1- or 2-step: Outcomes studied in GDM screening

Anne Paxton

June 2021—If screening for gestational diabetes mellitus were a dance competition, it might have a contest between quickstep and paso doble as its signature event. That tournament could pit the one-step testing protocol (twice as likely to diagnose GDM) against the two-step testing protocol (significantly easier for pregnant women to adhere to).

Which would prevail? In the real-life dance-off between one-step and two-step, despite helpful new input from a large-scale clinical trial, it may still be quite some time before the judges are able to declare a winner.

In Europe, consensus is strong that the one-step screening of pregnant women for gestational diabetes mellitus, which requires the patient to arrive in a fasting state, is preferable to two-step screening. But without conclusive evidence, medical opinion in the United States on this issue has long been divided. In a notable example of the disagreement, the American Diabetes Association (ADA) sides with Europe in favoring one-step while the American College of Obstetricians and Gynecologists favors the two-step approach.

The dissension about GDM screening goes beyond that, says David B. Sacks, MB, ChB, senior investigator and chief of the clinical chemistry service for the National Institutes of Health Clinical Center. From organization to organization, “People can’t agree on whether you should screen for GDM, when you should screen, how you should screen, how much glucose you should give during screening, whether the protocol should last two hours or three hours, or what your cutoff should be.”

A new study published March 11 in the New England Journal of Medicine doesn’t try to settle the dispute over which protocol is superior (Hillier TA, et al. N Engl J Med. 2021;384​[10]:895–904). But the large pragmatic, randomized clinical trial of gestational diabetes screening offers findings on a related matter: Are there differences in maternal and perinatal primary outcomes depending on which screening test is used?

GDM experts agree that the findings of the study, despite its limitations, shed light on the best way to screen pregnant women for a condition that affects six percent to 25 percent of them (depending on diagnostic criteria) by increasing the risk of stillbirth, neonatal death, and multiple serious conditions in mothers and their babies.

The Kaiser study consisted of 23,792 pregnant women who received care at Kaiser Permanente Northwest or Kaiser Permanente Hawaii and were randomly assigned to one-step screening or two-step screening. (The one-step protocol entails a glucose tolerance test in which the blood glucose levels are obtained fasting and for two hours after oral administration of a 75-g glucose load. Two-step screening includes a glucose challenge test in which the blood glucose level is obtained one hour after oral administration of a 50-g glucose load in the nonfasting state, and, if the first test is positive, a three-hour oral glucose tolerance test with a 100-g glucose load in the fasting state as the second diagnostic step.)

The trial found that the single-step approach resulted in detection of GDM in twice as many women as the two-step screening, but there were no significant between-group differences in the risks of the primary outcomes relating to perinatal and maternal complications. Primary outcomes were a GDM diagnosis, hypertensive disorders of pregnancy, primary cesarean delivery, large-for-gestational-age infants, and the perinatal composite outcome.

In the same issue, Brian Casey, MD, of the University of Alabama Department of Obstetrics and Gynecology, in his editorial, writes that the perinatal benefits of the GDM diagnosis with the use of the single-step approach “appear to be insufficient to justify the associated patient and health care costs of broadening the diagnosis.” He adds: “Refocusing attention on interventions in women who are at risk for the development of diabetes is more likely to yield substantive benefits.”

There’s good evidence from earlier trials that screening for gestational diabetes in pregnancy improves perinatal outcomes, says the lead author of the study, endocrinologist Teresa A. Hillier, MD, MS, distinguished investigator at the Kaiser Permanente Center for Health Research in Portland, Ore., and Honolulu, Hawaii. “And those trials’ data related to the two-step screening. But another study called the Hyperglycemia and Adverse Pregnancy Outcome [HAPO] trial looked at one-step testing. It found in 2008 that there was a linear relationship: The higher the glucose, the more problems with large-for-gestational-age infants and other outcomes. The study told us that C-peptide levels increased with hyperglycemia too.”

Because of this linear relationship, influential standard-setting groups including the ADA and the International Association of the Diabetes and Pregnancy Study Groups (IADPSG) recommended a single-step approach to diagnose GDM using a fasting sample. But since outcome research had been restricted to two-step–tested patients, “there wasn’t any evidence comparing outcomes head to head with both approaches,” Dr. Hillier explains.

That led to a much more robust debate among experts at the ADA and ACOG and no resolution. “At a consensus conference in 2013, the National Institutes of Health said, ‘We can’t make a recommendation yet based on current evidence.’ Everybody was hoping that identifying milder cases would improve outcomes, but we didn’t know.”

“So what we set out to do in our study that was unique is, because the one-step and two-step were both clinically recommended test protocols, we could randomize the study as part of clinical care and waive the need for individual consent, as long as participants have the right to opt out.” That was the right design for answering the question, Dr. Hillier says, because it allowed the researchers to do the trial on the scale of the whole population of pregnant women, making it possible to scrutinize different associated outcomes in mother and baby. “So it included vulnerable groups that typically wouldn’t agree to be involved in a clinical trial, including our Medicaid population and other demographics that don’t tend to be as involved in research.”

