Development of the DONOR prediction model on the risk of hypertensive complications in oocyte donation pregnancy: study protocol for a multicentre cohort study in the Netherlands

Study setting and design

The present study is performed within the DONOR project, on the DONation of Oocytes in Reproduction. This is a multicentre cohort study designed to determine the association between fetal–maternal HLA mismatches and the development of hypertensive complications in pregnancy.31 The study will be performed in seven fertility centres in the Netherlands, with the Leiden University Medical Center as coordinating centre.

Eligibility criteria

Patients with the following criteria are eligible for inclusion:

1. Pregnant or delivered after OD, embryo donation or surrogacy pregnancy. The fertility procedure can be performed in the Netherlands or abroad.

2. Pregnancy duration ≥20 weeks.

3. Visit(ed) one of the participating centres.

Patients who are mentally or legally incapable of signing the informed consent form and patients with known chromosomal abnormalities (such as Turner syndrome) or fetal abnormalities will be excluded.

Study population and recruitment

In the prospective cohort, patients who visit the gynaecology department of the participating centres will be assessed for eligibility by the attending physician or nurse. Prospective patient recruitment for the DONOR study started at the coordinating centre on 1 September 2016. Recruitment at Erasmus MC has started in 2019. All other centres will start recruitment in November 2023. Prospective patient recruitment is expected to end on 1 January 2026. In the retrospective cohort, patients who underwent successful OD at one of the participating centres between 2004 and the start of the current study will be approached for participation.

Prior to inclusion, all patients will receive written information and an informed consent form, which comprises a request to obtain permission for gathering data from medical records and a sample of maternal peripheral blood and one sample of umbilical cord blood in the prospective cohort and one sample of saliva of the child in the retrospective cohort. Participants are explicitly informed that participation is entirely voluntary and that they retain the right to withdraw at any time without consequences for subsequent care.

In order to optimise inclusion rate and enable representation of the Dutch population that applies to OD, more centres will be approached for collaboration.

Data collection

For the prospective cohort, maternal, paternal and donor baseline characteristics and obstetric history will be collected at the initial visit. After delivery, obstetric data will be acquired from the patients’ medical records. In the DONOR study, the follow-up period extends from the initial visit until 6 weeks after delivery. Throughout this period, patients will have regular prenatal and postnatal check-ups in the hospital or primary care, which will be documented in medical records by the attending physician or midwife. Only data from participants with a pregnancy duration ≥20 weeks will be used in further analysis.

From the retrospective cohort, maternal baseline and pregnancy characteristics will be collected from the patients’ medical records. All available data will be entered into the data management system Castor.32 An overview of patient data that will be collected is shown in table 1.

Blood and saliva sampling

In the prospective cohort, peripheral maternal blood and umbilical cord blood will be collected for DNA isolation and HLA typing. In the retrospective cohort, peripheral maternal blood and a saliva sample from the child will be obtained. From these samples, DNA will be extracted and HLA typing will be performed for loci HLA-A, HLA-B, HLA-C, HLA-DQ and HLA-DR, using the Reverse Sequence-Specific Oligonucleotides PCR technique.33 Additionally, the number of fetal–maternal HLA mismatches will be calculated at the national reference laboratory for histocompatibility testing (Leiden University Medical Center, LUMC), based on discrepancy in the HLA-A, HLA-B, HLA-C, HLA-DR and HLA-DQ. The definition of high number of HLA mismatches is defined as >5 fetal–maternal HLA mismatches.

In case the donor is known and willing to participate, one sample of peripheral blood will be obtained in order to perform HLA typing as well.

Control of bias

The Prediction model Risk Of Bias Assessment Tool (PROBAST) will be used to minimise the risk of bias.34 The PROBAST is developed as a general tool for critical appraisal of prediction model studies and consists of 4 domains (participants, predictors, outcome and analysis) containing 20 signalling questions to minimise risk of bias. In this study, most (18/20) PROBAST criteria will be met. The following two PROBAST criteria need attention to minimise the risk of bias.

• Item 1.2 (Domain 1: participants). It is possible that some participants appear multiple times in our dataset, as some participants will conceive through OD more than once. As repeated measurements of the same participant are not independent, robust standard errors will be calculated to account for this. Additionally, using information from medical records poses a risk of information bias and the retrospective cohort could be prone to response and recall bias, for which caution will be taken.

Furthermore, we aim to implement the prediction model as a tool for preconceptional counselling in women considering OD, whereas development and validation will be performed on a population that is already ≥20 weeks pregnant through OD. We anticipate that this approach will not significantly impact the performance of the prediction model since the majority of covariates remain constant during pregnancy and we expect the time interval between preconceptional counselling and pregnancy to be short.

• Item 2.3 (Domain 2: predictors). The predictive value of fetal–maternal HLA (mis)matches will be assessed, as previous work from our group28 29 35 suggests this could be of significance. However, HLA typing of the child cannot be performed before or during pregnancy. As the prediction model is intended to be applied in this exact period, the variable of fetal–maternal mismatches will not be used as a candidate predictor in the model. Instead, the variable will be assessed to determine its individual prognostic value on the risk of developing hypertensive disease. Moreover, proximate predictors such as ‘the expected amount of fetal–maternal mismatches’ based on the donor HLA typing or ‘relation between patient and oocyte donor (familial “yes” or “no”)’ will be used as candidate predictors in our prediction model.

