Risk of gestational diabetes mellitus in women achieving singleton pregnancy spontaneously or after ART: a systematic review and meta-analysis.

A computerized literature search in MEDLINE, Scopus and Cochrane (CENTRAL) was performed independently by two reviewers (J.K.B and P.G.A), covering the period between 1978 and July 2019. This systematic review followed the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) guidelines ( Liberati et al ., 2009 ) (PROSPERO registration number: CRD42019124251). The following PICO (Population, Intervention or exposure, Comparison, Outcome) elements were applied as inclusion criteria for this systematic review: Population: singleton pregnancies; Intervention: ART; Comparator: SC; Outcome: GDM. A search strategy with various synonyms was entered as free-text terms in the electronic databases in an attempt to maximize the sensitivity of the search strategy. The following search string was used: (microinjection[tiab] OR ‘intra-cytoplasmic sperm injection’[tiab] OR ICSI[tiab] OR ‘intracytoplasmic sperm injection’[tiab] OR IVF[tiab] OR ‘in-vitro fertilization’[tiab] OR ‘in vitro fertilization’[tiab] OR ‘in-vitro fertilization’[tiab] OR ‘in vitro fertilization’[tiab]) AND (‘Diabetes, Gestational’[MeSH] OR ‘gestational diabetes’[tiab] OR ‘pregnancy complications’[tiab] OR ‘obstetric complications’[tiab] OR (pregnancy[tiab] AND (diabet*[tiab] OR ‘hyperglycaemia’[tiab] OR ‘hyperglycemia’[tiab] OR ‘high blood glucose’[tiab] OR ‘high plasma glucose’[tiab]))) NOT (Animal[MeSH] NOT Human[MeSH]) NOT (letter[pt] OR comment[pt] OR editorial[pt] OR Review[pt] OR ‘practice guideline’[ptyp] OR ‘case reports’[ptyp]). No language limitations were applied. Institutional Board Review was not obtained as previously published data were used. Criteria for inclusion/exclusion of studies were established prior to the literature search. Studies had to fulfil the following criteria for eligibility: comparative data regarding the risk of GDM in women achieving singleton pregnancy by ART or spontaneously; ovarian stimulation, performed by gonadotropins and GnRH analogues. ART pregnancies included those achieved by IVF or ICSI, after fresh and/or frozen/thawed embryo transfer with autologous gametes. Studies were excluded if pregnancies were achieved using donor gametes, surrogacy, gamete intrafallopian transfer or zygote intrafallopian transfer. Studies performed exclusively in women with PCOS were also excluded due to the known association between PCOS and GDM ( Toulis et al ., 2009 , Yu et al ., 2016 ). Selection of eligible studies was performed independently by two of the reviewers (J.K.B and E.M.K). Any disagreement was resolved by discussion. Data extraction was performed independently by two of the reviewers (J.K.B and E.M.K). When a study provided data separately for the method of fertilization and type of ET, the relevant datasets were used for subgroup analyses. Any disagreement between the two reviewers responsible for data extraction was resolved by discussion. In case of missing data or ambiguities in study design or trial conduction, the study authors were contacted by e-mail to request additional information. The Newcastle-Ottawa Scale (NOS) was used for assessing the quality of each study. Briefly, this system evaluates studies based on three criteria: participant selection; comparability of study groups; and assessment of outcome or exposure. A study can be awarded a maximum of four stars for the selection category, a maximum of two stars for the comparability category and a maximum of three stars for the outcome/exposure category ( Wells et al ., 2014 ). The influence of various factors, such as type of study (matched versus unmatched, prospective versus retrospective) method of fertilization (IVF or ICSI), type of ET (fresh or frozen), inclusion or not of patients with PCOS and study quality (‘good quality’ versus ‘poor quality’ studies), was explored by performing pre-planned subgroup analyses and meta-regression. The dichotomous data results for each of the eligible studies were expressed as risk ratio (RR) with 95% CI. These results were combined for meta-analysis using the random effects model ( DerSimonian and Laird, 1986 ). Study-to-study variation was assessed by using the Chi 2 statistic (the hypothesis tested was that the studies are all drawn from the same population, i.e. from a population with the same effect size). In addition, the use of the I 2 statistic was employed to indicate heterogeneity between studies that could not be attributed to chance, with I 2  ≥ 40% ( Higgins and Green, 2011 ) indicating significant heterogeneity. The presence of publication bias was tested by using the Harbord–Egger’s test ( Harbord et al ., 2006 ). Statistical significance was set at a P level of 0.05. A meta-analysis of weighted average effect sizes was performed using STATA v14.0 (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX, USA: StataCorp LP).

