Rounding errors

Tests got faster; deaths did not fall

When bacteria get into the blood, doctors start a broad-spectrum antibiotic at once and wait for the laboratory to say which drugs the organism yields to. Standard testing takes about two days, and a small industry has worked to shorten it. The industry has made real progress, and the race is not over. In FAST, a trial published in April, a rapid test reported susceptibility a median of 7.5 hours after the Gram stain, the first look at the bacteria under a microscope, against 44 hours for the standard method, among patients with a single organism on the test's panel.11Banerjee R, Komarow L, Li Y, et al. Fast Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia: The FAST Randomized Clinical Trial. JAMA. 2026;335(20):1762. doi:10.1001/jama.2026.5487. Published online April 18, 2026. Introduction ("5 prospective clinical trials ... decreased time to optimal therapy, but did not reduce mortality or shorten duration of hospitalization"), Table 1, Table 2, Figure 4 legend (patients with monomicrobial, on-panel infections) and Discussion ("sites enrolling patients with resistant organisms did not always have the most effective antibiotic therapies available"). https://jamanetwork.com/journals/jama/fullarticle/2847970 (T2 -- peer-reviewed randomised trial) A decade earlier, at the Mayo Clinic, a rapid DNA test named the bacterium 1.3 hours after the Gram stain, against 22.3.22Banerjee R, Teng CB, Cunningham SA, et al. Randomized Trial of Rapid Multiplex Polymerase Chain Reaction-Based Blood Culture Identification and Susceptibility Testing. Clin Infect Dis. 2015;61(7):1071-1080. doi:10.1093/cid/civ447. PMID 26197846. Abstract -- "Time from BCB Gram stain to microorganism identification was shorter in the intervention group (1.3 hours) vs control (22.3 hours)"; "piperacillin-tazobactam (control 56 hours, rmPCR 44 hours, rmPCR/AS 45 hours)"; "less treatment of contaminants (control 25%, rmPCR 11%, rmPCR/AS 8%)"; "de-escalation: rmPCR/AS 21 hours, control 34 hours, rmPCR 38 hours"; "escalation: rmPCR/AS 5 hours, control 24 hours, rmPCR 6 hours". https://pubmed.ncbi.nlm.nih.gov/26197846/ (T2 -- peer-reviewed randomised trial)

What speed has not done is save lives. FAST randomised 899 patients in Greece, India, Israel and Spain and found no difference in how they fared at 30 days; deaths numbered 100 and 99.11Banerjee R, Komarow L, Li Y, et al. Fast Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia: The FAST Randomized Clinical Trial. JAMA. 2026;335(20):1762. doi:10.1001/jama.2026.5487. Published online April 18, 2026. Introduction ("5 prospective clinical trials ... decreased time to optimal therapy, but did not reduce mortality or shorten duration of hospitalization"), Table 1, Table 2, Figure 4 legend (patients with monomicrobial, on-panel infections) and Discussion ("sites enrolling patients with resistant organisms did not always have the most effective antibiotic therapies available"). https://jamanetwork.com/journals/jama/fullarticle/2847970 (T2 -- peer-reviewed randomised trial)33ClinicalTrials.gov, NCT06174649, "Fast Antibiotic Susceptibility Testing for Gram Negative Bacteremia Trial", results first posted July 28, 2026 -- primary outcome categories and all-cause mortality by arm. https://clinicaltrials.gov/study/NCT06174649 (T1 -- trial registry, sponsor-submitted results) Its authors count five earlier trials of rapid susceptibility testing for the same infections. All five got patients onto the right drug sooner; none reduced deaths or shortened hospital stays.11Banerjee R, Komarow L, Li Y, et al. Fast Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia: The FAST Randomized Clinical Trial. JAMA. 2026;335(20):1762. doi:10.1001/jama.2026.5487. Published online April 18, 2026. Introduction ("5 prospective clinical trials ... decreased time to optimal therapy, but did not reduce mortality or shorten duration of hospitalization"), Table 1, Table 2, Figure 4 legend (patients with monomicrobial, on-panel infections) and Discussion ("sites enrolling patients with resistant organisms did not always have the most effective antibiotic therapies available"). https://jamanetwork.com/journals/jama/fullarticle/2847970 (T2 -- peer-reviewed randomised trial) Six trials, one pattern: a faster answer, and no measurable change in who survives.

