT01121666 Rettenbacher et al. , 2015 (Bemfola) Only fresh Stratified by age (<35,  ≥ 35) randomization using a minimization algorithm 2:1 ratio ISRCTN74772901 Strowitzki et al. , 2016 (Ovaleap) Only fresh Computer generated randomization stratified by using a block size of 2 1:1 ratio NCT01687712 2017 (Afolia) Only fresh NCT03088137 Barakhoeva et al. , 2019 (Primapur) Only fresh NCT03506243 Hu et al. , 2020 (Follitrope ® ) Only fresh NCT05149924 Hu et al. , 2023a (QL1012) Only fresh IRCT201011155181N1 Rashidi et al. , 2021 (Cinnal-f) Only fresh NCT02454556 Pasqualini et al. , 2021 (Folitime) Only fresh Population characteristics and details of ART treatment protocol used of the randomized controlled trials included in the meta-analysis. NCT01121666 Rettenbacher et al. , 2015 (Bemfola ® ) ISRCTN74772901 Strowitzki et al. , 2016 (Ovaleap) NCT01687712 2017 (Afolia) NCT03088137   Barakhoeva et al. , 2019 (Primapur) NCT03506243 Hu et al. , 2020   (Follitrope ® ) 150 IU/day aged 20–30 150–225 IU/day aged 31–35 225–300 IU/day aged >35 NCT05149924   Hu et al. , 2023a (QL1012) 75–150 IU <30 years 150–225 IU for ages 30–34 225–300 IU ≥ 35 years IRCT201011155181N1 Rashidi et al. , 2021 Cinnal-f Age < 35 years; FSH levels ≤8 mIU/ml on the second day of the cycle; regular menstrual cycles of 25–35 days; and receiving gonadotropin-releasing hormone (GnRH) agonist for pituitary suppression Age 18–38 years, BMI 18–32 kg/m2, regular menstrual cycles of 25–35 days, basal FSH < 10 IU/L (cycle day 2–5), undergoing their first or second cycle in the present series of IVF, total AFC ≥ 8 to ≤18 follicles, with a follicle diameter <10 mm (sum of both ovaries) in the early follicular phase; documented history of infertility due to tubal factor, male factor, or unexplained infertility; presence of both ovaries and a normal uterine cavity as confirmed by transvaginal ultrasound within three months of randomization Quality assessment for RCTs was evaluated according to the revised Cochrane Risk of Bias tool 2.0 ( Sterne et al. , 2019 ) by two independent reviewers (K.I.K. and C.A.V.). The overall quality of evidence was graded according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group guidelines, using the GRADEpro tool ( GRADEPro GDT, 2024 ). The main outcome measures were LBR per randomized patient (defined as the delivery of a single viable infant from 22 weeks of gestation onwards; deliveries of both singletons and multiples were registered as one delivery ( Zegers-Hochschild et al. , 2017 ), using data only from the first cycle), ongoing pregnancy (at least one foetus with positive cardiac activity at 10–12 weeks of gestation), and clinical pregnancy (a viable pregnancy at 5–8 weeks of gestation). Cumulative live birth, ongoing pregnancy, and clinical pregnancy (defined as above) per oocyte collection were also evaluated by including all embryo transfers (either fresh or frozen) originating from a single attempted/planned oocyte retrieval until a live birth, ongoing pregnancy, or clinical pregnancy, respectively, was achieved or all embryos were used, whichever occurred first. Secondary outcomes were miscarriage, multiple pregnancy, ectopic pregnancy, ovarian hyperstimulation syndrome (OHSS) (moderate to severe), total dose of gonadotrophins, duration of gonadotrophin administration, number of cumulus-oocyte complexes (COCs) retrieved, fertilization rates, number of embryos, and number of good and/or top-quality embryos. Miscarriage, ectopic pregnancy, and multiple pregnancy were defined as in The International Glossary on Infertility and Fertility Care, 2017 ( Zegers-Hochschild et al. , 2017 ). The dichotomous data results for each of the eligible studies for meta-analysis were expressed as risk ratio (RR) and the continuous data as mean difference with 95% CIs. Data were analysed according to the intention-to-treat principle. These results were combined for meta-analysis using the Mantel/Haenszel model when using the fixed effects model and the restricted maximum likelihood method with Hartung-Knapp-Sidik-Jonkman correction ( Hartung and Knapp, 2001 ; IntHout et al. , 2014 ) when using the random effects model (in case of high heterogeneity, i.e. I 2 ≥ 50%). All results were combined for meta-analysis with the STATA Software (StataCorp. 2021, Stata Statistical Software: Release 17, College Station, TX, StataCorp LLC). Statistical heterogeneity was estimated with the I 2 statistic ( Deeks et al. , 2019 ). Statistical significance was set at a P level of 0.05. Publication bias was explored using the Egger’s test ( Egger et al. , 1997 ). A sensitivity analysis was performed for live birth, ongoing pregnancy, clinical pregnancy, and OHSS by excluding studies judged to be at overall high risk of bias according to RoB-2 ( Sterne et al. , 2019 ). Subgroup analyses on the primary outcome measure were planned according to different types of biosimilars, biosimilars approved by the FDA and/or EMA, and according to the type of patients (poor, normal, or high responders).

