General ID: December 2018

Seven days of antibiotics is as good as fourteen days for gram-negative bacteremia.  For me, this was the big RTC that came out this month, maybe the biggest of 2018 (POET and MERINO, I still love you).  Why?  Because it’s a highly generalizable study about a common infection, builds upon an existing body of retrospective literature reaching the same conclusions, and hence has tremendous potential to actually change practice.

The authors performed an unblinded non-inferiority trial randomizing hospitalized adults with aerobic gram-negative bacteremia who had become afebrile and hemodynamically stable by day seven of appropriate antibiotic therapy 1:1 to receive either no additional antibiotic therapy or another seven days of antibiotics.  They included both community and hospital-acquired bacteremias, and enrolled bacteremias from intraabdominal infections, infections of the urinary tract, respiratory tract, skin or soft tissue, or central venous catheters, as well as infections of an unknown source.  Bacteremias due to endocarditis and other endovascular infections, osteomyelitis, uncontrolled sources (e.g. undrained intraabdominal abscess, central lines that were not removed), CNS infections, and empyema were excluded; also excluded were polymicrobial bacteremias, bacteremia due to Salmonella or Brucella, and patients with HIV infection, unengrafted or recent stem cell transplantation, or neutropenia within 48hr of randomization.  The primary outcome was a 90-day composite of all-cause mortality, clinical failure (relapsed bacteremia or infectious complications), re-admission, and extended hospitalization (>14 days).  The authors chose a 10% threshold for non-inferiority.

A total 604 patients enrolled in the trial, which took place between 2013 and 2017 at several Israeli medical centers.  Baseline patient characteristics were comparable; most patients had Enterobacteriaceae infections (90%), most bacteremias were due to urinary tract infection (68%), and the route and class of the initial antibiotics given were similar.  The primary outcome occurred in 46% of the patients who received 7 days of antibiotics versus 48% of those who received 14 days of antibiotics (p>0.05), with a confidence interval of -11% to 5%, meeting the threshold for non-inferiority.  No differences were observed for any of the components of the composite outcome – and specifically, 90-day mortality was 12% in the 7-day group versus 11% in the 14-day group.  Analysis of the per-protocol population (n=556) produced similar results.  The only prespecified subgroup populations for whom 7 days did not meet the non-inferiority threshold were those limited by small samples sizes: patients who received inappropriate antibiotic therapy and patients with MDR pathogens.

Total duration of hospitalization was no different between the 7-day and 14-day antibiotic recipients; however, patients who received 7-day therapy recovered to their baseline functional status more quickly (median 2 vs 3 weeks; p=0.01), were less likely to be bedridden or dependent on others for ADLs at 30 days (51% vs 57%; p=0.03), and received fewer total days of antibiotics over the 90-day period (median 10 vs 16 days; p<0.001).

Bottom line?  Clinical outcomes are non-inferior (and in some regards may be superior) with 7 versus 14 days of antibiotic therapy for adults with gram-negative bacteremia who have reached clinical stability by day 7.  These data are widely generalizable, except for patients who have concomitant infections requiring longer durations of therapy (e.g. osteomyelitis, endovascular infections), infections with uncontrolled sources (e.g. undrained intraabdominal abscess, a central venous catheter left in place), infections due to special pathogens (Salmonella and Brucella), or in patients with specific major immune compromising conditions (HIV, current neutropenia, or recent organ transplantation).  30535100


Prescription of aspirin (and a macrolide) reduces 30-day mortality in severe community acquired pneumonia.  Infections are inflammatory and inflammation is pro-thrombotic, so it’s no surprise that cardiovascular events occur with and shortly after severe infections.  The idea that immunomodulatory agents (I’m including macrolides in this category, as they downregulate TNFA and Th2 cytokine activation) might be of benefit beyond antibiotic therapy in pneumonia has been around for a while (disclosure: Dr. Musher is one of my mentors).  While statins and steroids have gotten the bulk of the attention in this regard, we have both retrospective and randomized clinical trial data to suggest that aspirin therapy reduces cardiovascular mortality after pneumonia.  This observational study expands on the present body of literature by examining the effects of using aspirin, a macrolide, both, or neither on mortality in patients hospitalized with severe community-acquired pneumonia.