Posting a faster result did not speed narrowing; a team did

The reason appears at the other end of the clock: not when the answer arrives, but when anyone acts on it.

The test is fast; the switch is slow

Top, a Mayo Clinic trial: hours after the Gram stain, the first microscope look at an infected blood sample. Open dot, lab result ready; filled dot, antibiotic changed; red bar, the gap between those two medians; grey dots, when guidelines expect a review. Bottom: what an hour costs a patient in septic shock who is not yet on a drug that works.

Trial rows are medians from one randomised trial (Banerjee 2015, 617 patients); each red bar is the gap between two medians. Guideline points: Surviving Sepsis Campaign 2021 (daily review for de-escalation) and CDC Core Elements 2014 (48-hour time-out), both timed from the start of antibiotics. Septic shock panel: Kumar et al, Crit Care Med 2006, 2,154 patients, clock starting at the onset of low blood pressure; three values stated in the text, the rest measured from its Figure 1. It is observational and applies to patients not yet on an effective drug; most FAST patients were already on one, and FAST found no survival gain from faster therapy.

The Mayo Clinic trial is the cleaner experiment: it held the test fixed and varied the hospital. It randomised patients to standard testing, to the rapid test with its result posted in the record, or to the same rapid test with a stewardship team (specialists who review antibiotic orders) acting on each result as it arrived.22Banerjee R, Teng CB, Cunningham SA, et al. Randomized Trial of Rapid Multiplex Polymerase Chain Reaction-Based Blood Culture Identification and Susceptibility Testing. Clin Infect Dis. 2015;61(7):1071-1080. doi:10.1093/cid/civ447. PMID 26197846. Abstract -- "Time from BCB Gram stain to microorganism identification was shorter in the intervention group (1.3 hours) vs control (22.3 hours)"; "piperacillin-tazobactam (control 56 hours, rmPCR 44 hours, rmPCR/AS 45 hours)"; "less treatment of contaminants (control 25%, rmPCR 11%, rmPCR/AS 8%)"; "de-escalation: rmPCR/AS 21 hours, control 34 hours, rmPCR 38 hours"; "escalation: rmPCR/AS 5 hours, control 24 hours, rmPCR 6 hours". https://pubmed.ncbi.nlm.nih.gov/26197846/ (T2 -- peer-reviewed randomised trial) When a result meant moving a patient to a stronger drug, both rapid arms moved quickly: five or six hours after the Gram stain, against 24 with standard testing. When it meant narrowing a drug that was broader than needed, they parted. With the team, the narrower drug came at 21 hours; without it, at 38, later than the 34 hours of standard testing. Posting the faster result did not speed narrowing at all. The test was identical in both rapid arms; only the team differed.

FAST points the same way. Patients with carbapenem-resistant bacteria, the likeliest to have started on a drug that did not work, reached an effective one a median of 9.5 hours after the Gram stain with rapid testing, against 28 hours with standard testing.11Banerjee R, Komarow L, Li Y, et al. Fast Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia: The FAST Randomized Clinical Trial. JAMA. 2026;335(20):1762. doi:10.1001/jama.2026.5487. Published online April 18, 2026. Introduction ("5 prospective clinical trials ... decreased time to optimal therapy, but did not reduce mortality or shorten duration of hospitalization"), Table 1, Table 2, Figure 4 legend (patients with monomicrobial, on-panel infections) and Discussion ("sites enrolling patients with resistant organisms did not always have the most effective antibiotic therapies available"). https://jamanetwork.com/journals/jama/fullarticle/2847970 (T2 -- peer-reviewed randomised trial) An urgent correction is made quickly. An optional one waits.

