{"paper_id":"437a3931-b86e-4d63-b8cb-10483fa8decf","body_text":"Progesterone is needed to maintain early pregnancy, and progesterone supplementation\nin assisted reproduction technology (ART) cycles is a well-accepted procedure ( Shapiro  et al ., 2014 ;  Holmdahl  et al ., 1971 ;  van der Linden  et al ., 2015 ).\nLuteal phase deficiency affects women undergoing ART for many reasons. The most\nwidely accepted theory posits that luteal phase deficiency originates from premature\nnegative feedback on LH secretion in the pituitary caused by supra-physiological\nlevels of steroids during controlled ovarian stimulation (COS) sustained after\noocyte aspiration by multiple corpora lutea ( van der\nLinden  et al ., 2015 ;  Fatemi 2009 ).\nThere is evidence that luteal phase support (LPS) with progesterone, human chorionic\ngonadotropin (hCG) or gonadotropin-releasing hormone (GnRH) agonists improves\nreproductive outcomes in women undergoing in vitro fertilization (IVF) ( Shapiro  et al ., 2014 ;  van der Linden  et al ., 2015 ;\n Fatemi  et al ., 2007 ;\n Vaisbuch  et al ., 2012 ;\n Merriam  et al ., 2015 ;\n Martins  et al ., 2016 ).\nSince hCG correlates with higher risk of ovarian hyper stimulation syndrome ( van der Linden  et al ., 2015 ;\n Fatemi  et al ., 2007 ;\n Vaisbuch  et al ., 2012 )\nand evidence of the benefits of GnRH agonists is still of very low quality ( Martins  et al ., 2016 ),\nprogesterone appears to be the best option for LPS.\nProgesterone can be administered orally, intramuscularly, vaginally or rectally; all\nroutes seem to present similar levels of efficacy ( Shapiro  et al ., 2014 ;  van der Linden  et al ., 2015 ;  Vaisbuch  et al ., 2012 ;  Merriam  et al ., 2015 ). First-pass metabolism\nsubstantially reduces the bioavailability of oral progesterone to <10% ( Nahoul  et al ., 1993 ).\nIntramuscular progesterone has been associated with pain caused by daily injections,\ninflammatory response, and local abscess ( van der\nLinden  et al ., 2015 ;  Fatemi  et al ., 2007 ;  Vaisbuch  et al ., 2012 ;  Ghanem & Al-Boghdady, 2012 ). Although fewer adverse events are\nobserved with the vaginal route ( Maher  et\nal ., 2013 ), vaginal progesterone causes local irritation,\ndischarge, and bleeding; it is also affected by coitus, since absorption is\ndecreased after intercourse ( Merriam  et\nal ., 2015 ;  Ghanem &\nAl-Boghdady, 2012 ).\nDydrogesterone is a synthetic progestin with enhanced oral bioavailability, known for\nbeing highly selective for the progesterone receptor ( Kupferminc  et al ., 1990 ;  Domitrz  et al ., 1999 ). It is effective in\ntreating reproductive disorders such as threatened abortion and recurrent pregnancy\nloss, and has also been investigated in the prevention of gestational hypertension\nand preterm birth ( Carp, 2012 ;  2015 ;  Hudic\n et al ., 2016 ;  Mohamad Razi  et al ., 2016 ). Dydrogesterone has also\nbeen described to provide similar reproductive results as vaginal progesterone\n( van der Linden  et al .,\n2015 ;  Barbosa  et al .,\n2016 ). The oral route of administration is thought to be a more\npatient-friendly regimen that might improve compliance to treatment.\nThe objective of this systematic review and meta-analysis was to identify, appraise,\nand summarize the evidence from randomized controlled trials examining the efficacy,\nsafety, and tolerability of oral dydrogesterone compared to vaginal progesterone\ncapsules for LPS in women undergoing ART.\n\nThe protocol of this review was registered at PROSPERO(CRD42017071571).\nTrue randomized controlled trials (RCTs) comparing oral dydrogesterone to vaginal\nprogesterone capsules for LPS in women undergoing ART (fresh or frozen embryo\ntransfer following IVF/ICSI) were included. Quasi and pseudo-randomized trials\nwere not included.\nThe following electronic databases were searched for RCTs: PubMed, Scopus, and\nEmbase. The references of the included studies and related reviews were also\nhand-searched.