The initial literature search yielded 1356 studies, 73 of which were further evaluated by retrieving their full text and 34 of these were excluded ( Supplementary Table SI ). Eventually, 38 eligible studies were included in the systematic review, 37 of which offered extractable data for the meta-analysis. A flow diagram of this process is present in Fig. 1 . Flow diagram for selection of studies on risk of gestational diabetes mellitus after spontaneous and ART pregnancies . Thirty-eight cross-sectional studies (17 matched and 21 unmatched; 13 prospective and 25 retrospective), published between 1995 and 2019, were eligible for the systematic review, including a total of 1 934 494 women. Characteristics of the studies included in the systematic review are presented in Table I . Of the 38 studies, 24 were graded as being of ‘good quality’ and 14 of ‘poor quality’, according to the NOS ( Supplementary Table SII ). The definition of GDM was reported in 12 out of the 38 studies. After communication with the corresponding authors, further data on the definition of GDM was obtained for 23 studies ( Table I ). In the current systematic review and meta-analysis, studies including only patients with PCOS were excluded, as per protocol. In two of the eligible studies, no patients with PCOS were included ( Suzuki and Miyake, 2007 , Ashrafi et al ., 2014 ), while in 15 studies, they were included in the population analysed ( Reubinoff et al ., 1997 , Koivurova et al ., 2002 , Ochsenkuhn et al ., 2003 , Katalinic et al ., 2004 , Sazonova et al ., 2011 , Farhi et al ., 2013 , Stojnic et al ., 2013 , Machtinger et al ., 2015 , Cai et al ., 2017 , Luke et al ., 2017 , Dayan et al ., 2018 , Frankenthal et al ., 2018 , Harlev et al ., 2018 , Nagata et al ., 2019 , Yang et al ., 2019 ). In the remaining 21 eligible studies, it was unclear whether patients with PCOS were included or not ( Verlaenen et al ., 1995 , Maman et al ., 1998 , Isaksson et al ., 2002 , Zadori et al ., 2003 , Barros Delgadillo et al ., 2006 , De Geyter et al ., 2006 , Schieve et al ., 2007 , Caserta et al ., 2008 , Knoester et al ., 2008 , Sebastiani et al ., 2009 , Chaveeva et al ., 2011 , Tomic and Tomic, 2011 , Silberstein et al ., 2014 , Xu et al ., 2014 , Xu et al ., 2015 , Beyer and Amari, 2016 , Valenzuela-Alcaraz et al ., 2016 , Zhu et al ., 2016 , Qin et al ., 2017 , Lee et al ., 2018 , Szymusik et al ., 2019 ), although this specific information was requested from the corresponding authors ( Table I ). No data regarding the proportion of patients with PCOS were available in 12 out of the 15 studies that included women with PCOS, while this proportion was reported in the remaining three studies ( Farhi et al ., 2013 : 12.5%, Machtinger et al ., 2015 : 2%, Frankenthal et al ., 2018 : 6.5%). Diagnosis of GDM was present in 4776 out of 63 760 women who achieved singleton pregnancy after ART and in 158 526 out of 1 870 734 women who achieved singleton pregnancy spontaneously. In studies evaluating GDM after ART, IVF/ICSI was performed in 22 studies, IVF only in 5 and ICSI only in 3, whereas this information was not present in eight studies. Fresh and frozen ET were performed in 10 studies, fresh ET only in 11 and frozen ET only in 1, whereas this information was not present in 16 