Daniel Feikin, MD, MSPH
Special thanks to Daniel Feikin, MD, MSPH, for sharing his insight on his team’s latest publication. Dr. Feikin has been with the U.S. Centers for Disease Control and Prevention (CDC) for 15 years, where he spent 6 years with the Respiratory Diseases Branch and also served as the Epidemiology Section Chief for the International Emerging Infections Program in Kenya. Dr. Feikin has worked on a number of projects including surveillance establishing the burden and epidemiology of respiratory and diarrheal illness. Since 2010, Dr. Feikin has served as the Director of the Epidemiology team at the International Vaccine Access Center (IVAC).
Published in the September issue of PLOS Medicine, Serotype-Specific Changes in Invasive Pneumococcal Disease after Pneumococcal Conjugate Vaccine Introduction: A Pooled Analysis of Multiple Surveillance Sites analyzes the rates of invasive pneumococcal disease after the introduction of the seven-valent pneumococcal conjugate vaccine (PCV7). Dr. Feikin and colleagues from IVAC and CDC conducted this meta-analysis which showed that a reduction in all-cause and vaccine-type invasive pneumococcal disease happened rather quickly and was sustained for seven years after the introduction of PCV across the study sites. In the following interview, Dr. Feikin shares his insight on the issue of serotype replacement and how this it relates to decision making regarding PCV use.
What is serotype replacement (SR) and why is it an issue?
We have a unique situation with pneumococcus and pneumococcal vaccines with respect to serotype replacement. With pneumococcus, serotypes are defined by the kind of sugar capsule that surrounds the bacterium, and 90 different types have been defined. This creates a challenge for vaccines, which are based on specific serotypes. The first iteration of this conjugate vaccine could only fit in seven serotypes. They put the seven most common serotypes that you’d find in the US and Europe into the vaccine – that covered about 85% of the disease in the U.S.
First, SR was observed in the nose, which is where pneumococcus resides. Most of the time, the bacteria doesn’t cause disease, it just lives there in back of the nose (the nasopharynx) and a small percentage of the time, the bacteria in the nose will go on to cause an infection. When you get vaccinated, you get rid of the seven pneumococcal serotypes contained in the vaccine from the nose – quite effectively and quite rapidly. But what we also saw was that the seven serotypes that were wiped out by the vaccine were replaced by other serotypes of pneumococcus. That phenomenon – when you get rid of [serotypes] and they are replaced by others – is dubbed SR. The concern was: this is what happens to the bacteria living in the nose; does the same thing happen with disease caused by those bacteria? Are you just getting rid of the disease caused by the seven serotypes in the vaccine and having it be replaced by pneumococcal disease (PD) from the other serotypes? If that were to happen, you have the paradoxical situation of a vaccine that works against the seven serotypes you want it to work against, but your overall disease rates might not change in the long run because of SR. It was a really important question about whether this vaccine had an overall impact on disease – we needed to clarify whether SR exists and how much it exists.
What type of study did you choose to look at this issue and why did you feel it was the best way to research this question?
We conducted a meta-analysis, using 21 databases that were identified, to look at IPD rates after PCV7 introduction. When you are considering PD, you have to think about several different types of disease. The first type of disease is called invasive pneumococcal disease (IPD). IPD is the most straightforward because it means bacteria have invaded a normally sterile site – usually the cerebrospinal fluid, (the fluid surrounding the meninges) or the blood – and you can grow the bacteria from this fluid so it can be detected and positively identified. But the much bigger burden of PD is pneumonia, and most of the time when you have pneumococcal pneumonia you’re not going to isolate the bacteria since you can’t get a sample from the lung to grow and it normally isn’t a sterile site. So, when you want to look at SR, you are limited to looking at IPD – where you are able to culture the pneumococcus. That’s why the study focuses on IPD.
The second thing to understand is that there have been lots of individual studies from different countries, or different sites from the same country, that have looked at the issue of SR. There were many reports that came out after vaccine introduction and they were variable. There were some sites that showed very little SR and big impact of the vaccine. Then, there were other studies that showed up to full SR, meaning that you get rid of the vaccine serotype [with] full replacement by non-vaccine serotype (NVT), so in those sites there were no changes in overall disease rates.
This led to a lot of confusion. As countries in Africa and Asia were thinking about introducing PCV, they were concerned and somewhat confused about this issue of SR and wanted to know if PCV was worth their investment. That was the impetus in doing the study and it’s unclear why there were different results in different places. Some of it might have been that there truly was more SR in some sites, but it also could have been methodological differences or that some sites had very small samples sizes. One outlier site was potentially getting a lot of attention and people were looking at that saying there was complete SR, whereas another site was not showing that. So the idea behind the study was to put all the sites together, weigh all the studies according to their size and findings, and see if we could come up with some sort of summary of SR.