\nThe following terms were used, adjusting for each database as necessary: (IVF OR\nICSI OR embryo OR blastocyst OR oocyte OR egg OR retrieval OR luteal) AND\n(dydrogesterone OR duphaston OR isopregnenone OR dehydrogesterone). There was no\nlimitation regarding language, publication date or publication status.\nTwo authors (MWPB and CON) independently screened publications for titles and\nabstracts based on the pre-established inclusion criteria and checked for\nduplicates. The same authors examined the full text articles of the studies\nselected for inclusion in the review; a third author (WPM) was involved to solve\ndisagreements as needed. The authors corresponded with original study authors to\nclarify study eligibility when required.\nA data extraction form designed and pilot-tested by the authors was used to\nextract data from the included trials. In the event of studies with multiple\npublications, the main trial report was used as reference and additional details\nwere supplemented from secondary reports. The authors corresponded with trial\nauthors to get clarification when required. Data were extracted independently in\na standardized manner by two authors (MWPB and CON); a third author (WPM) was\ninvolved to solve disagreements as needed.\nThe following data were collected to characterize the included trials: authors;\ncountry; institution; funding sources; conflicts of interest; informed consent;\napproval by ethics committees; study design; period of enrollment; eligibility\ncriteria; number of participants in each group at each stage; age and BMI\n(mean±SD) of participants; COS protocol and trigger; number of embryos\ntransferred per woman; and implantation rate.\nThe primary outcomes for effectiveness were live birth and/or ongoing pregnancy\nrates, while the primary outcome for adverse effect was dissatisfaction. Ongoing\npregnancy was used a surrogate indicator of live birth in trials not reporting\nthe latter. Ongoing pregnancy was defined as evidence of fetal cardiac activity\non ultrasound examination after 10-12 weeks of gestation ( Daya, 2003 ). Ongoing pregnancy was calculated as the number\nof clinical pregnancies minus the number of miscarriages in the trials in which\nit was not described. Secondary outcomes were clinical pregnancy; miscarriage\nper clinical pregnancy (single fetal demise in twin or triplet pregnancies was\nnot counted as miscarriage); and any reported side effects.\nAdditional unreported data were collected from the authors of the studies. Where\ndata could not be obtained, clinical pregnancy (and subsequent miscarriage or\nlive birth) was assumed not to have occurred in women with cycle cancellation.\nNo assumption was made for women lost to follow up for other reasons.\nTwo authors (MWPB and CON) independently assessed the risk of selection bias\n(random sequence generation and allocation concealment); performance bias\n(blinding of participants and personnel); detection bias (blinding of outcome\nassessors); attrition bias (incomplete outcome data); reporting bias (selective\noutcome reporting), and other potential sources of bias (e.g.: difference in the\nnumber of embryos transferred, age of participants, co-interventions, early\nstopping). A third author (WPM) was involved to solve disagreements as needed.\nThe Cochrane Collaboration criteria for judging risk of bias was used in this\nreview ( Higgins & Green, 2011 ): the\ntrials were assigned 'low', 'high' or 'unclear' risk of bias. Blinding was not\nconsidered as a factor likely to affect the risk of performance and detection\nbias on reproductive outcomes, but it might be detrimental to the evaluation of\nparticipant satisfaction with treatment, since the main adverse effects related\nto the route of drug administration.\nDichotomous variables were expressed as risk ratios (RR) and the precision of the\nestimates was evaluated by the 95% confidence interval (CI). The clinical\nrelevance of all comparisons was assessed based on the precision of the\nestimates. A random effects model was used to address the differences in true\neffect size across studies, since doses were different. The random effects model\nalso incorporated the heterogeneity observed among studies and thus produced\nmore conservative confidence intervals ( Higgins\n& Green, 2011 ).