AMR deaths in Europe (part 2)

“33000 people die every year due to infections with antibiotic-resistant bacteria” this is what ECDC released on Nov 6, 2018, on their website. “Superbugs kill 33,000 in Europe every year” said CNN and the same wording was used (in Dutch) by our Telegraaf. Naturally, the headings were based on the ECDC study published that day in Lancet ID, which happened to be the most downloaded paper ever of the journal. But was this really what was published? Valentijn Schweitzer and I got lost in translation when trying to answer that question.

What is your interpretation of this conclusion and the newspaper headings?

  1. Many patients develop infections caused by antibiotic-resistant bacteria and 33,000 die from it, only because these bacteria were resistant to some antibiotics. They would have survived if the bacteria had been susceptible. Let’s call this AMR mortality.


  1. Many patients develop infections caused by antibiotic-resistant bacteria and 33,000 die because of the infection, but not necessarily because the bacteria were resistant to some antibiotics. Many would have died even if the bacteria had been susceptible. Let’s call this the total AMR infection mortality.

The total AMR infection mortality is the sum of AMR mortality and baseline infection mortality. The latter being the mortality rate of that infection when caused by a susceptible bug. For bloodstream infections (BSI) this is mortality due to non-AMR BSI (baseline BSI mortality).

Total AMR BSI mortality = AMR mortality + baseline BSI mortality

So how was this calculated in the study? Let’s use MRSA BSI as an example. Based on the systematic review performed (in the supplement, page 172) they set total MRSA BSI mortality at 17.9% and baseline BSI (non-MRSA/MSSA) mortality at 13.7% (range 7.1%-20.3%). Thus, MRSA mortality must be 4.2% (17.9% minus 13.7%, see formula). So, what does this mean for the Netherlands?  With 42 MRSA BSI/year and a total MRSA BSI mortality of 17.9%, 7 of these patients die each year. Yet, 5 of them (the 13.7%) would also have died from MSSA BSI, so the proportion of deaths truly (and exclusively) attributable to methicillin-resistance is 23% of 7, is 2. In other words, for MRSA BSI in the Netherlands the interpretation is that a total of 7 patients die with MRSA BSI, and that 2 of these die because of methicillin-resistance. For all pathogens it was estimated that 157 patients died after an AMR BSI in the Netherlands, and that 50 (32%) died because of the resistance. As mentioned in 2 previous blogs, we strongly feel that the attributable mortality figures for the super-killer-bugs in the Netherlands (i.e., colistin-resistant E. coli and K. pneumoniae and 3rd generation cephalosporin resistant E. coli and K. pneumoniae) were hugely overestimated, reducing the number of deaths because of antibiotic resistance in our country to the numbers expected (close to zero).

In this Table we have listed the figures for BSI, for the Netherlands.

blog ecdc_table_page_1

In the Netherlands, 76% (157/206) of the total AMR infection deaths were due to BSI. If we assume that the same proportion of AMR infections are due to BSI in Europe,  25.164 patients would die with AMR BSI infections, of whom 8,053 (again 32%) would die because of the resistance, which includes the deaths following BSI caused by colistin-resistant but carbapenem susceptible bacteria. This paints quite a different picture than the news headlines. Why is this distinction relevant? First of all, research is intended to deliver a correct answer to a question and that answer should be communicated correctly. Dying with or dying from are different things. Preventing selection and spread of resistance, without preventing infections, will at most prevent the “AMR mortality” fraction of the “total AMR infection mortality”. The correct interpretation of the study data was not immediately clear to us (and the press coverage didn’t help in getting us there). As we think all others knew exactly what was meant, we sincerely apologize for the confusion created by our previous blogs.



2 thoughts on “AMR deaths in Europe (part 2)

  1. I, for one, certainly interpreted the figure of 33,000 as AMR mortality, as you did initially. Thanks for really digging deep into this issue and getting my interpretation straight. I would like to add one aspect though: to what extent do resistant bacterial clones drive an increase in infections? To cite from the title of an article authored by one of the authors of this post: “Secular trends in nosocomial bloodstream infections: antibiotic-resistant bacteria increase the total burden of infection” (Clin Infect Dis. 2013 Mar;56(6):798-805). This seems to imply a causal effect of antibiotic resistant bacteria on the incidence of incidence. If this were the case, then the counterfactual for an infection with resistant bacteria would not be an infection with their susceptible counterpart, but no infection. That would make the approach taken by the ECDC (as explained in this post) or the article in ICHE (discussed in your previous blog post on this issue) more defendable (although, to what extent do these mortality figures truly represent infection-related mortality?).
    I am pretty skeptical about this, but increasing incidences seems to be a very influential theory. For example, in a recent paper estimating worldwide incidences of Gram-negative infections (Lancet Glob Health. 2018 Sep;6(9):e969-e979), infections with resistant bacteria would only start to replace infections with susceptible bacteria if the resistance level surpassed 30%. If I understand correctly: any infection with resistant Gram-negative infection occurring in the Netherlands (and many other countries!) would not have occurred in its susceptible form, had antibiotic resistance not existed.
    Maybe someone should be drawing a directed acyclic graph (DAG) to get to grips with the issue of increasing infection incidences due to antibiotic resistance.


  2. Thabk you Wouter.
    And yes, the associations between resistance and the occurrence of infections (replacement of susceptible infections or adding up to increase the total number of infections) is becoming very relevant. I think there is enough evidence to conclude that a resistant E. coli is not more virulent than its susceptible counterpart. It is also likely that in a setting where transmission control of resistant bacteria fails, other infection control measures ars also less effective than in settings with succesful transmission control, and this would lead to the association of more resistance = more infections. A clear effect of resistance on the total number of infections exists when infections can no longer be prevented with antibiotics (because of resistance to these antibiotics). If that happens, we usually change prophylactic protocols. Would be interesting to look for data that could quantify this fraction.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s