Nelson, Scott M.

Treatment-induced ovarian function loss is a significant concern for many young patients with breast cancer. Accurately predicting this risk is crucial for counselling young patients and informing their fertility-related decision-making. However, current risk prediction models for treatment-related ovarian function loss have limitations. To provide a broader representation of patient cohorts and improve feature selection, we combined retrospective data from six datasets within the FoRECAsT (Infertility after Cancer Predictor) databank, including 2679 pre-menopausal women diagnosed with breast cancer. This combined dataset presented notable missingness, prompting us to employ cross imputation using the k-nearest neighbours (KNN) machine learning (ML) algorithm. Employing Lasso regression, we developed an ML model to forecast the risk of treatment-related amenorrhea as a surrogate marker of ovarian function loss at 12 months after starting chemotherapy. Our model identified 20 variables significantly associated with risk of developing amenorrhea. Internal validation resulted in an area under the receiver operating characteristic curve (AUC) of 0.820 (95% CI: 0.817–0.823), while external validation with another dataset demonstrated an AUC of 0.743 (95% CI: 0.666–0.818). A cutoff of 0.20 was chosen to achieve higher sensitivity in validation, as false negatives—patients incorrectly classified as likely to regain menses—could miss timely opportunities for fertility preservation if desired. At this threshold, internal validation yielded sensitivity and precision rates of 91.3% and 61.7%, respectively, while external validation showed 92.9% and 60.0%. Leveraging ML methodologies, we not only devised a model for personalised risk prediction of amenorrhea, demonstrating substantial enhancements over existing models but also showcased a robust framework for maximally harnessing available data sources.

Background: International guidelines recommend targeted screening to identify gestational thyroid dysfunction. However, currently used risk factors have questionable discriminative ability. We quantified the risk for thyroid function test abnormalities for a subset of risk factors currently used in international guidelines. Methods: We included prospective cohort studies with data on gestational maternal thyroid function and potential risk factors (maternal age, body mass index [BMI], parity, smoking status, pregnancy through in vitro fertilization, twin pregnancy, gestational age, maternal education, and thyroid peroxidase antibody [TPOAb] or thyroglobulin antibody [TgAb] positivity). Exclusion criteria were pre-existing thyroid disease and use of thyroid interfering medication. We analyzed individual participant data using mixed-effects regression models. Primary outcomes were overt and subclinical hypothyroidism and a treatment indication (defined as overt hypothyroidism, subclinical hypothyroidism with thyrotropin >10 mU/L, or subclinical hypothyroidism with TPOAb positivity). Results: The study population comprised 65,559 participants in 25 cohorts. The screening rate in cohorts using risk factors currently recommended (age >30 years, parity ≥2, BMI ≥40) was 58%, with a detection rate for overt and subclinical hypothyroidism of 59%. The absolute risk for overt or subclinical hypothyroidism varied <2% over the full range of age and BMI and for any parity. Receiver operating characteristic curves, fitted using maternal age, BMI, smoking status, parity, and gestational age at blood sampling as explanatory variables, yielded areas under the curve ranging from 0.58 to 0.63 for the primary outcomes. TPOAbs/TgAbs positivity was associated with overt hypothyroidism (approximate risk for antibody negativity 0.1%, isolated TgAb positivity 2.4%, isolated TPOAb positivity 3.8%, combined antibody positivity 7.0%; p < 0.001), subclinical hypothyroidism (risk for antibody negativity 2.2%, isolated TgAb positivity 8.1%, isolated TPOAb positivity 14.2%, combined antibody positivity 20.0%; p < 0.001) and a treatment indication (risk for antibody negativity 0.2%, isolated TgAb positivity 2.2%, isolated TPOAb positivity 3.0%, and combined antibody positivity 5.1%; p < 0.001). Twin pregnancy was associated with a higher risk of overt hyperthyroidism (5.6% vs. 0.7%; p < 0.001). Conclusions: The risk factors assessed in this study had poor predictive ability for detecting thyroid function test abnormalities, questioning their clinical usability for targeted screening. As expected, TPOAb positivity (used as a benchmark) was a relevant risk factor for (subclinical) hypothyroidism. These results provide insights into different risk factors for gestational thyroid dysfunction.

Abstract Context Interpretation of thyroid function tests during pregnancy is limited by the generalizability of reference intervals between cohorts due to inconsistent methodology. Objective (1) To provide an overview of published reference intervals for thyrotropin (TSH) and free thyroxine (FT4) in pregnancy, (2) to assess the consequences of common methodological between-study differences by combining raw data from different cohorts. Methods (1) Ovid MEDLINE, EMBASE, and Web of Science were searched until December 12, 2021. Studies were assessed in duplicate. (2) The individual participant data (IPD) meta-analysis was performed in participating cohorts in the Consortium on Thyroid and Pregnancy. Results (1) Large between-study methodological differences were identified, 11 of 102 included studies were in accordance with current guidelines; (2) 22 cohorts involving 63 198 participants were included in the meta-analysis. Not excluding thyroid peroxidase antibody–positive participants led to a rise in the upper limits of TSH in all cohorts, especially in the first (mean +17.4%; range +1.6 to +30.3%) and second trimester (mean +9.8%; range +0.6 to +32.3%). The use of the 95th percentile led to considerable changes in upper limits, varying from –10.8% to –21.8% for TSH and –1.2% to –13.2% for FT4. All other additional exclusion criteria changed reference interval cut-offs by a maximum of 3.5%. Applying these findings to the 102 studies included in the systematic review, 48 studies could be used in a clinical setting. Conclusion We provide an overview of clinically relevant reference intervals for TSH and FT4 in pregnancy. The results of the meta-analysis indicate that future studies can adopt a simplified study setup without additional exclusion criteria.