\nReview Manager 5.3.5 (Copenhagen: The Nordic Cochrane Centre, The Cochrane\nCollaboration, 2014) was used to combine the results comprised in the\nmeta-analysis. The I 2  index was used to assess heterogeneity.\nIncreases in the risk of positive (e.g.: live birth) or negative (e.g.:\nmiscarriage) outcomes in the meta-analysis were plotted to the right of the\ncenterline, while decreases in the risk such outcomes were plotted to the left\nof the centerline. Since one multi-arm study was included, we were careful not\nto double count controls.\nIn view of the difficulty detecting and correcting for publication bias and other\nreporting biases, the authors aimed to minimize the potential impact by\nperforming a comprehensive search for eligible studies and by preventing the\nduplication of data. Additionally, a funnel plot was used to assess the presence\nof small-study effects suggestive of publication bias.\nSensitivity analysis was performed for primary outcomes to verify whether the\nconclusions would have been different if eligibility was restricted to studies\nat low risk of bias.\nA table was generated to summarize the review findings. The quality of the\nevidence for the main outcomes was evaluated following the Grading of\nRecommendations Assessment, Development and Evaluation (GRADE) Working Group\nrecommendation ( Guyatt  et al .,\n2011 ): the limitations of included studies, inconsistency of effect,\nimprecision, indirectness, and risk of publication bias were considered.\nThe quality of the evidence was graded in the following levels ( Balshem  et al ., 2011 ): High\nquality = We are very confident that the true effect lies close to the effect\nobserved in this review; Moderate quality = We are moderately confident in the\neffect estimate: the true effect is likely to be close to the effect observed in\nthis review, but it might be substantially different; Low quality = Our\nconfidence in the effect estimate is limited: the true effect may be\nsubstantially different from the effect observed in this review; Very low\nquality = We have very little confidence in the effect estimate: the true effect\nis likely to be substantially different from the effect observed in this\nreview.\n\nAn electronic search run in June 7, 2017 retrieved 376 records (PubMed = 77;\nScopus = 216; Embase = 83). Additional papers hand-searched from the references\nof the included studies or related reviews were not included. Three hundred and\nfour papers were excluded after their titles and abstracts were read: 128 were\nduplicates and 238 clearly did not meet the eligibility criteria. Ten studies\nwere further examined for eligibility: one study was excluded because it\ncompared dydrogesterone with vaginal progesterone capsules for luteal support in\nIUI cycles ( Khosravi  et al .,\n2015 ). Nine studies were included in our quantitative analysis ( Chakravarty  et al ., 2005a ;\n Ganesh  et al ., 2011 ;\n Patki & Pawar, 2007 ;  Rashidi  et al ., 2016 ;\n Saharkhiz  et al .,\n2016 ;  Salehpour  et\nal ., 2013 ;  Tournaye\n et al ., 2017 ;  Zarei  et al ., 2017 ;  Zargar  et al ., 2016 ); four of the nine studies had\ngroups given medication other than oral dydrogesterone and vaginal progesterone\ncapsules: vaginal progesterone gel ( Ganesh\n et al ., 2011 ); intramuscular progesterone ( Rashidi  et al ., 2016 ;\n Zargar  et al .,\n2016 ); dydrogesterone combined with either GnRH agonist or hCG ( Zarei  et al ., 2017 ). The\nindividuals in these groups were not included in the quantitative analysis.\n Figure 1  shows the study flow\ndiagram.\nFigure 1 Flowchart of study selection.\nFlowchart of study selection.