View all News & Blog posts »
View all projects »
View all publications »
Interactive data visualizations of antibiotic use and resistance in North America and Europe
Why strep matters
Streptococcus pneumoniae is a microbe that frequently inhabits our nostrils without causing health problems. However, when it invades normally sterile parts of the body, it can cause invasive pneumococcal disease (IPD) that can be deadly. IPD carries a high mortality risk of 10%-20% even in developed countries and claims between 850,000 – 1 million children worldwide, making it the number one vaccine-preventable disease in the world. In the United States, Streptococcus pneumoniae is the main pathogen behind community-acquired pneumonia (CAP) and a frequent cause of bacterial meningitis and middle ear infections. The bacteria strikes primarily pediatric populations and, to a lesser extent, elderly patients. Whereas life-threatening conditions were once resolved with penicillin, treatment has become harder and costlier due to the rapid proliferation of drug-resistant streptococcus pneumonia (DRSP) since the 1970s.
Fortunately, pneumococcal disease is vaccine-preventable - in 2000, the heptavalent protein-conjugate pneumococcal vaccine (PCV-7) was approved by the FDA. Marketed as Prevnar, PCV-7 has been universally administered to children under the age of 2, only recently succeeded by the updated PCV-13 vaccine in February 2010.. While PCV-7 decreased the incidence of disease, particularly in its first years, the vaccine’s long-term effects on patterns of pneumococcal drug resistance are less clear.
Through our ResistanceMap animations, Extending the Cure has already visualized the regional variation and national spread of pneumococcal resistance against two of the most common therapies: beta-lactams (e.g. penicillin) and macrolides (e.g. erythromycin). However, whereas the maps draw a compelling picture of regional differences, the objective of this blog post is to discuss the relationship between pneumococcal resistance and vaccine introduction, while also placing ResistanceMap data in the context of additional surveillance data sources.
Resistance Data in Context
The ResistanceMap video for erythromycin, a first-line pneumococcal therapy since the 1990s, shows an initial drop in outpatient resistance levels from 36.7% to 31.6% in the first three years, followed by a period of increasing resistance where national levels surpass the 40% mark by 2009. Regionally, levels are higher in the north central and southeastern parts of the country.
In the same series, penicillin, the traditional treatment against less invasive forms of S. pneumoniae, shows lower levels of resistance and an even greater initial decrease from 23.8% to 13.6% between 2000 and 2003. Afterwards, resistance levels gradually increase to 16% in 2007. Rates are markedly higher in South-Atlantic and (initially) East South Central states.
To explore this resistance data more deeply, the graph below plots national trends in non-susceptibility to these two drugs, where non-susceptibility is defined as the ratio of resistant and intermediate-resistant results to total sample size (rather than strictly resistant isolates). Our non-susceptibility data for the 2000-2009 period has the same source as ResistanceMap - a nationally and regionally representative repository of susceptibility results called The Surveillance Network-USA (TSN). We compare this data with samples from two other nationally-representative sources: the US SENTRY Surveillance Program (totaling 12,049 isolates from hospitals across the country), and the CDC-Active Bacterial Core Surveillance reports (totaling 34,681 isolates from cases of confirmed pneumococcal infections in ten states). Note that below we pool inpatient and outpatient samples, and explore non-susceptibility trends, instead of using strictly resistant samples as in ResistanceMap. Including intermediate resistance is important in the case of pneumococcus, as more than one source report penicillin intermediate resistance increased in the period, while strictly defined resistance remained stable.
All three sources use the universally accepted Clinical and Laboratory Standards Institute (CLSI) susceptibility criteria for the given year. In 2008 the CLSI raised breakpoints of the minimum concentrations of penicillin required to stop a non-meningeal infection (the vast majority of cases), effectively causing fewer isolates to be classified as resistant.
Based on all three sources, pneumococcal non-susceptibility was on the rise four years after vaccine introduction. Erythromycin non-susceptibility (blue) saw an initial decrease that lasted until 2002-2004, and was followed by a steady increase. By 2009, rates are between 16% (CDC and ResistanceMap data) and 52% (SENTRY) higher than in 2000. Penicillin non-susceptibility (yellow line) increased after 2004. Although growing at a slower pace, it remained higher than erythromycin by 5 to 15 percentage points. Overall, until 2008 when guidelines were revised, the direction of the trend was the same across drugs for each source: resistance increased. ResistanceMap rates (solid line) were significantly higher than SENTRY (dashed line) and CDC (dashed-dotted line) numbers. Although it is impossible to point to the cause of this variation without knowing more about the sampled populations, it is worth noting that ResistanceMap data has a sample size two to seven times higher than the other two sources.
Vaccination and Disease Incidence
While our focus falls on resistance, it is important to note that the effectiveness of a vaccine should be judged based on the number of prevented instances of disease. The burden of pneumococcal resistance is lowered if fewer people are ultimately infected with the bacteria. The interactive graph below uses data from the CDC-ABCs reports to visualize penicillin and erythromycin non-susceptibility against the estimated number of IPD cases. The motion chart shows non-susceptibility and number of cases fall steadily in the early 2000s in response to the vaccine. However, in 2002-2003 resistance rates of erythromycin and penicillin begin to rise, and in 2004, disease incidence follows suit. The estimated number of invasive infections, which fell from 58,000 to 38,000, climbs up to 44,000 in 2008. The pace at which infection rates are making a comeback is slower when population trends are adjusted for (see Estimated number of cases/100,000 people). However, the numbers still imply there is a reversal of the initial impact of the vaccine in the mid-2000s, right around the time studies report changes in the prevalence of different pneumococcal strains.