studies. Maternal age ( n  = 16), parity ( n  = 11), ethnic origin ( n  = 7), date of delivery ( n  = 6) and BMI ( n  = 3) were the most commonly used variables for matching pregnant women after ART with their counterparts after SC. Additional matching variables included smoking ( n  = 3), social class ( n  = 3), gravidity ( n  = 3), fertility history ( n  = 3), height ( n  = 2), weight ( n  = 2), gestational age ( n  = 2), education ( n  = 1) and obstetric outcome ( n  = 1). Characteristics of the 38 eligible studies included in the systematic review. GDM: gestational diabetes mellitus; OGTT: oral glucose tolerance test; rFSH: recombinant FSH, CC: clomiphene citrate; E2: estradiol, CC; Clomiphene citrate, ACOG: SC: spontaneous conception, ET: embryo transfer, PCOS: polycystic ovary syndrome, PN: pronuclei, WHO: World Health Organization, OHSS: ovarian hyperstimulation syndrome Thirty-seven studies (17 matched, 20 unmatched) provided data for the main comparison. Women with singleton pregnancies achieved by ART showed a higher risk of GDM compared with those women who achieved singleton pregnancy spontaneously (RR 1.53, 95% CI 1.39–1.69, I 2 78.6%, 1 893 599 women) ( Fig. 2 ). No evidence for publication bias was detected using the Harbord–Egger’s test for the primary outcome ( P  = 0.84). Gestational diabetes mellitus after ART versus after spontaneous conception in matched and unmatched studies. RR: risk ratio. Subgroup analysis was performed according to whether the eligible studies were matched ( n  = 17) or unmatched ( n  = 20). This, however, did not change the direction or the magnitude of the effect observed regarding the type of conception and the presence of GDM (matched studies: RR 1.42, 95% CI 1.17–1.72, I 2 61.5%, 21 606 women—unmatched studies: RR 1.58, 95% CI 1.40–1.78, I 2 84.1%, 1 871 993 women) ( Fig. 2 ). Meta-regression analysis confirmed that the type of study (matched versus. unmatched) did not have a significant effect on the association between type of conception and GDM (coefficient: 0.91, 95% CI 0.67–1.22, P  = 0.51). Subgroup analysis was performed according to whether eligible studies were prospective ( n  = 12) or retrospective ( n  = 25). This, however, did not change the direction or the magnitude of the effect observed regarding the type of conception and the presence of GDM (prospective studies: RR 1.52, 95% CI 1.27–1.83, I 2 62.2%, 112 954 women—retrospective studies: RR 1.53, 95% CI 1.36–1.72, I 2 82.5%, 1 780 645 women) ( Supplementary Fig. S1 ). Meta-regression analysis confirmed that the type of study (prospective versus retrospective) did not have a significant effect on the association between type of conception and GDM (coefficient: 0.99, 95% CI 0.74–1.35, P  = 0.99). Subgroup analysis was performed according to whether pregnancies after ART were achieved exclusively either by fresh or by frozen ET ( n  = 17). Compared to women achieving pregnancy spontaneously, a higher risk of GDM was observed in women achieving singleton pregnancy after fresh ET ( n  = 14) (RR 1.38, 95% CI 1.03–1.85, I 2 75.4%, 605 740 women). This association was not present when women achieving pregnancy spontaneously were compared with those achieving singleton pregnancy after frozen ET ( n  = 3) (RR 0.46, 95% CI 0.10–2.19; I 2 73.1%, 12 