\nThe characteristics of the nine parallel studies included in the quantitative\nanalysis are reported in  Table 1 . One\nstudy was held in two centers ( Saharkhiz\n et al ., 2016 ), one in 38 different sites ( Tournaye  et al ., 2017 ),\nand the remaining seven were carried out in single centers. Five studies were\nconducted in Iran ( Rashidi  et\nal ., 2016 ;  Saharkhiz\n et al ., 2016 ;  Salehpour  et al ., 2013 ;  Zarei  et al ., 2017 ;  Zargar  et al ., 2016 ), three in India\n( Chakravarty  et al .,\n2005a ;  Ganesh  et\nal ., 2011 ;  Patki &\nPawar, 2007 ), and one in multiple centers in Austria, Belgium,\nGermany, Finland, Israel, Russia, and Spain ( Tournaye  et al ., 2017 ). Despite attempts to contact\nthe authors of all studies, additional details were collected from only one\nstudy ( Saharkhiz  et al .,\n2016 ). All studies were published as full articles. Only patients\nusing vaginal progesterone capsules were used in the comparisons.\nCharacteristics of the included studies.\nNR = not reported; ART = assisted reproductive technology; FET =\nfrozen embryo transfer; VP = vaginal progesterone;\n* = study vs. control\nNOTES: One study reported a conflict of interest  ( Tournaye  et al. ,\n2017 ); Two studies reported funding sources ( Tournaye  et al. ,\n2017 ;  Zarei  et\nal. , 2017 ); All studies obtained approval\nfrom ethics committees, and one study did not provide informed\nconsent  ( Chakravarty   et\nal. , 2005a ); All were parallel studies; Two\nstudies reported an mean of three embryos transferred ( Ganesh  et al. ,\n2011 ;  Salehpour\n et al. , 2013 ), and one study\nreported an mean of two embryos transferred  ( Rashidi  et al. , 2016 ).\nParticipants: 4,061 women submitted to ART in nine studies were included; 1,905\nwere allocated to groups prescribed dydrogesterone for luteal phase\nsupplementation, and 2,156 were allocated to groups on vaginal progesterone\ncapsules. The eligibility criteria, and therefore the characteristics of the\nincluded participants, were different across studies and are reported on  Table 1 .\nInterventions: The nine studies assessed the use of daily oral dydrogesterone in\ndoses ranging from 20mg to 40mg versus vaginal progesterone capsules in doses\nranging from 600 mg/day to 800 mg/day.\nOutcomes: Two of nine studies reported live births ( Rashidi  et al ., 2016 ;  Tournaye  et al ., 2017 );\n3/9 reported ongoing pregnancies ( Chakravarty\n et al ., 2005a ;  Saharkhiz  et al ., 2016 ;  Zarei  et al ., 2017 ); 8/9 reported clinical\npregnancies ( Ganesh  et al .,\n2011 ;  Patki & Pawar, 2007 ;\n Rashidi  et al .,\n2016 ;  Saharkhiz  et al .,\n2016 ;  Salehpour  et\nal ., 2013 ;  Tournaye\n et al ., 2017 ;  Zarei  et al ., 2017 ;  Zargar  et al ., 2016 ); 7/9 reported miscarriages\n( Chakravarty  et al .,\n2005a ;  Ganesh  et\nal ., 2011 ;  Rashidi\n et al ., 2016 ;  Saharkhiz  et al ., 2016 ;  Salehpour  et al ., 2013 ;  Zarei  et al ., 2017 ;  Zargar  et al ., 2016 ); 2/9\nreported female patient dissatisfaction ( Chakravarty  et al ., 2005a ;  Saharkhiz  et al ., 2016 ); and 3/9 reported\nside effects ( Saharkhiz  et al .,\n2016 ;  Salehpour  et\nal ., 2013 ;  Tournaye\n et al ., 2017 ). In one study, the number of\nclinical pregnancies was assumed to be equal to the summation of ongoing\npregnancies and miscarriages ( Chakravarty\n et al ., 2005a ). In three studies, the number of\nongoing pregnancies was assumed to be equal to the number of clinical\npregnancies minus miscarriages ( Ganesh\n et al ., 2011 ;  Salehpour  et al ., 2013 ;  Zargar  et al ., 2016 ).\nSix studies described adequate methods of randomization ( Ganesh  et al ., 2011 ;  Rashidi  et al ., 2016 ;  Saharkhiz  et al ., 2016 ;\n Salehpour  et al .,\n2013 ;  Tournaye  et\nal ., 2017 ;  Zarei  et\nal ., 2017 ;  Zargar\n et al ., 2016 ) and two studies did not report the\nmethod used ( Chakravarty  et al .,\n2005a ;  Patki & Pawar,\n2007 ). Six studies described allocation concealment through sealed\nenvelopes ( Ganesh  et al .,\n2011 ;  Rashidi  et\nal ., 2016 ;  Saharkhiz\n et al ., 2016 ;  Salehpour  et al ., 2013 ;  Tournaye  et al ., 2017 ;  Zargar  et al ., 2016 ). One\nstudy blinded participants and care providers ( Tournaye  et al ., 2017 ). In six studies outcome\nassessors were blinded to allocation ( Ganesh\n et al ., 2011 ;  Rashidi  et al ., 2016 ;  Saharkhiz  et al ., 2016 ;  Salehpour  et al ., 2013 ;\n Tournaye  et al .,\n2017 ;  Zargar  et al .,\n2016 ) and the remaining three studies did not report whether outcome\nassessors were blinded.