In the course of this analysis, was there anything that was surprising to you?
I have two answers for that. First, although a lot of countries had introduced PCV7, there weren’t many sites that had done the type of surveillance that we felt would be appropriate for this type of analysis. There were sites that were doing surveillance, but when you boiled it down to who met stringent criteria for being able to evaluate SR, there weren’t as many as I thought. That was a lesson because people are out there publishing studies and making conclusions about SR with limitations in their surveillance, which we felt would affect their interpretation of SR, so we did not include those types of studies. We were fairly rigorous in our inclusion and exclusion criteria, so we only ended up including 21 studies.
The second thing that was somewhat surprising to me in the results was that when we looked at children, and when we looked at the overall impact of this vaccine on children, there was about a 50% reduction in overall IPD. So, overall IPD decreased by half. We saw that decrease happen by the first year after the vaccine was introduced and it stayed at roughly 50% reduction all the way out to about seven years. I thought that initially there would be a growing reduction in overall PD over time. The drop in the first year was greater than I thought – suggesting that this vaccine had a pretty fast impact on pneumococcal epidemiology in children – and it didn’t change a lot after that first year. Now, there was a lot going on with each progressive year, so vaccine types started to go down and they continued to go down out to seven years until they were virtually gone. You were getting a lot of change in the different serotypes starting in the first year after vaccine introduction and continuing out, but changes in the vaccine types and changes in the NVTs balanced each other out, so this overall 50% reduction remained fairly consistent over the seven years.
How could this study contribute to policy making?
I think that for policy makers (people that don’t think about PD everyday), I think their take-home point is that there’s about a 50% reduction in IPD in children and it lasts all the way out to seven years. There is SR; it exists. It’s a real phenomenon, but the proportion of disease caused by vaccine serotypes is greater than non-vaccine serotypes, so the vaccines do have an overall impact on disease in children. I think that’s the first message.
The second is that you do see a herd effect in adults. There is a decrease in overall IPD in adults. It’s not as great as what is seen in children and it’s a bit delayed, but at least in the countries that introduced PCV, there was herd protection among adults – which could potentially have a big impact for countries, especially low-income countries, that have a lot of PD in adults. For a decision maker, hopefully that’s the message they get from this.
Are there any plans that you know of to build on the evidence provided by this study, and investigate SR in higher valency PCV?
There are some clear limitations to this study. The first limitation is that this study looked at the impact of the 7-valent PCV. That vaccine is no longer made. Countries are introducing either 10-valent vaccine or 13-valent vaccine. Much of the SR that we saw occurring came from the six extra serotypes that were not in the 7-valent vaccine but are now in the higher valency vaccines, so we suspect that these higher valency vaccines would take care of a good deal of the SR that we saw.
What we don’t know is: Are there other serotypes that are waiting in the wing that will cause SR in the same way? It would be important to follow countries as they introduce the higher valency vaccines and do a similar type of analysis several years out after the vaccine has been introduced. The problem is: we showed that you really need to go out to five or more years after vaccine introduction to see what exactly is going to happen to this dynamic process overtime. You can’t do the studies yet because the vaccines have just been introduced in the last couple of years so we need to wait, particularly for developing countries because they are just introducing.
The second major limitation is that the studies we looked at in our analysis were mostly from developed countries (North America, Europe, and Australia). There were a few indigenous populations in our data set – the Navajo and the Australian indigenous – which might be more like the low-income country populations, but not exactly the same. So, it is possible that in Africa and South Asia, where the epidemiology of PD is different, the results of the vaccine could potentially be different as well. It would be important to do a similar type of analysis in those countries. I think this type of analysis needs to be repeated in a few years, and it needs to be repeated in low-income settings that have the higher valency vaccines.
Do you have any other thoughts that you want to convey that we have not covered?
I think there is one thing, not based on the data itself, but just a comment. One of the things this study showed is the power of collaboration. Any single site would not be able to do this type of analysis, but they showed data from their own site and they made conclusions for their own site. But in order to understand an epidemiologic concept in a broader way that may vary site by site, you really need to have data from different places. The willingness of investigators to collaborate on this project, to lend their data to be looked at, to work together to make some sort of conclusion that was bigger than their own data was a really valuable exercise. This type of collaboration between sites is becoming more and more important, and we’re seeing more of it happening in the field of epidemiology. So, the example of doing a multi-site analysis, where sites give their data and they participate in the interpretation, was a valuable lesson and exercise.