Prelabor screening at term using the cerebroplacental ratio and placental growth factor: a pragmatic randomized open-label phase 2 trial

Exploring the Limits of Artificial Intelligence for Referencing Scientific Articles

OBJECTIVE

 To evaluate the reliability of three artificial intelligence (AI) chatbots (ChatGPT, Google Bard, and Chatsonic) in generating accurate references from existing obstetric literature.

STUDY DESIGN

 Between mid-March and late April 2023, ChatGPT, Google Bard, and Chatsonic were prompted to provide references for specific obstetrical randomized controlled trials (RCTs) published in 2020. RCTs were considered for inclusion if they were mentioned in a previous article that primarily evaluated RCTs published by the top medical and obstetrics and gynecology journals with the highest impact factors in 2020 as well as RCTs published in a new journal focused on publishing obstetric RCTs. The selection of the three AI models was based on their popularity, performance in natural language processing, and public availability. Data collection involved prompting the AI chatbots to provide references according to a standardized protocol. The primary evaluation metric was the accuracy of each AI model in correctly citing references, including authors, publication title, journal name, and digital object identifier (DOI). Statistical analysis was performed using a permutation test to compare the performance of the AI models.

RESULTS

 Among the 44 RCTs analyzed, Google Bard demonstrated the highest accuracy, correctly citing 13.6% of the requested RCTs, whereas ChatGPT and Chatsonic exhibited lower accuracy rates of 2.4 and 0%, respectively. Google Bard often substantially outperformed Chatsonic and ChatGPT in correctly citing the studied reference components. The majority of references from all AI models studied were noted to provide DOIs for unrelated studies or DOIs that do not exist.

CONCLUSION

 To ensure the reliability of scientific information being disseminated, authors must exercise caution when utilizing AI for scientific writing and literature search. However, despite their limitations, collaborative partnerships between AI systems and researchers have the potential to drive synergistic advancements, leading to improved patient care and outcomes.

KEY POINTS

· AI chatbots often cite scientific articles incorrectly.. · AI chatbots can create false references.. · Responsible AI use in research is vital..

Term planned delivery based on fetal growth assessment with or without the cerebroplacental ratio in low-risk pregnancies (RATIO37): an international, multicentre, open-label, randomised controlled trial

BACKGROUND

The cerebroplacental ratio is associated with perinatal mortality and morbidity, but it is unknown whether routine measurement improves pregnancy outcomes. We aimed to evaluate whether the addition of cerebroplacental ratio measurement to the standard ultrasound growth assessment near term reduces perinatal mortality and severe neonatal morbidity, compared with growth assessment alone.

METHODS

RATIO37 was a randomised, open-label, multicentre, pragmatic trial, conducted in low-risk pregnant women, recruited from nine hospitals over six countries. The eligibility criteria were designed to be broad; participants were required to be 18 years or older, with an ultrasound-dated confirmed singleton pregnancy in the first trimester, an alive fetus with no congenital malformations at the routine second-trimester ultrasound, an absence of adverse medical or obstetric history, and the capacity to give informed consent. Women were randomly assigned in a 1:1 ratio (block size 100) using a web-based system to either the concealed group or revealed group. In the revealed group, the cerebroplacental ratio value was known by clinicians, and if below the fifth centile, a planned delivery after 37 weeks was recommended. In the concealed group, women and clinicians were blinded to the cerebroplacental ratio value. All participants underwent ultrasound at 36 + 0 to 37 + 6 weeks of gestation with growth assessment and Doppler evaluation. In both groups, planned delivery was recommended when the estimated fetal weight was below the tenth centile. The primary outcome was perinatal mortality from 24 weeks' gestation to infant discharge. The study is registered at ClinicalTrials.gov (NCT02907242) and is now closed.