This systematic review and meta-analysis of eight RCTs including ∼3000 patients suggest a relative reduction of 17% in the probability of live birth when using biosimilars compared to using the originator of follitropin alfa (n = 6 studies, 2335 patients). This finding was persistent in the sensitivity analysis performed, which excluded studies at high risk of bias. Moreover, the meta-analysis demonstrated a similarly lower probability with biosimilar preparations of follitropin alfa compared to the originator in ancestor outcomes, such as ongoing and clinical pregnancy, suggesting that there is no differential effect beyond the clinical pregnancy stage. The most recent systematic review and meta-analysis published on this subject ( Chua et al. , 2021 ) has reviewed and pooled data from five RCTs published up to October 2020, all of which have been Phase III clinical trials evaluating biosimilars of follitropin alfa. Regarding live birth, their analysis eventually pooled data from four RCTs and detected a significantly decreased rate in the biosimilars group when compared to the originator (RR: 0.83, 95% CI: 0.71–0.97; four RCTs, n = 1881, I 2 = 0%). Moreover, the study suggested, ongoing pregnancy, and clinical pregnancy rates were significantly lower with biosimilars of follitropin alfa compared with the originator. At the same time, that meta-analysis associated the biosimilars with a shorter duration of gonadotrophin administration, leading to a higher number of COCs retrieved. The outcomes of the current systematic review are consistent with those reported in the aforementioned publication, and the same effect estimate has been maintained with increased precision (narrower CI). Two more RCTs ( Pasqualini et al. , 2021 ; Hu et al. , 2023a ) have been included in the analysis for the primary outcome (RR: 0.83, 95% CI: 0.72–0.96; I 2 = 0%, 6 studies, n = 2335) increasing the total sample size by ∼20%. Similarly, the statistical syntheses of seven studies for clinical and ongoing pregnancy were both in favour of the originator and were not materially altered in the sensitivity analyses performed. In contrast with the previously published systematic review and meta-analysis, the current one did not detect a difference in the number of oocytes retrieved between the biosimilars and the originator of follitropin alfa. Nevertheless, a recent corrigendum ( Venetis et al. , 2023 ) supported that the difference in the number of oocytes retrieved between the originator and the biosimilars of follitropin alfa was the result of incorrectly presented data in the original manuscript ( Hu et al. , 2020 ), and this was confirmed by an issued correction by the authors of the original article ( Hu et al. , 2023b ). Incorporating the correct figures in the original meta-analysis showed that a significant difference in the number of oocytes retrieved was never actually present ( Venetis et al. , 2023 ). An important consideration when assessing the results of this meta-analysis is the considerable homogeneity between the eligible studies. Even though these studies evaluated seven different biosimilars, the population and the clinical protocols used in these trials were similar. This was also confirmed by the fact that the statistical heterogeneity, as measured by the I 2 , was negligible. Importantly though, the effect sizes on live birth, ongoing pregnancy, and clinical pregnancy observed in the six largest individual RCTs are very similar to the pooled effect size. The two RCTs ( Barakhoeva et al. , 2019 ; Pasqualini et al. , 2021 ) where the effect size is not suggestive of lower pregnancy rates in the biosimilars group are the two smallest ones with significant uncertainty as expressed by wide 95% CIs. Furthermore, a recently published study using real-world evidence (RWE) from the French National Health System comparing the outcomes of different types of gonadotrophins also suggested that biosimilars of follitropin alfa were associated with 19% lower odds of live birth (adjusted odds ratio: 0.81, 95% CI: 0.76–0.86) compared to the originator ( Grynberg et al. , 2023 ). This large (>240,000 cycles analysed) study, which statistically controlled for a number of potential confounders, including female age, number of previous oocyte pick-ups, and parity, also suggested lower odds of cumulative live birth with biosimilars of follitropin alfa compared to the originator (adjusted hazard ratio: 0.86, 95% CI: 0.82–0.89). This agreement between two different sources of evidence (meta-analysis of RCTs and RWE) is important, as it essentially triangulates the evidence and adds confidence to the findings ( Heale and Forbes, 2013 ). The explanation of the differences in pregnancy rates despite producing a similar yield of oocytes remains elusive. The most likely explanations are a differential effect on oocyte quality or endometrial receptivity. A discrepancy in the quality of the oocyte could be the result of variations