Medical Device Daily Washington Editor
SILVER SPRING, Maryland – Monday's workshop on interpartum electronic fetal monitoring aided by computer algorithms, sponsored jointly by FDA and the National Institutes of Health, was replete with explanations as to why this approach to diagnosis has made little or no dent on modern healthcare despite more than a decade's worth of effort.
Still, the frustration has not blunted the interest in computer assisted diagnosis (CAD) for fetal health, which many hope can reduce the numbers of caesarian sections and other conditions that plague mother and child before and after birth. However, the meeting also served to reinforce a trend at FDA for the greater use of device-specific guidance, such as the guidance that will come out sometime in the next 18 months for CAD (Medical Device Daily, March 6, 2008). Whether that guidance will have to find its way into print before FDA can publish guidance for CAD fetal monitors remains to be seen.
Gary Hankins, MD, chairman of the department of gynecology at the University of Texas Medical Branch (Galveston, Texas) gave a clinical overview of intrapartum fetal monitoring and said "you're going to be monitored in the overwhelming majority" of cases, with continuous monitoring of the fetal heart rate and the uterine artery the most common.
Hankins said "electronic fetal heart rate monitoring is the predominant method used for assessing fetal oxygenation," even while other methods offer better measures. The problem is that "the fetus is largely inaccessible," to methods such as pulse oximetry. After discussing several other aspects of the current status of fetal monitoring, Hankins said FDA's question is "how do you judge a device that someone wants to bring to market?"
Colin Pollard, the branch chief for obstetric and gynecology devices at the Office of Device Evaluation at FDA, gave a quick look at the field of fetal monitors. He said that a company that developed a fetal pulse oximeter obtained a PMA for the device in 2000 with "significant post-approval study conditions," but that the clinical community was not happy with the device. Pollard said the post-approval study (PAS) for this device, the FOX trial, showed that there was "no impact on caesarean sections" and the $25 million the sponsor plowed into the device "basically went nowhere." This apparently is a reference to the Oxifirst, made by Nellcor (Boulder, Colorado).
Pollard also mentioned the Stan S31 fetal heart monitor, which was the subject of a pivotal study of nearly 5,000 cases done in Sweden. He mentioned that the advisory committee "was concerned that this was done in Sweden," but the sponsor, Neoventa Medical (Molndal, Sweden) conducted additional studies and won a PMA in 2005. All the same, Pollard said, doctors are "taking a healthy skepticism" toward the device, which functions by placing an electrode on the fetus's scalp.
Pollard said that given the uneven pace of progress in this area of medical science, "we feel that FDA has to communicate its methods and decisions better ... but we also need to access input from stakeholders," including the clinical community and payers, where clinical trial design is concerned.
Julian Parer, MD, a professor of obstetrics and gynecology at the University of California at San Francisco, discussed clinical decision-making and how technology interfaces with that process. He mentioned an article published in 1995 that indicated that fetal monitoring is associated with a 21% increase in caesarean section.
Parer said that despite the iffy outcomes associated with the use of fetal monitoring, "the vast majority of parents are quite supportive." Still, he said he sees "a poor relationship between the results of randomized, controlled clinical trials (RCTs) and clinical acceptance" of the 11 fetal monitoring technologies approved by FDA, but he also remarked, "the simpler the technique, the more likely there is to be acceptance of RCT results." Parer said "the striking discordance" between heart rate monitoring and other ways of assessing fetal health has a lot to do with the low rate of acceptance.
RCTs focus on "hard information" and measures "that are easily countable," Parer observed, and those trials "assess some clinical findings poorly," such as patterns of illness severity and short-term outcomes. He said that broadly speaking, RCTs are "the Holy Grail, but just have not been good enough" for fetal heart rate monitoring.
Gary Eglington, MD, of New York Hospital Medical Center (Queens) discussed the difficulty of establishing endpoints for trials for intrapartum fetal monitors. He led with the observation that the mathematics used to assess fetal monitoring data "has been well worked out for more than 100 years" and that Sir Joseph Barcroft tackled this application just after World War I. Hence, "fetal heart-rate investigation has been going on for some time."
"It turned out to be not the rate that was important, but the rhythms" of the fetal heart, Eglington said, but the lack of a standardized interpretation still bedevils clinical practice despite efforts on the parts of the National Institute of Child Health and Human Development at NIH and the American College of Obstetricians and Gynecologists (Washington).
As for clinical trial design, Eglington said "there are a number of candidate endpoints, but cerebral palsy is probably not one of them."
With some fetal monitoring systems on the market, there are more than 400 rules to determine whether a clinician should test the fetal scalp for oxygen saturation, which he said is a lot to ask the practicing doctor, who is usually pretty swamped. This highlights the need for automation, Eglington said.
"Perhaps the issue again is that the experts – who taught the machines the rules – haven't figured out the rules." However, "an adaptive system, a machine that can learn," could perhaps get the rules right.
Deborah Wing, PhD, associate professor and the director of the division of maternal fetal medicine at the Irvine Medical Center at the University of California Irvine, said, "I worship at the altar of the RCT," she said, but acknowledged that a large trial is needed to capture "rare events such as intrapartum fetal death."
Wing noted that the current draft of a trial that is tentatively dubbed the INFANT trial is remarkable in at least one respect. "The thing that's shocking to me that this is funded to the tune of six million pounds," which is not much considering the number of patients to be followed.
This trial is "a huge trial of 46,000 women," Wing said, randomized and followed to 35 weeks using K2MS Guardian system, made by K2 Medical (Devon, UK), which "can be used to allocate patients randomly" to CAD or to unaided review. She said the Guardian was chosen by the UK's National Health System (NHS) because the algorithms "were designed to be flexible." Another point of interest on the part of NHS was that the trial "adopts a rule-based approach so that a clinician can ask why a treatment is indicated" by the system. In this trial, "poor neonatal outcomes is a composite" made of all deaths, and several morbidities, but "they also will be looking at process outcomes and will evaluate the amount of time taken to respond to alarms and alerts."
Sponsors of new devices will have to validate the algorithms used in their machines prior to enrolling for a pivotal study, and Wing asked rhetorically, "are existing databases okay?" She said one option is the data set from the Fox trial, which enrolled 10,000 subjects. "Safe Labor has promise" as a data source, she said, but she noted that some have concerns about data harmonization.
Large providers may offer some data as well, such as the Kaiser Permanente networks, but Wing warned that "clinical management is going to be relatively heterogeneous" in such a data set.