\nSaharkhiz  et al ., 2016 \nwas judged to be at high risk of attrition bias, since 24/234 (10.3%)\nparticipants were excluded after randomization; loss to follow up was unbalanced\nbetween groups, with 21/117 (17.9%) participants in the dydrogesterone group and\n3/117 (2.6%) in the progesterone group. The other eight studies were judged to\nbe at a low risk of attrition bias. Five studies analyzed all randomized women\n( Chakravarty  et al .,\n2005a ;  Ganesh  et\nal ., 2011 ;  Patki &\nPawar, 2007 ;  Salehpour  et\nal ., 2013 ;  Zargar\n et al ., 2016 ).  Rashidi  et al . (2016 ) excluded one of 120\nparticipants from the analysis because she failed to come to embryo transfer due\nto a car accident; this study was deemed to present low risk of attrition bias\nsince the withdrawal rate was low.  Tournaye\n et al . (2017)  excluded 57/1031 (5.5%)\nparticipants after randomization; loss to follow-up was balanced between groups\n- 23/520 (4.4%) in the dydrogesterone group and 34/511 (6.6%) in the\nprogesterone group - and the study was considered to present low risk of\nattrition bias. For the same reasons the study by  Zarei  et al ., (2017)  was assigned low risk\nof attrition bias: 22/222 (10%) participants were excluded after randomization,\nbut loss to follow-up was balanced between groups, with 10/110 in the\ndydrogesterone group and 12/112 in the progesterone group.\nThe study by  Zargar  et al .\n(2016 ) was judged to present high risk of selective reporting bias,\nas three outcomes described in the registered protocol were not reported (live\nbirth, preterm delivery, and perineal irritation caused by vaginal\nprogesterone). Four studies reported all outcomes described in the registered\nprotocol ( Rashidi  et al .,\n2016 ;  Saharkhiz  et\nal ., 2016 ;  Salehpour\n et al ., 2013 ;  Tournaye  et al ., 2017 ) and the remaining four were\nnot assessed as presenting selective reporting bias.\nOne study was deemed at high risk of bias for containing a larger proportion of\nwomen aged 40+ years in the dydrogesterone group ( Chakravarty  et al ., 2005a ). There was no suspicion\nof other sources of bias in the other eight studies.\nForest plots were used to show the results of each individual study and their\nrespective possible biases ( Figures 2 - 4 )\nFigure 2 Forest plot for live birth/ongoing pregnancy. Risk of bias legend: A\n= Selection bias (random sequence generation); B = Selection bias\n(allocation concealment); C = Performance bias; D = Detection bias;\nE = Attrition bias; F = Reporting bias; G = Other biases.\nForest plot for live birth/ongoing pregnancy. Risk of bias legend: A\n= Selection bias (random sequence generation); B = Selection bias\n(allocation concealment); C = Performance bias; D = Detection bias;\nE = Attrition bias; F = Reporting bias; G = Other biases.\nFigure 4 Forest plot for miscarriage. Risk of bias legend: A = Selection bias\n(random sequence generation); B = Selection bias (allocation\nconcealment); C = Performance bias; D = Detection bias; E =\nAttrition bias; F = Reporting bias; G = Other biases.\nForest plot for miscarriage. Risk of bias legend: A = Selection bias\n(random sequence generation); B = Selection bias (allocation\nconcealment); C = Performance bias; D = Detection bias; E =\nAttrition bias; F = Reporting bias; G = Other biases.\nLive birth / Ongoing pregnancy ( Table\n2 )\nSummary of findings.\nAll outcomes, except miscarriage, were analyzed per randomized\nwomen.\nCI = confidence interval; RR = relative risk;\na  = The absolute risk in the Vaginal Progesterone group\nwas determined as the mean risk in these groups; the absolute risk\nin the Oral Dydrogesterone group and its 95% CI was determined using\nthe RR and its 95% CI;\n1 . Downgraded one level because of imprecision.