between these products in terms of glycosylation and the resulting isoforms ( Wang et al. , 2016 ; Lunenfeld et al. , 2019 ; Dias and Ulloa-Aguirre, 2021 ) or even in terms of batch-to-batch variability. Such differences between the originator of follitropin alfa and some of its biosimilars (i.e. Bemfola and Ovaleap) have already been demonstrated and could provide a pharmacological basis for the observed clinical differences ( Mastrangeli et al. , 2017 ; Riccetti et al. , 2019 ; Manzi et al. , 2022 ). It should be noted that biosimilars are associated with a shorter duration of stimulation, without any detectable differences in the total dose of FSH and the number of oocytes retrieved. This difference in the stimulation pattern might be associated with differences in oocyte quality and therefore with the observed differences in LBRs. These hypotheses, however, could not be explored as data on embryo quality were not available for most studies and/or could not be statistically synthesized. An alternative hypothesis could be a negative effect of biosimilars on the endometrium either directly or via increased steroids (e.g. late follicular progesterone), which could impact negatively on endometrial receptivity ( Bosch et al. , 2010 ; Venetis et al. , 2013 ; Venetis et al. , 2015 ; Venetis et al. , 2016 ). Data on late follicular steroids, and particularly serum progesterone, were not unfortunately available and therefore this hypothesis could also not be explored. Finally, it has been argued by some ( de Mora and Howles, 2023 ) that the observed differences in pregnancy rates might be attributable to product-unrelated factors such as the hCG dose given to trigger final oocyte maturation, the time interval between oocyte collection and the laboratory procedures followed or even the individual embryo transfer operator. However, such a hypothesis is highly problematic as these are rigorously designed RCTs with strict protocols intended to be used for regulatory approval. Importantly, all included RCTs have been designed, executed, and/or sponsored by entities that were interested in obtaining regulatory approval and commercially benefiting from the biosimilars. Therefore, the scenario of consistent performance bias in favour of the originator product appears to be highly unlikely ( Venetis and Mol, 2023 ). On the other hand, there are some limitations that need to be considered. First, the meta-analysis includes RCTs from seven different biosimilars, which might not have the same pharmacodynamic profile. Nevertheless, the clinical data from the individual studies appear to be rather homogeneous and this was also tested through formal subgroup analyses, which failed to detect any significant differences. Another limitation of this meta-analysis is that only data regarding the fresh embryo transfer cycle were available, and, therefore, it is currently unknown whether there are differences in terms of cumulative live birth, ongoing pregnancy, and clinical pregnancy rates per aspiration. Robust data on oocyte or even better embryo quality would have been helpful in the interpretation of the findings of this meta-analysis, but unfortunately only three studies ( Rettenbacher et al. , 2015 ; Pasqualini et al. , 2021 ; Hu et al. , 2023a ) compared embryo quality (using different outcome measures) and none supported a difference between the biosimilars and the originator of follitropin alfa. Considering the heterogeneous way these comparisons were performed and that in all cases the analysis was not been performed ‘per intention to treat’ but ‘per protocol’, no reliable conclusions could be drawn ( Supplementary Table S3 ). Finally, the population examined in the eligible RCTs does not explicitly include poor or high responders, and therefore, caution should be exercised before extrapolating these results to these populations. Patients and physicians should be informed that biosimilars of follitropin alfa, although more affordable in many countries, are likely to be associated with reduced pregnancy rates compared to the originator. This could affect the cost-effectiveness of these products, which could lead to the originator being more cost-effective despite its higher overall cost ( Schwarze et al. , 2022 ). Moreover, as biosimilars are key products in the pharmaceutical industry, it is imperative that more research is targeted in confirming or rebutting the findings of this meta-analysis, and importantly elucidating the physiological mechanisms by which biosimilars might be affecting pregnancy outcomes. This would ideally result in the optimization and the production of biosimilars so that they lead to similar LBRs compared to the originator of follitropin alfa. In conclusion, based on the best available evidence originating from eight RCTs, biosimilars of follitropin alfa are associated with similar number of oocytes but lower live birth, ongoing pregnancy, and clinical pregnancy rates compared to the originator.