Source of data: CDC-ABCs Why Are Disease Incidence And Resistance Rising?
How does the observed trend of increasing resistance and disease incidence relate to vaccination? The explanation advanced by the literature is that the PCV-7 vaccine altered the distribution of pneumococcal serotypes and led to the increased prevalence of microbes that carry higher resistance.
At a microbiological level, a pneumococcal serotype includes all strains that are grouped based on similarities of their polysaccharide coating. That protective coating is the primary target of the body’s immune response, which aims to destroy the polysaccharide capsule as soon as it detects an invasion. Vaccines work by teaching the body to produce an arsenal of antibodies that can fight off such an invasion. The problem is that they are only active against serotypes used to prepare them. PCV-7 includes the top 7 serotypes, which in 2000 accounted for over 80% of invasive pneumococcal disease and 89% of penicillin resistance. However, there are over 90 known serotypes of the bacteria worldwide.
Theoretically, while not as effective as an ideal-case scenario where everyone is vaccinated against all serotypes in existence, the employed strategy was expected to deliver a substantial hit to DRSP. First, vaccinated children would be less likely to get colonized or infected. By extension, there would be fewer opportunities for resistance genes to spread because the most resistant strains are targeted by the vaccine. Second, the decreased incidence of disease should decrease antibiotic consumption, thereby decreasing selection pressure on the non-vaccine strains.
There was indeed a decrease in the prevalence of vaccine serotypes following vaccine introduction: this review summarizes some of the evidence that during the first four years of vaccination there were significantly fewer cases of disease and decreasing rates of drug resistance (as also observed in our data). However, the vaccine did not cover the entire population, or the full array of serotypes. Predictably, most of the decrease in disease and resistance was observed among covered serotypes, while multiple studies document a parallel increase in the global prevalence of the non-vaccine serotypes as these microbes fill a newly freed ecological niche. As serotype replacement took place in the community, there were growing numbers of non-vaccine serotypes both colonizing and causing pneumococcal disease, which in turn increased their exposure to antibiotics. The rest is the old story about antibiotics, selection pressure and the resultant increase in resistant and (especially) intermediate-resistant strains of the non-vaccine serotypes.
This process of non-vaccine replacement and increase in overall resistance levels is exemplified by the evolution of the 19A serotype. The PCV-7 vaccine did not protect against this serotype, despite its clinical importance and high drug-resistance. A review of the literature shows that five years after the launch of the vaccine, 19A asserts itself as the primary cause of invasive pneumococcal disease and exhibits rapidly increasing non-susceptibility to both penicillin and erythromycin. There is evidence that this trend has been global, occurring even in places where PCV-7 was not introduced. While the causal link between the vaccine and this serotype is yet to be proven, it is clear that the rise of the 19A strain accounts for the resurgence of resistance observed in our data.
How did the non-vaccine organisms develop resistance? A recent article in Science maps the evolutionary history of S. pneumoniae going back more than 30 years. Researchers sequence the genome of 240 isolates of the PMEN 1 drug-resistant strain to establish how the bug has mutated in response to selection pressure from antibiotic use and vaccinations. They reveal multiple ways in which the strain has adapted to antibiotic overuse and vaccination interventions by recombining, substituting and exchanging genetic elements that confer resistance. Some of these modifications consisted of capsule switching – strains swapped a portion of their genome and evolved a different polysaccharide coating, effectively passing from one serotype to another. Even before the introduction of PCV-7 some strains had passed from the 23F strain covered by the vaccine to the notorious19A strain discussed above. The discovery is a true testament to the adaptive powers of this organism. It raises a new question: even though we now have a vaccine (PCV-13) to protect against the 19A strain, how long will it be before a new non-vaccine strain emerges to take the place of 19A?
Our brief look at data and literature indicates that pneumococcal vaccination has caused a shift in strains and a compensatory resurgence in drug resistance. The new generation of PCV-13 vaccines covers the strains that were most responsible for this compensatory increase such as 19A. Upcoming data will show how fast strains that remain uncovered are adapting to this enhanced intervention. Genomic research points to the inevitability of such bad news for serotype-based vaccines. However, the improved access to genomic information also provides encouraging news: the February 17th edition of Cell Host & Micobe reports on the development of a powerful non-capsular protein-based vaccine. The new approach promises not only much lower development and immunization costs, but also protection against all known serotypes.
 Another type of vaccine – the 23-valent polysaccharide vaccine (PPVSV) - has been available for the remaining at-risk population of adults and elderly. However, as it is not universally administered, it has failed to attract as much attention
 This introduces an interesting caveat in our penicillin trend lines after 2008 – only data from the CDC-ABCs take into account the differential diagnosis criteria. SENTRY applies the new breakpoints across the board. While not officially confirmed, it seems our data provider also adheres to the latter. The implication is that comparing penicillin non-susceptibility after 2008 should be done with a lot of caution.
 A serotype or serovar groups pneumococcal strains based on the body’s immune response
Image credit: Flikr: prashant_zi