FINDINGS

Between July 29, 2016, and Aug 3, 2021, we enrolled 11 214 women, of whom 9492 (84·6%) completed the trial and were eligible for analysis (4774 in the concealed group and 4718 in the revealed group). Perinatal mortality occurred in 13 (0·3%) of 4774 pregnancies in the concealed group and 13 (0·3%) of 4718 in the revealed group (OR 1·45 [95% CI 0·76-2·76]; p=0·262). Overall, severe neonatal morbidity occurred in 35 (0·73%) newborns in the concealed group and 18 (0·38%) in the revealed group (OR 0·58 [95% CI 0·40-0·83]; p=0·003). Severe neurological morbidity occurred in 13 (0·27%) newborns in the concealed group and nine (0·19%) in the revealed group (OR 0·56 [95% CI 0·25-1·24]; p=0·153). Severe non-neurological morbidity occurred in 23 (0·48%) newborns in the concealed group and nine (0·19%) in the revealed group (0·58 [95% CI 0·39-0·87]; p=0·009). Maternal adverse events were not collected.

INTERPRETATION

Planned delivery at term based on ultrasound fetal growth assessment and cerebroplacental ratio at term was not followed by a reduction of perinatal mortality although significantly reduced severe neonatal morbidity compared with fetal growth assessment alone.

FUNDING

La Caixa foundation, Cerebra Foundation for the Brain Injured Child, Agència per la Gestió d'Ajuts Universitaris i de Recerca, and Instituto de Salud Carlos III.

Number and quality of randomized controlled trials in obstetrics published in the top general medical and obstetrics and gynecology journals

BACKGROUND

There has been an increasing number of randomized controlled trials published in obstetrics and maternal-fetal medicine to reduce biases of treatment effect and to provide insights on the cause-effect of the relationship between treatment and outcomes.

OBJECTIVE

This study aimed to identify obstetrical randomized controlled trials published in top weekly general medical journals and monthly obstetrics and gynecology journals, to assess their quality in reporting and identify factors associated with publication in different journals.

STUDY DESIGN

The 4 weekly medical journals with the highest 2019 impact factor (New England Journal of Medicine, The Lancet, The Journal of the American Medical Association, and British Medical Journal), the top 4 monthly obstetrics and gynecology journals with obstetrics-related research (American Journal of Obstetrics & Gynecology, Ultrasound in Obstetrics & Gynecology, Obstetrics & Gynecology, and the British Journal of Obstetrics and Gynaecology), and the American Journal of Obstetrics & Gynecology Maternal-Fetal Medicine were searched for obstetrical randomized controlled trials in the years 2018 to 2020. The primary outcome was the number of obstetrical randomized controlled trials published in the obstetrics and gynecology journals vs the weekly medical journals and the percentage of trials published, overall and per journal. The secondary outcomes included the proportion of positive vs negative trials overall and per journal and the assessment of the study characteristics of published trials, including quality assessment criteria.

RESULTS

Of the 4024 original research articles published in the 9 journals during the 3-year study period, 1221 (30.3%) were randomized controlled trials, with 137 (11.2%) randomized controlled trials being in obstetrics (46 in 2018, 47 in 2019, and 44 studies in 2020). Furthermore, 33 (24.1%) were published in weekly medical journals, and 104 (75.9%) were published in obstetrics and gynecology journals. The percentage of obstetrical randomized controlled trials published ranged from 1.5% to 9.6% per journal. Overall, 34.3% of obstetrical trials were statistically significant or "positive" for the primary outcome. Notably, 24.8% of the trials were retrospectively registered after the enrollment of the first study patient. Trials published in the 4 weekly medical journals enrolled significantly more patients (1801 vs 180; P<.001), received more often funding from the federal government (78.8% vs 35.6%; P<.001), and were more likely to be multicenter (90.9% vs 42.3%; P<.001), non-United States based (69.7% vs 49.0%; P=.03), and double blinded (45.5% vs 18.3%; P=.003) than trials published in the obstetrics and gynecology journals. There was no difference in study type (noninferiority vs superiority) and trial quality characteristics, including pretrial registration, ethics approval statement, informed consent statement, and adherence to the Consolidated Standards of Reporting Trials guidelines statement between studies published in weekly medical journals and studies published in obstetrics and gynecology journals.