186 women) ( Fig. 3 ). Meta-regression analysis did not detect a significant effect of type of ET (fresh versus frozen) on the association between type of conception and GDM (coefficient: 0.53, 95% CI 0.19–1.44, P  = 0.19). Gestational diabetes mellitus after ART versus after spontaneous conception according to type of embryo transfer. ET: embryo transfer. Subgroup analysis was performed according to whether pregnancies were achieved exclusively after IVF or ICSI ( n  = 13). Compared to women achieving pregnancy spontaneously, a higher risk of GDM was observed in women achieving singleton pregnancy by IVF ( n  = 7) (RR 1.95, 95% CI 1.56–2.44, I 2 43.1%, 265 253 women). This association was not present when women achieving singleton pregnancy spontaneously were compared with those achieving singleton pregnancy by ICSI ( n  = 6) (RR 1.42, 95% CI 0.94–2.15, I 2 73.5%, 103 402 women) ( Fig. 4 ). Meta-regression analysis did not detect a significant effect of method of fertilization (IVF versus ICSI) on the association between type of conception and GDM (coefficient: 0.80, 95% CI 0.45–1.41, P  = 0.40). Gestational diabetes mellitus after ART versus after spontaneous conception according to method of fertilization . Gestational diabetes mellitus after ART versus after spontaneous conception in studies including patients with PCOS or not, or whether this information was unclear . PCOS: polycystic ovary syndrome Subgroup analysis was performed according to whether studies included patients with PCOS ( n  = 15), excluded specifically patients with PCOS ( n  = 2) or this information was unclear ( n  = 20). This, however, did not change the significance or the direction of the effect observed regarding the type of conception and the presence of GDM (patients with PCOS excluded: RR 4.12, 95% CI 2.63–6.45, I 2 0%, − patients with PCOS included: RR 1.49, 95% CI 1.33–1.66, I 2 75.0%, − unclear information: RR 1.46, 95% CI 1.22–1.75, I 2 77.7%) ( Fig. 5 ). Meta-regression analysis detected a significant effect ( P  < 0.03) of the population analysed on the association between type of conception and the presence of GDM. More specifically, the RR of GDM after ART compared to SC was significantly higher in studies that specifically excluded patients with PCOS compared to those which included patients with PCOS ( P  < 0.01) or to those in which this information was unclear ( P  < 0.01). Subgroup analysis was performed according to whether eligible studies were classified as of ‘good quality’ ( n  = 24) or as of ‘poor quality’ ( n  = 13). This, however, did not change the direction or the magnitude of the effect observed regarding the type of conception and the presence of GDM (‘good quality’ studies: RR 1.53, 95% CI 1.35–1.74, I 2 74.8%, 709 503 women—‘poor quality’ studies: RR 1.50, 95% CI 1.26–1.79, I 2 83.9%, 1 184 096 women) ( Supplementary Fig. S2 ).

In conclusion, the present systematic review and meta-analysis, by analysing 1 893 599 women, showed a higher risk of GDM in women achieving singleton pregnancy by ART compared with those achieving pregnancy spontaneously. This finding highlights the importance of early detection of GDM in women treated by ART, which could lead to timely and effective interventions, prior to ART as well as during early pregnancy.