The “pragmatic” element of the study was that although patients were assigned randomly initially to either one-step or two-step, they and their clinicians were free to change them to the other testing protocol based on clinical judgment. The institutional review boards waived the need for individual consent because both tests were clinically recommended, as long as there was an option to opt out of randomized assignment based on clinical judgment. Over the course of the trial, there was more opting-out for the two-step approach, and the research team needed to extend the duration of the trial to obtain sufficient numbers. “But that was the pragmatic part of the trial; the participants had to have the option to opt out. And we did have that adherence issue,” Dr. Hillier says.

“We designed the study to ask, ‘Are there any differences between the two groups?’ Not ‘Is one better than the other?’ because that would require a different statistical model and setup. And we found there were differences in GDM incidence, with it being twice as common with the one-step versus the two-step. But there were no differences in any of our primary outcomes: large-for-gestational age, perinatal composite outcome, hypertension, and C-section rates, or in the secondary or safety outcomes.”

“There has been hope that diagnosing twice as many women by using the one-step protocol would improve outcomes,” Dr. Hillier says. “But there had never been a head-to-head trial comparing the treatment outcomes of the two tested cohorts. It was another advantage of the trial having been done at an integrated health care system where everybody received the same treatment after the diagnosis.”

The pragmatic advantage of the two-step is that everybody first does a nonfasting step. “Only about 20 to 25 percent of women fail that and have to go on to the longer three-hour glucose tolerance test. But if they don’t fail”—and that’s the majority—“they’re done.” To minimize the different levels of adherence, “we went out to providers and talked to them and we started getting pretty consistent feedback that they needed to switch to the two-step protocol with some patients just to ensure screening happened. And we couldn’t argue with that.”

In the end, 27 percent of the women randomized to the one-step ended up having the two-step. So the level of adherence was one of the study’s limitations. An additional limitation was the level of participation of racial minority populations. “Our two populations in Hawaii and the Pacific Northwest are representative of those areas, but not of the overall U.S.,” Dr. Hillier says. Black and American Indian populations were underrepresented in the study.

Some aspects of the laboratory testing performed and reported in the study raise additional concerns, potentially undermining its value, says the NIH’s Dr. Sacks. Part of his critique relates to different cutoffs used in the glucose testing performed at Kaiser’s two study sites. “At one site, they called 130 a positive and at the other they picked a cutoff of 140. Obviously, if you pick 130 as the cutoff you can identify more people than if you pick 140. That said, the number of people who went on to the second test was different but they just combined everybody.” (Dr. Hillier acknowledges there were two different regional standards for cutoffs. “Among the controversies around testing for GDM is what are the right thresholds for the one-step. Should it be 130, 135, or 140? And our two regions had different standards of care for the one-step,” she says. Additionally, she points to the comment about this in the supplementary appendix published online: “Only 92 cases of GDM were diagnosed at KPNW due to a GCT in the 130–139 range, suggesting that the difference in thresholds is unlikely to affect overall results.”)

In addition, Dr. Sacks notes, 18 percent, or 165, of the women in the two-step protocol did not meet the criteria for GDM based on that protocol, but they were treated as having GDM because their fasting glucose was increased. In such a case, “you’re not strictly following the protocol and it muddies the waters,” he says.

Also unsettling to him is the absence of information about how the glucose was measured for the trial. “People think glucose is a perfect test and it’s not even close to that,” he says. In this study, “there is no mention of how they handled the glucose samples, and this is critical because you can get glycolysis.” In the Hyperglycemia and Adverse Pregnancy Outcome study, which Dr. Sacks considers the best designed in terms of glucose handling, “they were very, very stringent. Everybody had to immediately place the samples on ice and spin down the blood within 30 minutes to get rid of the cells to minimize glycolysis. I have no idea what they did in the Kaiser study.”

One other limitation he sees is that the study could address only short-term outcomes, which leaves out significant factors that should influence a choice between one-step and two-step. For example, he says, “The HAPO study shows that 10 or 11 years following a GDM pregnancy, there was a big increase—a doubling—of obesity in the offspring of the GDM women.”

The overarching concern in any discussion of GDM screening is that the one-step protocol diagnoses a much larger number of women with GDM. “The main concern the ACOG gave in 2013 for not using the one-step protocol backed by IADPSG is that it increases by two- to maybe even threefold the number of women who are diagnosed with GDM,” he says. That leads to having 18 to 20 percent of pregnant women labeled as having GDM. It was for this reason that ACOG opted to support the two-step approach of a nonfasting glucose tolerance test with a three-hour test in a fasting state if glucose concentrations are high.

But a screening test, Dr. Sacks says, should err on the side of false-positives rather than risk missing cases through false-negatives. “You don’t want to miss anybody. So it’s better to have false-positives and send them off for a diagnostic test, which would be the second test.”