Pregnant patients with sickle cell disease (SCD) have a significantly elevated risk for severe maternal morbidity (SMM) and preterm birth, but a novel risk calculator shows promise for predicting maternal complications and improving outcomes, according to investigators.

The findings from two studies highlight progress made toward improving pregnancy outcomes for patients with SCD, reported Aaron N. Cheng, MD, first author of the research projects. He presented both studies in two sessions at the American Society of Hematology (ASH) annual meeting.

SMM Incidence in Patients With SCD

For one study, the investigators conducted a nationwide electronic health record database analysis. The analysis showed an SMM incidence of 25% among 2679 pregnant individuals with SCD, compared with 2.4% among more than 2.8 million Black individuals without SCD, Cheng reported at the ASH annual meeting.

After adjusting for covariates including year of delivery, gestational age at birth, pregnancy-related conditions, and socioeconomic status, patients with SCD had a significantly greater risk for SMM than all controls without SCD (adjusted odds ratio [aOR], 3.96) and Black controls without SCD (aOR, 3.66).

The differences persisted when SMM was redefined to exclude sickle cell crisis and transfusion, noted Cheng, a hematology and oncology fellow at the University of Pennsylvania, Philadelphia.

The risk for preterm birth, which was encoded as a binary variable for the analysis, was also significantly greater among those with SCD vs all controls without SCD (aOR, 1.70) and only Black controls without SCD (aOR, 1.78), he said.

Earlier studies, including a 2023 analysis of the National Inpatient Sample from 2012 to 2018, have also demonstrated an increased risk for SMM in individuals with SCD, but they have been limited by older data and a lack of information on key clinical variables, Cheng explained.

To address those limitations, he and his colleagues performed the current analysis using data from 2020-2024, encompassing the COVID era, from the Epic Cosmos nationwide database of more than 300 million patients. 

The findings suggest that underlying disease biology and structural deficiencies in care, such as those related to racial differences in healthcare access and outcomes, contribute to the differences in outcomes among patients with SCD, he said. 

They also reinforce an ongoing urgency to improve maternal outcomes in this high-risk population, he added, noting that despite this well-established risk, there is no validated, widely adopted tool to predict whether a pregnancy is most likely to result in maternal complications.

New vs Old Prediction Model

To address the need for an accurate risk stratification tool to guide preconception counseling and perinatal care for patients with SCD, Cheng and his colleagues launched a multi-institutional effort in the United States to develop such a tool, as part of the other study he presented at the ASH annual meeting.

“Compared to the only previously published model [which was developed at Mount Sinai Hospital in Toronto], our model exhibited superior discriminative performance in predicting severe maternal complications,” he said during a separate presentation at the ASH meeting.

For the prediction model, which incorporates readily available clinical and hematologic predictors to risk-stratify pregnancies, the investigators conducted a retrospective cohort study of 231 pregnancies among 167 unique individuals with SCD from two large data repositories. Manual chart reviews were conducted to confirm clinical information, and baseline laboratory data were collected.

As a benchmark, the investigators applied the previously published prediction model, Cheng said, noting that the older model had important limitations.

The Toronto model was “a single-center model that has not been externally validated and may not be generalizable,” he said. Maternal events included in that model were:

• Acute anemia
• Cardiac, pulmonary, hepatobiliary, musculoskeletal, skin, splenic, neurologic, or renal complications
• Multi-organ failure
• Venous thromboembolism
• Admission-requiring vaso-occlusive events
• Red cell transfusion
• Mortality
• Hypertensive disorder of pregnancy

Fetal events included:

• Preterm birth
• Small-for-gestational-age
• perinatal mortality

The new model was developed using a more stringent and clinically focused composite outcome to identify potential complications, incorporating factors such as SCD-related acute chest syndrome, stroke, hemolytic crisis, urgent red cell exchange, and sepsis, as well as obstetric-related complications, such as preeclampsia and venous thromboembolism. 

The area under the receiver operator curve (AUC) achieved using the new model was 0.831, which was adjusted “only marginally” to 0.819 after internal validation. This was superior to the Toronto model, which showed an apparent AUC of 0.811 using its original composite outcome, which declined to 0.743 when evaluated using the more stringent and clinically actionable composite outcome of the new model. This indicated “substantially reduced discriminative performance” of the Toronto model, Cheng explained, noting that the model performance persisted after “internal validation using a bootstrapping approach to account for the potential risk for overfitting.”

“[These results] demonstrate the ability of our model to discriminate between risk profiles using easily obtainable clinical and laboratory features,” he said during the meeting.

Audience member Kevin H.M. Kuo, MD, co-first author of the Toronto model study, applauded the “excellent work” of Cheng and his team on the development of the new model — in particular for “externally validating a model with AUC 0.811.” Additionally, Kuo, of the University Health Network in Toronto, Ontario, Canada, expressed approval of the researchers’ leveraging of the Epic Cosmos database to assess SMM in patients with SCD.

Kuo also made a suggestion to Cheng and colleagues. He said using a continuous variable that looks at the number and range of factors instead of the binary variable used, which involved simply assessing for the presence or lack of presence of those factors to assess maternal risk, “would have been much more powerful” for the analysis.

Cheng responded that the binary approach allowed for higher power for certain “very rare events” included in the SMM definition, but the use of a continuous variable would be “something to look into down the line.”

Efforts are ongoing to better understand SMM in this population using the Epic Cosmos database, Cheng noted, adding that “we also need to better characterize the transfusion burden in this population and its association with baseline factors.” The lead author said that he and his team also plan to “leverage the [Epic Cosmos] database to look at additional outcomes, such as birthweight for gestational age and ICU admissions.”

They are currently working to refine and conduct further external validation of the new prediction model using independent databases, as well. The goal is to create an evidence-based tool to guide the risk-adapted management of pregnant individuals, Cheng noted.

“My ultimate vision is to create a tool that can help clinicians tailor their clinical decision-making,” he said. “I envision a clinical risk calculator that can identify individuals who are at the highest risk and might prompt discussion of more aggressive interventions like prophylactic transfusions or enrollment in a clinical trial.

“On the other hand, I think such a model can be very important in identifying the lowest-risk population who might be spared this type of intervention and simply proceed with routine surveillance.”

Cheng reported having no relevant disclosures. 

Sharon Worcester, MA, is an award-winning medical journalist based in Birmingham, Alabama, writing for Medscape, MDedge, and other affiliate sites. She currently covers oncology, but she has also written on a variety of other medical specialties and healthcare topics. She can be reached at sworcester@mdedge.com or on X: @SW_MedReporter.