This systematic review and meta-analysis, including 1 934 494 pregnant women and 163 302 women with GDM, showed an increased risk of GDM in women achieving singleton pregnancy by ART compared with those achieving singleton pregnancy spontaneously. This higher risk was observed after IVF but not after ICSI, and after fresh but not after frozen ET. Nevertheless, meta-regression analyses did not detect any significant effect of method of fertilization or type of ET on the association between GDM and type of conception. To accurately evaluate the association between ART and risk of GDM, studies including exclusively women with PCOS and multiple pregnancies were excluded, since they are considered as strong risk factors for the development of GDM ( Qin et al ., 2015 , Yu et al ., 2016 ). To the best of our knowledge, this is the largest systematic review and meta-analysis focusing on the association between ART and risk of GDM in singleton pregnancies. The present meta-analysis is sufficiently large to provide precise risk estimates. Moreover, it allowed us to perform subgroup analyses, aiming to evaluate the impact of fertilization method and type of ET on the risk of GDM. The definition of GDM was not reported or was unclear in several studies, while a high degree of heterogeneity in its definition was present among those studies that offered such data. Thus, no meaningful subgroup analysis was feasible. Moreover, although the quality of most of the studies was characterised as ‘good’ by NOS, the retrospective design in the majority of the included studies, as well as the fact that most of the studies were unmatched, are potential sources of bias. Nevertheless, the higher risk of GDM in women achieving singleton pregnancy after ART as compared to those achieving pregnancy after SC did not change in subgroup analyses, evaluating whether pooled studies were prospective/retrospective or matched/unmatched. Two previous meta-analyses evaluated the association between ART and risk of GDM in singleton pregnancies ( Jackson et al ., 2004 , Pandey et al ., 2012 ). Both meta-analyses showed a higher risk for GDM, although with a limited number of studies [ Jackson et al ., 2004 : odds ratio (OR) 2.00, 95% CI 1.36–2.99, n  = 4, 2291 women; Pandey et al ., 2012 : RR 1.48, 95% CI 1.33–1.66, n  = 6, 587 790 women]. In the present meta-analysis, the overall sample size increased from 587 790 to 1 934 494 women compared with the meta-analysis by Pandey et al. (2012 ). The underlying mechanisms regarding the increased risk of GDM in women achieving singleton pregnancy by ART compared with those achieving singleton pregnancy spontaneously remain unclear. Moreover, whether the association observed is explained by the presence of infertility per se or the ART procedure performed cannot be evaluated on the basis of the data presented ( Wang et al ., 2017 ). A potential explanation for the increased risk of GDM after ART might be the use of progesterone for luteal phase support in all ART cycles as well as during the first trimester of pregnancy ( Rebarber et al ., 2007 , Ashrafi et al ., 2014 ). Progesterone is known to increase insulin resistance ( Branisteanu and Mathieu, 2003 ), which can lead to GDM. Although a higher risk of GDM was observed after fresh but not after frozen ET, meta-regression analysis failed to detect a potential effect of the type of ET (fresh versus frozen) on the GDM risk. This might be due to the fact that the number of datasets pooled, comparing pregnancies after frozen ET versus pregnancies after SC, was limited ( n  = 3), in contrast to that comparing pregnancies after fresh ET versus pregnancies after SC ( n  = 14). Alternatively, the higher risk of GDM only after fresh ET might be due to the known adverse effects of ovarian stimulation on endometrial receptivity ( Kolibianakis et al ., 2002 , Van Vaerenbergh et al ., 2009 ). Endometrial quality is reported to be associated with the incidence of GDM in singleton pregnancies, since a higher probability of GDM is shown to be present after frozen ET in a hormonal replacement cycle compared with frozen ET in a natural cycle (adjusted OR 0.52, 95% CI 0.39–0.69) ( Saito et al ., 2019 ). The higher risk of GDM, observed only after fresh ET, might be attributed to differences in the quality of placentation between fresh cycles and frozen-thawed cycles ( Kansal Kalra et al ., 2011 ), explained by differences in the hormonal peri-implantation environment in these two clinical scenarios. It has been suggested that supraphysiologic steroid hormone levels during the fresh stimulated cycles may lead to abnormal endometrial angiogenesis and abnormal placentation ( Maheshwari et al ., 2018 ). Altered placental gene regulation has been associated with GDM, probably through epigenetic mechanisms involvement ( Nomura et al ., 2014 , Finer et al ., 2015 , Reichetzeder et al ., 2016 ). Regarding the method of fertilization, although the higher risk of GDM was statistically significant only after IVF but not after ICSI, the direction and