July 2020: General ID

Add brain abscesses to your list of infections that can be adequately treated with oral antibiotics. One of the oft-stated hesitancies of using oral antibiotics in bacterial *meningitis* is that that is an often rapidly fatal condition, and the minutes to hours of delay in achieving therapeutic CNS antimicrobial levels with gut absorption versus IV administration could be the difference between a good outcome and permanent neurological sequelae or death.  Fair enough.  Bacterial CNS abscesses, however, tend to be more indolent processes.  So as long as we use bioavailable antibiotics with good CNS concentrations, and especially if PO is given as a switch after a few doses of IV therapy, why shouldn’t they work well?

The authors reviewed all adult patients admitted to Nantes University Hospital with a pyogenic brain abscess between 2003 and 2016.  They included only cases with CNS imaging and either a clinical diagnosis treated with antibiotics or a positive culture; cases due to mycobacteria, fungi, and parasites were excluded.  They sorted these patients into those given all-IV therapy, or switch to an all-oral regimen.  Their primary outcome was the Glasgow Outcome Scale 3 months after diagnosis, as assessed by the study authors.  If you also hadn’t heard of the GOS, it’s a rather subjective five point scale that ranges from 1 (vegetative status) to 5 (“good recovery”).

They identified a total 109 patients, of whom 78 had a GOS>4 (no more than moderate disability), 30 had a GOS <3, and the rest died.  Of these patients, 28% had post-surgical infections, 19% an underlying malignancy, and 17% some other preexisting neurological condition.  Half had had an initial Glasgow Coma Score <15. A microbiologic etiology was identified in >95% of those with a relatively good 3mo outcome and 80% of those with a poor outcome; of these, the most common organisms were strep (30%, with S.anginosus being particularly common), followed by anaerobes (27%), MSSA (14%), and GNRs (9%).  Total duration of antimicrobial therapy was a mean 65 days (range 45-95); 44% switched to oral therapy after a mean 19 days (range 12-28).  In multivariate analysis, abnormal initial GCS (aOR 4.4 with 95% CI 1.6-14) and medical comorbidities (aOR 4.8 with 95% CI 1.4-19.4) were associated with a poor outcome, while switch to oral therapy was associated with a favorable outcome (aOR 0.2 for poor outcome with 95% CI 0-0.6).

In a few ways, this paper isn’t the slam dunk I want it to be.  The switch to oral antibiotics was, on average, after nearly 3 weeks of therapy – how much does that really mean for a disease state in which the standard duration of therapy can be as short as 4 weeks?  Multivariable analysis can also only do so much to account for the differences in who was likely to be switched to PO (my guess is clinical improvement at followup = switch to PO, and lack of improvement = continued IV).  Case-control matching or propensity weighting would have been welcome, though the small sample sizes would have made the data difficult to interpret. 32699907

Obtaining a FDG-PET is associated with changes in clinical management and better rates of survival in S.aureus bacteremia (SAB).  I’ve previously written about why I think the evidence for routine PET is at least as strong or stronger than the evidence for routine TEE. It’s surprising to me that current guidelines haven’t recommended routine PET in SAB despite the recommendation for routine echocardiography with a stated preference for TEEs.  This paper adds to the mounting evidence that FDG-PET is useful and potentially life-saving in SAB.

First, a word on the methodology.  The authors compared adults with SAB received FDG-PET, either because they’d enrolled in another clinical study or because the treating physicians thought it was a good idea, to adults with SAB seen over largely overlapping 3-year periods who hadn’t received FDG-PET, matched 1:1 for survival time from blood culture positivity to PET, methicillin resistance, Charlson comorbidity score, and age.  So don’t let the use of the world “prospective” in the title distract you: this is a matched cohort study and not a randomized trial.  In any case, they stated they encouraged basing treatment duration decisions on the FDG-PET (>2 weeks if residual foci of infection were found, and 2 weeks if not) and their primary outcome of interest was 90-day all-cause mortality.  Multiple episodes of SAB per patient were included in the study so long as they occurred at least 6mo apart.