The authors collected the data of adult patients hospitalized at one five hospitals in the US, China, Japan, and Italy between 2011 and 2015 who had clinical symptoms of pneumonia and met the 2008 IDSA CAP guidelines criteria for severe disease.  The methods section includes the nebulous statement that “only patients treated with aspirin before admission were included in the study, while patients who received aspirin during hospitalization were not recruited.”  I take this to mean that in order to qualify as being in the aspirin or aspirin-macrolide groups you had to have been on aspirin prior to your hospitalization, and in order to be in the macrolide or neither-drug groups you had to have not received any aspirin at all, but I’m not 100% on that.  Also, note that patients had to be on at least 100mg of aspirin a day to qualify – so you have to imagine that most of the patients were on full-dose aspirin, seeing as the two dosages we use in practice are 81mg and 325mg.

The study included 1295 patients with severe CAP.  Of these, 18% received aspirin, 23% received a macrolide, 11% received both, and 48% received neither. The baseline characteristics between these groups were similar other than a higher incidence of multicomorbidity in the aspirin-only group and a higher incidence of MDR infections in the neither-drug group.  More cardiovascular events were recorded in the aspirin-only group versus the others (no surprise – presumably they were on aspirin for cardiac risk mitigation and hence at a higher baseline risk).  Mortality rates were 28% (neither drug), 24% (macrolide only), 21% (aspirin only), and 16% (aspirin plus macrolide).  When comparing these groups directly, receipt of both aspirin and a macrolide was associated with lower mortality versus neither drug (p=0.001) or just a macrolide (p=0.04) but not versus just aspirin (p=0.12).  The statistical significance of each of these associations remained unchanged after propensity score matching and regression analysis.  To me, this indicates that the aspirin, rather the macrolide, provided the bulk of the protective effect.

Given that aspirin is a relatively benign intervention and that most of the excess cardiovascular risk is centered in the first 90 days after the episode of pneumonia, I think the balance of the evidence favors prescribing a 30-90 course of full dose aspirin for all adults with severe pneumonia. What do you think? 30509940


How long a delay to appropriate antimicrobial therapy does a patient with E.coli or K.pneumoniae bacteremia need to have before they’re at increased risk of a poor outcome? Not long.  The authors performed a systematic review of studies examining the association between time to appropriate antibiotic therapy and mortality in hospitalized patients with E.coli or K. pneumoniae infections.  They included studies who reported patients stratified by appropriate therapy delays of either 24hr or longer or after the results of culture and susceptibilities had been reported.  Twenty studies met the inclusion criteria, 19 of which looked at bacteremias.

Five of the twenty studies showed statistically significantly higher mortality in patients who had delayed versus non-delayed appropriate therapy, with a trend in that direction observed in most of the remaining studies.  Overall, delays in appropriate antimicrobial therapy of greater than 24hr or beyond the availability of culture and susceptibilities were associated with increased mortality (OR 1.6; 95% CI 1.3-2.1).  When restricting the analysis to the five papers that looked at delayed appropriate therapy >24 hours, there was a nonsignificant trend toward increased mortality (OR 1.2; 95% CI 0.9-1.8); when restricting the analysis to the four papers looking at delayed appropriate therapy >48 hours, a similar nonsignificant trend was found (OR 1.3; 95% CI 0.8-2.1). 

This seems like a case of inadequate sample size rather than lack of a genuine association, as nearly all of the studies are pointing the same way.  The bottom line here is that even relatively short delays in appropriate therapy can confer significant excess risk of death, so the “start broad, then narrow” approach to empiric antimicrobial therapy is almost certainly life-saving for severe infections versus starting conservatively and then escalating therapy once an antibiotic-resistant organism has been identified. 30518337

Infective endocarditis (IE) associated with drug use has exploded over the past decade.  The United States’ opiate epidemic shouldn’t be news to anyone working in healthcare today, but the degree of increased prevalence of this disease is staggering.  The authors reviewed data from adult hospital admissions in the North Caroline Hospital Discharge Database between 2007 and 2017.  This dataset includes demographic, diagnostic, procedural, and billing data from all nonfederal acute care hospitals in the state.  Cases of IE were identified by ICD-9 codes, and IE cases associated with drug use were identified by the accompanying presence of ICD-9 codes for substance use disorders; the authors further stratified the data by the type of substances used (e.g. opiates, cocaine, amphetamines, etc).