magnitude of the effect were similar in both groups, while meta-regression analysis did not detect any significant effect of the fertilization method on the association between GDM and type of conception. Thus, it appears that the method of fertilization does not affect the association between GDM and type of conception. The higher risk of GDM, observed only after IVF but not after ICSI, might be due to the expected higher proportion of women with female pathology associated not only with infertility, but also with GDM, such as advanced maternal age and obesity. On the contrary, in couples undergoing ICSI the expected main cause leading to infertility is male factor and the anticipated presence of the above risk factors in these couples is lower. Due to the fact that a higher risk of GDM has been reported among women with PCOS compared to those without PCOS ( Palomba et al ., 2015 , Azziz et al ., 2016 , Bahri Khomami et al ., 2018 ), the observed association between the type of conception and GDM could be partially attributed to the inclusion of women with PCOS in many of the eligible studies. However, by performing subgroup analysis and meta-regression, the higher risk of GDM after ART compared to SC was still present in studies that specifically excluded PCOS women. In fact, the RR of GDM after ART compared to SC was significantly higher in studies that specifically excluded patients with PCOS compared to those which included them or to those in which this information was unclear. Thus, the effect of the presence of patients with PCOS in many of the eligible studies is probably negligible, which might be attributed to the relatively low proportion of women with PCOS patients in these studies. Women achieving pregnancy after ART should be monitored for GDM, since the risk is increased compared with SC pregnancies. Early detection as well as appropriate support and care is warranted, aiming to avoid serious complications during pregnancy. Whether this risk is attributed to the underlying infertility status of the couples undergoing ART as compared with those who conceived spontaneously needs to be further elucidated.

The number of pregnancies resulting from ART is continuously increasing worldwide. Not unexpectedly, the interest in the potential risks to the mothers and children born after ART has also increased. Currently, a higher risk of obstetric and perinatal complications appears to be present in women achieving pregnancy after ART compared with those achieving pregnancy spontaneously ( Nassar et al ., 2003 , Jackson et al ., 2004 , Pandey et al ., 2012 , Qin et al ., 2015 , Vermey et al ., 2019 ). One of the most common and important complications of pregnancy is gestational diabetes mellitus (GDM). GDM has been associated with a higher risk of pre-eclampsia, caesarean section in the mother as well as macrosomia, shoulder dystocia, hypoglycaemia and jaundice in the newborn ( Ashrafi et al ., 2014 ). In women undergoing ART, major risk factors for GDM, such as advanced maternal age, obesity, multiple pregnancy and polycystic ovary syndrome (PCOS) are often encountered, suggesting a potential association between GDM and ART ( Szymanska et al ., 2011 ). Support for this association was offered by a meta-analysis published in 2012 ( Pandey et al ., 2012 ), including, however, a limited number of studies ( n  = 7). Since the publication of that meta-analysis, several studies evaluating the association between GDM and ART have been published ( Farhi et al ., 2013 , Stojnic et al ., 2013 , Ashrafi et al ., 2014 , Silberstein et al ., 2014 , Xu et al ., 2014 , Xu et al ., 2015 , Beyer and Amari, 2016 , Valenzuela-Alcaraz et al ., 2016 , Zhu et al ., 2016 , Cai et al ., 2017 , Luke et al ., 2017 , Qin et al ., 2017 , Dayan et al ., 2018 , Frankenthal et al ., 2018 , Harlev et al ., 2018 , Lee et al ., 2018 , Nagata et al ., 2019 , Szymusik et al ., 2019 , Yang et al ., 2019 ), with some of them including thousands of patients ( Xu et al ., 2014 , Luke et al ., 2017 ), allowing for more precise estimates to be obtained. Moreover, this is the first systematic review and meta-analysis evaluating the influence of various moderators, such as the method of fertilization and type of embryo transfer (ET), as well as of various confounders, such as study type, in the association between GDM and ART. The purpose of this systematic review and meta-analysis was to synthesize the best available evidence regarding the association between ART and GDM in singleton pregnancies. The specific research question asked was whether the risk of GDM is higher in women achieving singleton pregnancy by ART compared with those achieving singleton pregnancy spontaneously. In addition, the influence of various moderators, such as the method of fertilization (IVF or ICSI) and type of embryo transfer (fresh versus frozen), as well as of various confounders, such as type of study (matched versus unmatched, prospective versus retrospective), was explored.

Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file.