A total 149 patients underwent PET for a total 151 episodes of SAB; of these, 25 episodes had PET at the treating physician’s discretion, and the rest got it as part of the clinical trial.  The FDG-PET studies were obtained a media 11 (range 8-13) days after the first positive blood culture, and 71% identified at least one focus of infection, with 67% having a NEW infectious focus seen on PET.  The most commonly identified infectious foci were bone and joint (including vertebral osteomyelitis; 38% of all PET recipients, in half of whom such foci had not been clinically suspected), followed by skin/soft tissue and lungs (21%, again the majority of which had not been clinically suspected).  About a fifth of patients had a new source control procedure after getting the FDG-PET.  The primary outcome of 90-day mortality was lower in the PET recipients than matched controls (14% vs 29% with p=0.002) despite more prolonged bacteremia in the PET group (29% vs 17% having >3 days of bacteremia).  In multivariable analysis, receipt of PET continued to be associated with lower mortality (OR 0.4 with 95% CI 0.2-0.8 and p=0.016); other predictive factors included ago, Charlson and Pitt scores, prolonged bacteremia, and fever despite appropriate antibiotic therapy.  Of note, receipt of a TEE was also predictive of survival and to a similar degree (OR 0.4 with 95% 0.2-0.9 and p=0.018). So, more PETs in SAB = more infectious foci found, more infectious foci dealt with, and lower rates of all-cause mortality at 90 days.  What’s not to love? 32639560

Persistently positive cultures with gram-negative bacteria indicate a high risk of mortality.  So says a retrospective study of 1702 patients at Duke.  Positive followup blood cultures was associated with 12% infection-attributable mortality versus 7% for those whose cultures were negative, with a HR of 2.1 (95% CI 1.6-2.8) in propensity score-weighted analysis. No surprise here, as gram-negative bacteremias usually clear rapidly even in the absence of good source control, an indeed those with persistently positive cultures were likely to have been on ineffective antibiotic therapy, to have endovascular devises, or to have ESRD.  The authors also found that obtaining follow up blood cultures was associated with lower mortality, but I don’t think we can assume a causal relationship.  This association probably reflects immortal time bias because followup blood cultures were defined as cultures obtained up to 7 days later (i.e. you can’t get repeat blood cultures and be classified in the got-repeat-cultures group on day 7 if you died on day 4).  The authors stated they did an analysis excluding patients who died within 48 hours of the initial culture in the no-followup-cultures group and still found an association between repeat cultures and lower mortality, but they don’t show the data, and it seems to me that an analysis excluding anyone who died before hospital day 7 in both groups would be a more appropriate accounting for that bias.  But I’m not a biostatistician; feel free to yell at me on twitter if I’m wrong. 32114010

Decline in appropriate surgical management for infective endocarditis: the unintended consequence of public reporting on valve surgery outcomes?  Public reporting for aortic valve surgery, started by the Society of Thoracic Surgery in 2013, was intended to facilitate quality improvement and transparency, but as far as I can tell no one has looked at whether this actually improved clinical outcomes for patients that needed aortic valve surgery, and specifically for those with infective endocarditis.  One of the experiences from internship that still haunts me is caring for a young man in our county hospital’s CCU with severe multivalvular destruction resulting from IE. He was perpetually caught in the limbo of being too high-risk for valve surgery, too hemodynamically compromised to stay off ionotropes and vasopressors long enough to transfer to the floor (much less go home), and too young and otherwise healthy for his family - or frankly, his doctors - to be willing to accept hospice care.  He and his family would have accepted an attempt for surgical cure, even with tremendous risk of intraoperative mortality, because it would have offered at least the possibility of survival, which did not exist with medical management. Instead he languished in a CCU bed for months, unhappy and mostly alone, bodily functions slowly being supplanted or replaced by machines, until he finally developed HAP and died.

The authors of this paper used inpatient data from a large publically available database (the National Inpatient Sample) obtained between 2010 and 2015.  They collected data on adults 18-65 admitted with native valve endocarditis (determined based on ICD9 codes) and stratified them into those who did vs did not also have an ICD9 code for injection drug use or hepatitis C (the authors referenced literature indicating this has passable PPV for IDU-associated IE).  Their primary outcome was whether the patients received valve surgery, again determined by ICD9 code; in-hospital mortality was secondary outcome.  The authors found that, across 31,319 hospitalizations, valve surgery rates declined by 4% per quarter after public reporting was initiated in Q1 2013, with no difference in rates of decline between IDU-associated and non-IDU-associated cases.  Female gender and low income were associated with not getting surgery, whereas having private insurance was associated with getting surgery.  In-hospital mortality declined by 2% per quarter after public reporting, again with no difference between IDU and non-IDU cases.  This is not surprising, because sometimes patients die intraoperatively when you perform major cardiac surgery.  It also does not mean that the overall quality of care or rates of survival for IDU-associated and non-IDU-associated IE have improved, because there was no way to look at long-term survival; in fact, as other literature has associated valve surgery for IE with a host of improved outcomes, limiting access to that care may have done great hard.  Here’s one of the few times it’s actually correct to end your conclusion saying further research is desperately needed. 31598642