\nOverall, there was no evidence of relevant differences between oral\ndydrogesterone and vaginal progesterone on live birth/ongoing pregnancy rates:\nRR 1.08; 95% CI 0.92 to 1.26; I 2 =29%, 8 RCTs, 3,386 women; high\nquality evidence. In other words, considering a live birth/ongoing pregnancy\nrate of 24% in women using vaginal progesterone, this rate would be in the range\nof 22-30% in women using oral dydrogesterone. Sensitivity analysis excluding the\nthree studies at high risk of bias did not change the estimate: RR 1.10; 95% CI\n0.86 to 1.40; I 2 =48%, 5 RCTs, 2,334 women.\nClinical pregnancy ( Figure 3 )\nFigure 3 Forest plot for clinical pregnancy. Risk of bias legend: A = f4\nSelection bias (random sequence generation); B = Selection bias\n(allocation concealment); C = Performance bias; D = Detection bias;\nE = Attrition bias; F = Reporting bias; G = Other biases.\nForest plot for clinical pregnancy. Risk of bias legend: A = f4\nSelection bias (random sequence generation); B = Selection bias\n(allocation concealment); C = Performance bias; D = Detection bias;\nE = Attrition bias; F = Reporting bias; G = Other biases.\nOverall, there was no evidence indicating that clinical pregnancy was affected by\nthe use of oral dydrogesterone versus vaginal progesterone capsules: RR 1.10,\n95% CI 0.95 to 1.27; I 2 =43%; 9 RCTs; 4,061 women; high quality\nevidence. If 28% of the women using vaginal progesterone became pregnant, 27-36%\nof the women using oral dydrogesterone might also be clinically pregnant.\nSensitivity analysis excluding the three studies at high risk of bias did not\nchange the estimate: RR 1.08; 95% CI 0.86 to 1.36; I 2  = 51%, 5 RCTs,\n2,334 women.\nMiscarriage ( Figure 4 )\nOverall, there was no evidence indicating that miscarriage was affected by the\nuse of oral dydrogesterone versus vaginal progesterone: RR=0.92,\n95%CI=0.68-1.26, I 2 =6%, 8 RCTs, 988 clinical pregnancies; moderate\nquality evidence.\nTwo studies reported patient dissatisfaction with treatment ( Chakravarty  et al ., 2005a ;\n Saharkhiz  et al .,\n2016 ) (26,30). Since the two studies were significantly heterogeneous\n(I 2  = 91%), their results were not pooled together.  Saharkhiz  et al . (2016) \nreported no difference in dissatisfaction between groups: 8% in women using\ndydrogesterone  vs.  7% in women using vaginal progesterone\ncapsules; RR 1.19, 95% CI 0.46 to 3.04; 210 women. This study was deemed to be\nat high risk of bias.  Chakravarty  et\nal . (2005a)  described a great benefit of dydrogesterone\nin reducing patient dissatisfaction: 3% in women using dydrogesterone\n vs.  26% in women using vaginal progesterone capsules; RR\n0.10, 95% CI 0.02 to 0.39; 430 women.\nSubstantial heterogeneity (>50%) was found for all side effects reported by\nthe three studies and therefore the results were not pooled together. While two\nstudies did not describe differences in reported side effects between the two\ngroups ( Saharkhiz  et al .,\n2016 ;  Tournaye  et\nal ., 2017 ), one study showed that dydrogesterone was\nassociated with more cases of vaginal bleeding (RR 2.38; 95% CI 1.18 to 4.78),\nnausea (RR 21.00; 95% CI 1.27 to 346.66), and abdominal pain (RR 13.00; 95% CI\n0.76 to 223.33) when compared to vaginal progesterone capsules.\nAlthough suboptimal, since fewer than 10 studies were included, the funnel-plot\nanalysis for the only outcome reported in the nine studies - clinical pregnancy\n- was not suggestive of publication bias ( Figure\n5 ).\nFigure 5 Funnel plot analysis for clinical pregnancy.\nFunnel plot analysis for clinical pregnancy.\nSensitivity analysis was reported along with the synthesis of the results.\n\nNine studies were included in the comparison between oral dydrogesterone and\nvaginal progesterone capsules. Oral dydrogesterone was generally as effective as\nvaginal progesterone capsules for luteal phase support in women undergoing\nembryo transfers after IVF/ICSI. The assessment of patient dissatisfaction with\ntreatment revealed an important inconsistency between the two studies reporting\nthis outcome: one reported a significant difference favoring dydrogesterone\n( Chakravarty  et al .,\n2005a ) while the other found no differences between the regimens\n( Saharkhiz  et al .,\n2016 ). Possible explanations for this discrepancy are the different\ndoses of dydrogesterone and the potential differences in the characteristics of\nthe two patient populations.