Results wait for the next review

The likeliest cause is ordinary. Guidelines expect clinicians to review antibiotics at set moments: a daily assessment for narrowing, in the Surviving Sepsis Campaign's guidance, and a "time out" 48 hours after antibiotics start, in the 2014 version of America's Centers for Disease Control and Prevention (CDC) hospital stewardship standards.44Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021. doi:10.1007/s00134-021-06506-y. Recommendation 29 -- "we suggest daily assessment for de-escalation of antimicrobials" (weak recommendation, very low quality of evidence); rationale -- "Antimicrobial de-escalation should ideally be done as soon as possible, and rapid diagnostic techniques may facilitate this". https://link.springer.com/article/10.1007/s00134-021-06506-y (T1 -- international clinical guideline)55Centers for Disease Control and Prevention. Core Elements of Hospital Antibiotic Stewardship Programs. 2014 -- "All clinicians should perform a review of antibiotics 48 hours after antibiotics are initiated". The 2019 update adds that "the optimal timing of antibiotic timeouts has not been established" and suggests daily reviews. https://www.cdc.gov/antibiotic-use/healthcare/pdfs/core-elements.pdf (T1 -- federal agency guidance) The same sepsis guidance says narrowing "should ideally be done as soon as possible, and rapid diagnostic techniques may facilitate this".44Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021. doi:10.1007/s00134-021-06506-y. Recommendation 29 -- "we suggest daily assessment for de-escalation of antimicrobials" (weak recommendation, very low quality of evidence); rationale -- "Antimicrobial de-escalation should ideally be done as soon as possible, and rapid diagnostic techniques may facilitate this". https://link.springer.com/article/10.1007/s00134-021-06506-y (T1 -- international clinical guideline) A result that lands between reviews can wait for the next. FAST wrote this into its protocol: a result that arrived "in the evening or on a weekend" could wait for the stewardship team until "the beginning of the next working day".66Study protocol for the FAST trial. Trials. 2025;26:500, published November 12, 2025 -- "If the AST result is ready when the AS provider is unavailable (e.g., in the evening or on a weekend), the AS team will review the MRs and make recommendations no later than the beginning of the next working day." https://link.springer.com/article/10.1186/s13063-025-09228-4 (T2 -- peer-reviewed protocol) Without a team acting in real time, the Mayo trial's narrowing came at 38 hours, well beyond the 24 hours a daily review implies.

In septic shock, survival fell 7.6 points for each early hour of delay

For a patient on the wrong drug, an hour has a price. In 2006 Anand Kumar and colleagues traced 2,154 patients in septic shock who had not received an effective antibiotic before their blood pressure collapsed. Survival to discharge was 82.7% when an effective drug began within 30 minutes and 42.0% when it began in the sixth hour; over those six hours, each hour of delay cost an average of 7.6 percentage points.77Kumar A, Roberts D, Wood KE, et al. Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock. Crit Care Med. 2006;34(6):1589-1596. doi:10.1097/01.CCM.0000217961.75225.E9. PMID 16625125. Results -- "Survival was 82.7% if effective antimicrobials were administered within 30 mins of initial evidence of hypotension, 77.2% in the second half hour (79.9% aggregate in the first hour), and 42.0% in the sixth hour"; "mean decrease in survival of 7.6% (range 3.6-9.9%; Fig. 1)". The figure's other survival values were measured from the bar heights of Figure 1 and checked against the three stated values (script and data in figures/). https://pubmed.ncbi.nlm.nih.gov/16625125/ (T2 -- peer-reviewed observational cohort) The study was observational, and sicker patients may both be treated later and die more often; it describes association, not proof.