CONCLUSION

Approximately 45 trials in obstetrics are being published every year in the highest impact journals, with one-fourth being in the weekly medical journals and the remainder in the obstetrics and gynecology journals. Only about a third of published obstetrical trials are positive. Trials published in weekly medical journals are larger, more likely to be funded by the government, multicenter, international, and double blinded. Quality metrics are similar between weekly medical journals and obstetrics and gynecology journals.

The fetal cerebroplacental ratio in pregnancies complicated by hypertensive disorders of pregnancy

BACKGROUND

Hypertensive disorder in pregnancy is common and the optimal ultrasound surveillance of the fetus in this setting is unclear.

AIM

The aim of this study is to assess the relationship between the fetal cerebroplacental ratio (CPR) and perinatal outcomes in pregnancies complicated by maternal hypertension.

MATERIALS AND METHODS

A retrospective cohort study was performed over ten years at a single centre. All women who had an ultrasound scan between 34 and 37 weeks gestation with a non-anomalous singleton pregnancy were included. The hypertensive cohorts were compared to a non-hypertensive cohort. Each cohort was divided into low CPR for gestational age, or normal/high CPR and these were correlated with intrapartum and perinatal outcomes.

RESULTS

A low CPR in a hypertensive pregnancy is associated with an increased risk of induction of labour, emergency caesarean section and poor perinatal outcome. This significance persists when adjusted for gestational age and birth weight. The diagnosis of pre-eclampsia combined with a low CPR markedly increases the risk of poor perinatal outcome, with 52.6% (P < 0.001) of fetuses in this group having either neonatal intensive care unit admission, respiratory distress, low Apgar score, or acidosis. The odds ratio of a fetus with low CPR in a woman with pre-eclampsia having a poor composite outcome is 4.09 (95% CI: 1.85-9.06).

CONCLUSION

There is an association between low CPR and the perinatal outcomes of pregnancies complicated by a hypertensive disorder. This association appears to be stronger in pregnancies complicated by pre-eclampsia than in other types of hypertensive disorders.

The role of abnormal cerebroplacental ratio in predicting adverse fetal outcome in pregnancies with scheduled induction of labor

OBJECTIVE

To investigate the role of abnormal cerebroplacental ratio (CPR) in predicting adverse fetal outcome in pregnancies with induction of labor.

METHODS

This prospective observational study conducted at Medeniyet University Göztepe Training and Research Hospital between December 1, 2018 and October 31, 2019 enrolled women with scheduled induction of labor at or beyond 37 weeks of pregnancy. Women with singleton non-anomalous fetuses with cephalic presentation and who had Bishop scores of 5 or less in pelvic examination were included in the study. Exclusion criteria were the presence of uterine scar history, non-cephalic presentation, multiple pregnancy, vacuum- or forceps-assisted delivery, and shoulder dystocia. Using fetal Doppler ultrasound, CPR was calculated (the ratio of umbilical artery to middle cerebral artery pulsatility index) and categorized into abnormal CPR (<1) and normal CPR (≥1). Data on maternal and delivery characteristics, fetal birth weight, and fetal complications were compared between the groups.

RESULTS

A total of 145 women were included, 28 in the abnormal CPR group and 117 in the normal CPR group. Multivariate analysis revealed labor induction at a later week of pregnancy (odds ratio [OR] 10.33, P = 0.001), lack of intrauterine growth restriction (IUGR) (OR 13.21, P = 0.001), fetal distress (OR 8.14, P = 0.003) or meconium aspiration (OR 159.91, P = 0.001), and umbilical artery pH values greater than 7.31 (OR 17.51, P = 0.015) to be associated with an increased likelihood of having normal (≥1) CPR values. Receiver operating characteristic analysis revealed association of normal CPR values with later labor induction (cut-off value of >38.3 weeks, P = 0.001), higher birth weight (cut-off value of >2460 g, P = 0.022) and higher umbilical artery pH (cut-off value of >7.31, P = 0.007).

CONCLUSION

Our findings revealed the significant role of abnormal CPR in predicting adverse fetal outcome in singleton pregnancies with scheduled induction of labor.