During the 10-year period studied, 22,825 cases of IE occurred, of which the authors identified 2,602 cases (11%) as associated with drug use.  Between 2007 and 2017, the rate of IE admissions increased 36%, from 28 to 38 cases per 100,000 residents, driven by a 1200% increase in IE associated with drug use (from 0.9 to 11 cases per 100,000 residents).  Hospitalizations resulting in valve surgery doubled (1.6 to 3.3 per 100,000 residents), and by the end of the study IE associated with drug use accounted for 42% of all hospitalizations involving valve surgery.

Who is the patient with drug-associated IE?  In this study, the median age was 35, men and women were equally represented, nearly 70% were either uninsured or receiving Medicaid, 62% used opiates (vs 22% using cocaine, 11% amphetamines, and 19% other drugs), and 60% had HCV coinfection.  Compared to other IE cases, IE associated with drug use resulted in longer hospital stays (among all cases, median 11 vs 7 days; p < 0.001; among those receiving surgery, 27 vs 17 days, p<0.001), higher costs ($60k vs 35k; p<0.001), and discharge against medical advice in 13% of cases.  The total medical costs of drug use-associated IE exceeded 78 million dollars.

To me, this study compellingly demonstrates the need for evidence-based substance use disorder treatment and other interventions aimed at harm reduction.  Even if you’re only willing to be moved by the cold-hearted economic argument – drug-associated IE is costing North Carolina 8 million dollars alone, and probably much more once you factor in ED and outpatient costs and economic losses due to unemployment and criminal activity.  The state could easily recoup those costs by providing safe injection centers, subsidizing suboxone treatment, and implementing other similar interventions, all of which would be lifesaving at the same time.  It just makes sense.  30508432

Maybe oral beta-lactams aren’t so terrible for pyelonephritis?  Beta-lactams have traditionally been disfavored for the treatment of urinary tract infections due to a history of increased failures versus alternative agents in trials of cystitis.  Which is all well and good, except that antibiotic resistance is on the march, and with the rise of quinolone and trimethoprim-sulfa nonsusceptibility in community-acquired uropathogens, there really isn’t a reliable drug for the outpatient treatment of pyelonephritis in many parts of the world (remember that nitrofurantoin and fosfomycin, which we ought to have been using more of for UTI over the past 30 years, achieve poor levels in renal parenchyma and are really only intended for cystitis).  So now we find ourselves taking a look at alternative agents.

The authors performed a systematic literature review of RTCs examining the efficacy of oral antibiotics for acute pyelonephritis.  They located a total 5 studies involving the antibiotics cefaclor, ciprofloxacin, gatifloxacin, levofloxacin, lomefloxacin, loracarbef, norfloxacin, rufloxacin, and trimethoprim-sulfa.  All studies had clinical criteria for pyelonephritis as well as laboratory criteria for bacteriuria and (in all but one case) pyuria.  Sample sizes were moderate (84 – 378 participants), and women made up the large proportion of participants in every case.

The studies most commonly assessed clinical cure at 5-9 days and 4-6 weeks after antibiotic treatment.  They found that cefaclor, ciprofloxacin, levofloxacin, loracarbef, and norfloxacin all produced high rates of clinical cure at both 5-9 days (84-95%) and 4-6 weeks (83-95%), though microbiologic cure rates were lower (50-80%) at all time points for the beta-lactams.  Trimethoprim-sulfa produced clinical cure at lower rates for both the early (85%) and later (72-78%) time points.  Both ciprofloxacin and trimethoprim-sulfa were associated with high rates of adverse events (24% and 33%), mostly GI and headache.

I’d be hesitant to pronounce the equivalence in efficacy of beta-lactams as a class to ciprofloxacin and trimethoprim-sulfa on the basis of these data.  Cefaclor is primarily (60-85%) excreted in urine, which is also the case for cephalexin and cefadroxil (>90% each) but not the case for other drugs, like cefdinir (12-18%), cefpodoxime (29-33%), or even amox-clav (60% for amoxicillin but only 25-40% for clavulanate).  Still, there is RTC data that step-down therapy with oral cephalosporins works just as well as all-IV therapy (22951426 - though note that everybody got 3 days of 2g IV ceftriaxone, and recent data says 3-5 days is probably enough for pyelo), so perhaps we ought to stop turning up our noses at oral beta-lactams entirely. 30191339