My previous blog on “mcr-1 and the end of the world” evoked responses on the important effects of colistin resistance on patient outcome, referring to a new study in CID with the following abstract closure: “Importantly, mortality was increased in patients with colistin-resistant isolates.” The wording is correct, but I’m afraid that it will be interpreted incorrectly.
This is the next great study in 2017: an observational study describing 246 patients with carbapenem-resistant K. pneumoniae (CRKp) isolates of which 31 (13%) were also resistant to colistin (no mcr-1 or mcr-2). In Cox regression and Kaplan-Meier analysis colistin-resistance was associated with hospital-mortality (adjusted odds ratio 3.48 (95% CI 1.73-6.57, p<0.001) and proportions surviving 30 days after culture results of app. 25% and app. 60%. Colistin-resistance clearly is a marker of a poor prognosis. Yet, is it also responsible for that outcome?
Quantifying the attributable effect of resistance on patient outcome is difficult, as one patient may resemble another one as an apple and a pear. Let’s look at the fruit collection in this study: 246 patients had a CRKp isolate, but only 111 (45%) met criteria for infection. So, 135 patients did NOT have an infection, and it seems unlikely that carriage attributes to mortality. So, the large outcome difference – if attributable to colistin resistance – must have come from the 111 infected ones. And I was, therefore, interested in the breakdowns of colistin susceptibility and mortality among the infected/non-infected patients. The lead author David van Duin kindly provided these upon request:
Infected colistin-R: 6/13 (46%) all-cause hospital mortality
Infected colistin-S: 40/98 (41%) all-cause hospital mortality
Non-infected colistin-R: 8/18 (44%) all-cause hospital mortality
Non-infected colistin-S: 22/117 (19%) all-cause hospital mortality
These data show a remarkably lower mortality among the non-infected with colistin-susceptible CRKp carriage. Hard to imagine that colistin-resistance among 6 infected patients created the observed 35% mortality difference. So, colistin resistance clearly is a prognostic marker, but I’m not sure (yet) that it causes mortality. For clarity, the authors addressed this aspect (though briefly) in the discussion, but I fear that some readers do not go that far …..
Also interesting in this study: the (lack of) reliability of E-tests for categorizing susceptibility in the breakpoint area!