Furthermore, the current study will be reported in line with the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guideline, which aims to improve the reporting of studies developing, validating, or updating a prediction model.36 37 Details regarding the control of bias can be found in online supplemental file 1. Finally, any discrepancies between the protocol and the final study will be thoroughly documented. Transparent reporting could help mitigate bias.

Sample size calculation

The required sample size was calculated according to the approach of van Smeden et al and Riley et al.38–40 Based on the current absolute risk of hypertensive complications in patients undergoing OD between 10% and 20%,14 16 18 19 a number of six candidate predictor parameters, an anticipated model performance of 0.15, a shrinkage factor of 0.90 and a mean absolute prediction error of 0.05 between observed and true outcome probabilities, the required sample size should be at least 541 OD pregnancies.

According to the latest data, approximately 80 successful oocyte or embryo donation procedures are performed yearly in the Netherlands.41 In addition, a significant number of Dutch women undergo oocyte or embryo donation procedures in countries outside the Netherlands such as Spain, where the laws and regulations concerning these procedures differ.42 For pregnancy check-ups and delivery, these women return to the Netherlands, which poses the opportunity of participation in our study. Currently, the DONOR cohort31 consists of 300 OD pregnancies, of which data on fetal–maternal HLA matching is available for 150 pregnancies. To increase inclusion rate and to optimise representation of the Dutch population that applies to OD, also other centres that perform OD are approached for collaboration. The objective is to collect data from approximately 150 additional pregnancies. Assuming that 50% of all eligible patients visiting one of the participating centres will be included in this study and that approximately 15% is lost to follow-up or does not reach pregnancy>20 weeks, it is estimated that the required sample size will be reached within 24 months.

Study outcomes

The primary outcome measure, hypertensive complication in pregnancy, is defined according to the 2021 International Society for the Study of Hypertension in Pregnancy (ISSHP) classification.43 PIH is defined as new onset hypertension with diastolic blood pressure ≥90 mm Hg and/or systolic blood pressure ≥140 mm Hg detected after 20 weeks of gestation. PE is defined as PIH accompanied by one or more of the following: (1) proteinuria, (2) maternal organ dysfunction and (3) uteroplacental dysfunction (fetal growth restriction or abnormal Doppler findings).22 43 Superimposed PE is defined by chronic hypertension combined with evidence of uteroplacental dysfunction. Severe hypertension is defined as blood pressure ≥160 mm Hg systolic or ≥110 mm Hg diastolic. HELLP syndrome, which is a serious manifestation of PE, is defined as haemolysis with a microangiopathic blood smear in combination with elevated liver enzymes and a low platelet count.43

To enhance generalisability and applicability of a final model, predictor selection should be well defined, standardised and reproducible. In addition, selection of predictors should be based on literature and clinical reasoning. Therefore, the following candidate predictors will be considered28 44–46:

1. Fetal–maternal genetical difference, including

2. Age recipient (continuous)

3. Nulliparity recipient (dichotomous)

4. Plurality recipient (categorical)

5. Body Mass Index recipient (continuous)

6. Smoking recipient (dichotomous)

7. Natural versus artificial cycle (dichotomous)

8. Ethnicity recipient (categorical)

9. Medical history of recipient related to hypertensive complications in pregnancy (dichotomous)

10. Family history of recipient related to hypertensive complications in pregnancy (dichotomous)

11. Age oocyte donor (continuous)

12. Sperm origin (father vs donor) (dichotomous)

13. Use of medication during pregnancy, specifically acetylsalicylic acid (dichotomous)

Data analysis

All statistical analyses will be performed using R.

Model development

A multivariate logistic regression model will be developed for the binary outcome hypertensive complications in pregnancy ‘yes’ and ‘no’. The model will include the preselected candidate predictors.

As part of the secondary objectives, we aim to predict the risk of developing HELLP and time to development of PE. To achieve this, multivariate logistic regression models will be developed for the binary outcomes early PE (≤34 weeks) and HELLP ‘yes’ or ‘no’.43 Additionally, in order to predict the time of development of PE, a multivariate linear regression model will be developed for the continuous outcome gestation age at development of PE, with a minimum of 20 weeks and a maximum of 42 weeks. Possible non-linear associations between the continuous predictors and the outcome will be examined, using spline models. Potential multiplicative interaction will be explored by adding interaction terms to the model, and use of Lasso regression for variable selection will be considered.47

Model performance

Internal validation will be performed to assess model performance, and discrimination and calibration will be determined. Discrimination describes the ability of the model to discriminate between events and non-events and will be evaluated with the area under the receiver curve. Calibration describes the relation between the observed risks within the population and the predicted risks and will be assessed by calibration plots, plotting the relation between the observed and predicted risks.48 Internal validation will be performed using the bootstrapping method with 100 bootstrap samples. The calibration slope from the bootstrapping procedure will be used as uniform shrinkage factor to calculate new regression coefficients to prevent overfitting.47 External validation will be performed using the ‘DONation of Oocytes in Reproduction individual participant data’ (DONOR IPD). The DONOR IPD database is currently being set up and will include at least 2301 women pregnant after OD and beyond 20 weeks of gestation.30 In this study original data from multiple studies will be combined into a single database is therefore suitable for model performance assessment and external validations.49 50

Handling of missing data

In case the amount of missing data in some of the predictors is substantial and cannot be ignored, a multiple imputation method will be applied.51 52

Patient and public involvement

During the development of the study protocol, Freya was consulted for input and advice. Freya is the Dutch society for patients with fertility problems and many patients who apply to OD are members of the association. Study information will be published on their website, and information on progress and results will be presented to patients during meetings, patient days and webinars.