\nOur findings were in agreement with the latest Cochrane review on the subject,\nwhich suggested a significant effect in favor of synthetic progesterone versus\nnatural progesterone ( van der Linden  et\nal ., 2015 ). Four studies were included in the\ncomparison, three of which also included in our review ( Chakravarty  et al ., 2005 a;  Ganesh  et al ., 2011 ;  Patki & Pawar, 2007 ). The other study\nwas not included in our review because it compared oral chlormadinone acetate to\nintramuscular progesterone ( Iwase  et\nal ., 2008 ). Another recent review showed that\ndydrogesterone provides similar reproductive results when compared to vaginal\nprogesterone ( Barbosa  et al .,\n2016 ). Seven studies were included in this review ( Chakravarty  et al .,\n2005a ; b;  et\nal . ,  2006 ;  Ganesh  et al ., 2011 ;  Patki & Pawar, 2007 ;  Saharkhiz  et al. , 2016 ;\n Salehpour  et al .,\n2013 ), five of which were also included in our review ( Chakravarty  et al ., 2005a ;\n Ganesh  et al ., 2011 ;\n Patki & Pawar, 2007 ;  Saharkhiz  et al ., 2016 ;\n Salehpour  et al .,\n2013 ). Two of the studies were not included in our review because\nthey were published as abstracts, thus yielding a high risk of bias to the\ncomparison ( Chakravarty  et al .,\n2005 b ;  2006 ). Four other\nstudies were included in our review ( Rashidi\n et al ., 2016 ;  Tournaye  et al ., 2017 ;  Zarei  et al ., 2017 ;  Zargar  et al ., 2016 ). One of these studies\nwas sponsored by a pharmaceutical company ( Tournaye  et al ., 2017 ), but its results were\nsimilar to the one described in other trials. The only difference between the\nstudy by  Tournaye  et al .\n(2017)  and the others included was the double-blinding procedure,\nwhich in fact minimizes the risk of bias ( Lexchin  et al ., 2003 ). With the addition of more\nstudies in our review, and by excluding the abstracts, the authors believe that\nthis review provides a robust body of evidence for the comparison between\ndydrogesterone and vaginal progesterone capsules for LPS in women undergoing\nembryo transfers.\nIn terms of dissatisfaction with treatment, our review included the same studies\nas the cited review ( Barbosa  et\nal ., 2016 ). The discrepancy between the two studies in\nregards to this outcomes makes it difficult to draw firm conclusions. Different\nside effects were reported in three studies ( Saharkhiz  et al ., 2016 ;  Salehpour  et al ., 2013 ;  Tournaye  et al ., 2017 ),\nand two of them did not report differences between the two groups ( Saharkhiz  et al ., 2016 ;\n Tournaye  et al .,\n2017 ). Additionally, a systematic review on the use of dydrogesterone\nfor recurrent miscarriage found 13 studies reporting apparently minimal adverse\neffects ( Carp, 2015 ).\nThe evidence available suffers from the limitations inherent to the included\nstudies: five of the nine studies had high risk of bias in at least one domain;\nand the use of different doses in case and control groups along with different\ndurations of LPS may have introduced some heterogeneity in the analysis. This\nissue was addressed with a random-effects model and by the incorporation of\nobserved heterogeneity in the interpretation of the findings and in the\nassessment of the quality of the evidence.\nThe quality of the evidence was considered to be high for live birth/ongoing\npregnancy and clinical pregnancy. It was downgraded one level for miscarriage\nbecause of imprecision: there was a relatively low number of events and a broad\nconfidence interval.\n\nOral dydrogesterone is as effective as vaginal progesterone capsules for luteal-phase\nsupplementation in ART cycles. Oral dydrogesterone might be a good option in\nclinical practice, since oral administration is more patient-friendly than the\nvaginal route. The choice for either should be based mainly on availability, cost,\nand side effects.","source_license":"CC-BY-4.0","license_restricted":false}