That curve applies to the minority of patients whom a rapid test finds on the wrong drug, and the trials suggest hospitals already move fastest for them. It does not apply to most of the long waits in the figure. In FAST, 62% of patients were already on an effective drug when they enrolled.11Banerjee R, Komarow L, Li Y, et al. Fast Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia: The FAST Randomized Clinical Trial. JAMA. 2026;335(20):1762. doi:10.1001/jama.2026.5487. Published online April 18, 2026. Introduction ("5 prospective clinical trials ... decreased time to optimal therapy, but did not reduce mortality or shorten duration of hospitalization"), Table 1, Table 2, Figure 4 legend (patients with monomicrobial, on-panel infections) and Discussion ("sites enrolling patients with resistant organisms did not always have the most effective antibiotic therapies available"). https://jamanetwork.com/journals/jama/fullarticle/2847970 (T2 -- peer-reviewed randomised trial) Among those with carbapenem-resistant bacteria, the rapid arm reached an effective drug 18 hours sooner at the median; even there, 30-day mortality was 47.7% against 40.0%, a difference the trial cannot tell from chance. Among the reasons its authors give is supply: sites with resistant infections "did not always have the most effective antibiotic therapies available", and study sites in Greece could not obtain aztreonam for several months during a national shortage.11Banerjee R, Komarow L, Li Y, et al. Fast Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia: The FAST Randomized Clinical Trial. JAMA. 2026;335(20):1762. doi:10.1001/jama.2026.5487. Published online April 18, 2026. Introduction ("5 prospective clinical trials ... decreased time to optimal therapy, but did not reduce mortality or shorten duration of hospitalization"), Table 1, Table 2, Figure 4 legend (patients with monomicrobial, on-panel infections) and Discussion ("sites enrolling patients with resistant organisms did not always have the most effective antibiotic therapies available"). https://jamanetwork.com/journals/jama/fullarticle/2847970 (T2 -- peer-reviewed randomised trial)

On this evidence, the slow lane costs mostly stewardship. Narrowing is how hospitals limit the harm broad antibiotics do. In the Mayo trial, rapid identification cut the hours patients spent on piperacillin-tazobactam, a broad-spectrum workhorse, from 56 to 44; the share treated for bacteria that were mere contaminants of the sample fell from 25% to 11%, and to 8% with the team.22Banerjee R, Teng CB, Cunningham SA, et al. Randomized Trial of Rapid Multiplex Polymerase Chain Reaction-Based Blood Culture Identification and Susceptibility Testing. Clin Infect Dis. 2015;61(7):1071-1080. doi:10.1093/cid/civ447. PMID 26197846. Abstract -- "Time from BCB Gram stain to microorganism identification was shorter in the intervention group (1.3 hours) vs control (22.3 hours)"; "piperacillin-tazobactam (control 56 hours, rmPCR 44 hours, rmPCR/AS 45 hours)"; "less treatment of contaminants (control 25%, rmPCR 11%, rmPCR/AS 8%)"; "de-escalation: rmPCR/AS 21 hours, control 34 hours, rmPCR 38 hours"; "escalation: rmPCR/AS 5 hours, control 24 hours, rmPCR 6 hours". https://pubmed.ncbi.nlm.nih.gov/26197846/ (T2 -- peer-reviewed randomised trial) A 30-day count of deaths will not see those gains.

Measure the time from result to prescription

Rapid diagnostics have been judged by the hours they save the laboratory. The trials suggest the interval that matters now is the next one: from a result appearing in the record to a prescription changing. Its length depends on whether anyone with authority to act is there to see the result. That length is an operational quantity, set by rotas, review rounds and paging rules, and neither trial here reports it directly. Faster tests will still matter. But in today's hospitals, paradoxically, the next hours will be won not in the laboratory but on the ward.

Correction, October 6th 2026: an earlier version of this article subtracted FAST's result time, which the trial reports only for patients with a single organism on the test's panel, from the times at which other groups' antibiotics changed. Those derived intervals have been removed from the text and the figure, and the Mayo Clinic trial's standard-testing arm, omitted from the earlier version, has been added.

