5 questions on new dengue vaccine analysis

Q: What was your involvement in the new analysis?

LUEDTKE: Sanofi had already run three large trials, they had collected the data and they were continuing follow-up on the participants. There was a certain “missing data” problem that we helped them tackle.

For most people, we were missing the baseline information on whether they were seropositive or seronegative upon vaccination. There is a test for that, but it’s not a rapid test, and so only a small section of people were tested; this prior exposure information was available for only 10 to 17 percent of the people in the trials.

So Sanofi developed a new test to measure prior infection with dengue, and they applied it to post-vaccination blood samples from almost all dengue cases in the trials. Our analysis applied machine-learning techniques to analyze information from this new test, and we found it could infer whether any of the participants were infected with dengue or not at the time of vaccination. It’s much more complicated than what statisticians usually run, but it solved the missing-data problem.

GILBERT: Sanofi had their own statisticians. They did a conventional analysis, which was still reasonably sophisticated. I hooked Alex into it, and he devised the new method. He is doing some pretty modern, cutting-edge statistical methods that are more robust and efficient than is traditional practice in clinical trials. Part of the excitement is that this modern method is being put into the NEJM along with the conventional methods. It’s part of the process of new methods starting to penetrate clinical practice.

The new statistical methods have so many applications. They really could apply almost in any clinical trial, not just vaccine trials. They could be used in any trial that’s measuring biomarkers and wants to ask questions about whether the biomarker modiifies the treatment effect.

Q: What about this vaccine? Since not every infection causes clear symptoms and there is no rapid test for prior exposure, what do we do now?

GILBERT: Where we are now is that health officials have to make these tough decisions. The WHO, after Sanofi’s press release back in the fall, revised their guidelines to recommend that physicians only give the vaccine to people they believe have been previously infected with dengue. It’s a little hard to imagine how each physician will know or not because they don’t have a [rapid] test.

If this study would never been done and they just gave the vaccine in highly endemic settings, it would still do way more good than harm just because the vaccine is so great in seropositives and there is a relatively small bit of harm in the serognegatives. So it starts to get into issues of what kind of risk are people willing to tolerate; how much individual risk is acceptable for a public good.

The vaccine is licensed [for children 9 and over] in over 13 countries now, and each one of those settings is wrestling with this issue. Some of those countries are still favoring using the vaccine in a widespread level, I assume because they’re opting for total amount of severe dengue cases saved. Then other countries are taking a more conservative position. They’re barring the vaccine totally because they think even a single case caused by this vaccine is unacceptable. That’s part of the discussion that is happening.

People do feel as though a point-of-care test to tell if you’re baseline seropositive or seronegative would be the ideal solution, if that could be found to be cheap and rapid. That is the only solution that could truly satisfy all concerns. But a much better solution is to get a vaccine that doesn’t need a point-of-care test.

Q: Is such a vaccine possible?

GILBERT: Historically, vaccines have tended to be very good, well over 95 percent efficacious. This could be representing a new era, and now that we’re trying to get vaccines for the tough, genetically diverse pathogens, we’re going to be seeing this type of issue more — the possibility that there’s protection for some and harm for others.

What’s also happening is that now we’re measuring more things. Systems vaccinology is becoming a hot topic. We’re measuring more biomarkers at baseline, during and after whole vaccination series. Now we have the opportunity to discover that the vaccine may be working in some and harming others, where before we never even did these analyses.

To throw in one more buzz phrase, which is “personalized” or “stratified” medicine: Historically vaccines have never been thought of as personalized or as “precision public health.” In fact, that’s exactly not been the goal. The goal has been universal coverage. This study is inching us toward that new era where we might have to personalize, or stratify, vaccines.