Footnotes

1. Banerjee R, Komarow L, Li Y, et al. Fast Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia: The FAST Randomized Clinical Trial. JAMA. 2026;335(20):1762. doi:10.1001/jama.2026.5487. Published online April 18, 2026. Introduction ("5 prospective clinical trials ... decreased time to optimal therapy, but did not reduce mortality or shorten duration of hospitalization"), Table 1, Table 2, Figure 4 legend (patients with monomicrobial, on-panel infections) and Discussion ("sites enrolling patients with resistant organisms did not always have the most effective antibiotic therapies available"). https://jamanetwork.com/journals/jama/fullarticle/2847970 (T2 -- peer-reviewed randomised trial) ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

2. Banerjee R, Teng CB, Cunningham SA, et al. Randomized Trial of Rapid Multiplex Polymerase Chain Reaction-Based Blood Culture Identification and Susceptibility Testing. Clin Infect Dis. 2015;61(7):1071-1080. doi:10.1093/cid/civ447. PMID 26197846. Abstract -- "Time from BCB Gram stain to microorganism identification was shorter in the intervention group (1.3 hours) vs control (22.3 hours)"; "piperacillin-tazobactam (control 56 hours, rmPCR 44 hours, rmPCR/AS 45 hours)"; "less treatment of contaminants (control 25%, rmPCR 11%, rmPCR/AS 8%)"; "de-escalation: rmPCR/AS 21 hours, control 34 hours, rmPCR 38 hours"; "escalation: rmPCR/AS 5 hours, control 24 hours, rmPCR 6 hours". https://pubmed.ncbi.nlm.nih.gov/26197846/ (T2 -- peer-reviewed randomised trial) ↩ ↩² ↩³

3. ClinicalTrials.gov, NCT06174649, "Fast Antibiotic Susceptibility Testing for Gram Negative Bacteremia Trial", results first posted July 28, 2026 -- primary outcome categories and all-cause mortality by arm. https://clinicaltrials.gov/study/NCT06174649 (T1 -- trial registry, sponsor-submitted results) ↩

4. Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021. doi:10.1007/s00134-021-06506-y. Recommendation 29 -- "we suggest daily assessment for de-escalation of antimicrobials" (weak recommendation, very low quality of evidence); rationale -- "Antimicrobial de-escalation should ideally be done as soon as possible, and rapid diagnostic techniques may facilitate this". https://link.springer.com/article/10.1007/s00134-021-06506-y (T1 -- international clinical guideline) ↩ ↩²

5. Centers for Disease Control and Prevention. Core Elements of Hospital Antibiotic Stewardship Programs. 2014 -- "All clinicians should perform a review of antibiotics 48 hours after antibiotics are initiated". The 2019 update adds that "the optimal timing of antibiotic timeouts has not been established" and suggests daily reviews. https://www.cdc.gov/antibiotic-use/healthcare/pdfs/core-elements.pdf (T1 -- federal agency guidance) ↩

6. Study protocol for the FAST trial. Trials. 2025;26:500, published November 12, 2025 -- "If the AST result is ready when the AS provider is unavailable (e.g., in the evening or on a weekend), the AS team will review the MRs and make recommendations no later than the beginning of the next working day." https://link.springer.com/article/10.1186/s13063-025-09228-4 (T2 -- peer-reviewed protocol) ↩

7. Kumar A, Roberts D, Wood KE, et al. Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock. Crit Care Med. 2006;34(6):1589-1596. doi:10.1097/01.CCM.0000217961.75225.E9. PMID 16625125. Results -- "Survival was 82.7% if effective antimicrobials were administered within 30 mins of initial evidence of hypotension, 77.2% in the second half hour (79.9% aggregate in the first hour), and 42.0% in the sixth hour"; "mean decrease in survival of 7.6% (range 3.6-9.9%; Fig. 1)". The figure's other survival values were measured from the bar heights of Figure 1 and checked against the three stated values (script and data in figures/). https://pubmed.ncbi.nlm.nih.gov/16625125/ (T2 -- peer-reviewed observational cohort) ↩