The 35% Problem: Why Isolate-Based AST Is Systematically Under-Detecting Resistance in Complicated Urinary Tract Infection

The clinician prescribed empirically. The lab report said susceptible. Three weeks later, the patient was back — same symptoms, same pathogen.

This sequence plays out hundreds of thousands of times each year in hospitals across the United States, and the diagnostic framework responsible for it has remained fundamentally unchanged for nearly seven decades. Standard urine culture followed by isolate-based antimicrobial susceptibility testing (AST) has been the cornerstone of complicated urinary tract infection (cUTI) management since the Kass criteria were first published in 1962. The model rests on three assumptions: the bladder is sterile unless infected by a single dominant pathogen, clinical cure equals microbiologic clearance, and the behavior of a subcultured isolate in the lab reflects what is happening inside the patient.

Converging evidence from multiple independent research groups now shows that all three assumptions are wrong — and that the consequences are measurable.

The Shadow Failures

In the largest FDA analysis of cUTI clinical trials ever published, Kadry and colleagues evaluated participant data from 13 independent Phase 3 trials submitted to the agency between 2011 and 2019. The study, published in Clinical Infectious Diseases in 2023, encompassed 4,842 patients and provides the most robust evidence to date that microbiologic persistence is a primary driver of clinical relapse in cUTI.

The numbers are difficult to dismiss. At the test-of-cure visit — typically five or more days after completing antibiotic therapy — 70.7% of patients achieved what the researchers called concordant success: clinical cure plus microbiologic eradication. That sounds like a passing grade until you look at the other 30%.

A striking 18% of participants were classified as discordant failures. They felt better. Their symptoms had resolved. But their urine still harbored the original pathogen at clinically significant levels. Another 6.7% failed on both counts — symptoms persisted and the bug was still there.

The clinical consequences of that 18% became apparent at the late follow-up visit, 21 to 28 days after randomization. More than a quarter of the discordant patients — 26.7% — experienced symptomatic relapse. Among the patients who had achieved true concordant success, the relapse rate was just 6.2%. Patients whose infections appeared cured but weren’t eradicated were five and a half times more likely to relapse within three weeks.

This is not a rare edge case. It is nearly one in five patients across thirteen major clinical trials. And the FDA’s rationale for requiring the composite endpoint — clinical cure plus microbiologic eradication — is validated by exactly this data.

The question the Kadry analysis does not answer is why these patients retain their infections despite apparently effective therapy. Three mechanisms, each supported by independent research, provide the explanation.

Where the Resistance Hides

The Isolation Tax

Standard urine culture selects the fastest-growing, most dominant organism in the specimen and tests that organism alone. Everything else — the slower-growing species, the minority resistant populations, the fastidious organisms that don’t thrive on standard agar — gets left behind on the plate. In clinical reporting, many specimens are labeled “mixed flora” or “contaminated” under the College of American Pathologists standard, which dictates that a urine culture containing more than two isolates should be treated as contamination.

Research published by Moreland and colleagues in Frontiers in Cellular and Infection Microbiology in 2025 demonstrates how much this standard misses. When the investigators compared standard urine culture to multiplex PCR in symptomatic patients aged 65 and older, M-PCR identified polymicrobial infections in 56.1% of the cohort. Most of these were reported as contaminated by standard methods.

The CAP contamination rule made sense in 1970, when the prevailing view was that the bladder was sterile. The urobiome discovery proved it is not. Applying Koch’s postulates to a polymicrobial reality is not conservative laboratory practice — it is a structural blind spot.

The Hidden Resistance Within

If isolation bias filters out entire organisms, heteroresistance hides resistance within organisms that appear susceptible.

Heteroresistance occurs when a bacterial population contains a minor subpopulation of resistant cells — often at frequencies of one in ten thousand to one in ten million — that can rapidly expand under antibiotic pressure. One of the most common methods of antimicrobial susceptibility testing, broth microdilution (BMD), uses a standardized inoculum of approximately 500,000 colony-forming units per milliliter. At that concentration, the math simply does not work: the minor populations of resistant bacteria are almost never present during the incubation time of the test. The lab reports “susceptible.” The clinician prescribes accordingly. And the resistant subpopulation, freed from competition by the antibiotic’s elimination of the susceptible majority, expands to cause relapse.

Van Riel and colleagues analyzed isolates from the ALLIUM Phase 3 clinical trial and found that 11.5% of cUTI isolates were heteroresistant to piperacillin-tazobactam. Every one of these isolates tested susceptible by standard BMD. Patients infected with heteroresistant isolates had more than double the rate of treatment failure, with an odds ratio of 2.13.

Separately, Orfano and colleagues examined 355 urinary E. coli isolates from the CDC-funded Georgia Emerging Infections Program and found that 7.6% were heteroresistant to fosfomycin and 3.6% to trimethoprim-sulfamethoxazole — despite testing susceptible by standard methods.

“When you test the whole specimen instead of a subcultured isolate, you see resistance that the isolate never showed you. It’s not that the isolate result is wrong — it’s that it only represents the dominant population. The minority populations that survive treatment are invisible to that workflow.”

Joseph D. Kittle Jr, CSO, CytoSPAR
Lead author, BLAST rapid bacterial labeling method, Diagnostic Microbiology & Infectious Disease, 2026

The 24-Hour Time Tax

The isolation step is not just a source of selection bias. It is a 24-hour delay before susceptibility testing can even begin. During that window, the patient is on empiric therapy that may or may not be appropriate.

Thorpe and Joski published the most comprehensive national analysis of this delay in the Journal of Medical Economics in 2025, using a representative dataset of privately insured adults in the United States.

The delay is not merely a time penalty. It is a 24-hour window during which the wrong antibiotic may be actively selecting for resistance, driving the patient toward the very treatment failure the susceptibility test was supposed to prevent.

The Market Signal

If the clinical evidence establishes that the current diagnostic model is broken, the market has already delivered its own verdict.

The Accelerate Pheno system used a proprietary instrument and cartridge system to deliver phenotypic AST from positive blood cultures in approximately seven hours — 40 hours faster than conventional methods. Accelerate Diagnostics received FDA clearance for the system in 2017 through the De Novo pathway, and the technology represented a genuine advance in rapid diagnostics.

The RAPIDS-GN study, conducted across multiple large healthcare centers, demonstrated statistically significant reductions in time to first antibiotic change. The verdict was clear. The science was sound. The test was instituted at multiple major hospitals and healthcare systems. They believed this was the opportunity to improve outcomes for their patients.

However, by December 2024, Accelerate had accumulated $718.9 million in losses driven by the high costs of manufacturing and maintaining the proprietary test platform and ultimately by low adoption rates that could not sustain the business. The company filed for Chapter 11 bankruptcy in May 2025.

The Accelerate bankruptcy proves two things simultaneously. First, the clinical need for rapid phenotypic AST is real — multiple hospitals and large hospital systems would not have deployed the system otherwise. Second, the next generation of rapid AST cannot depend on dedicated hardware. Kit-based approaches that run on existing laboratory infrastructure have a structural commercial advantage that instrument-dependent platforms do not.

The Paradigm Shift

The most significant development in this space may not be a new technology at all. It may be a new way of using technology that is already familiar to every lab in the country.

In 2026, the Antibacterial Resistance Leadership Group published a paper in Clinical Microbiology and Infection proposing a fundamentally different framework for how we assess new diagnostic methods. The paper, authored by Doernberg and colleagues from the ARLG Laboratory Center at Mayo Clinic, argues that we should no longer use “agreement with culture” as the sole benchmark for evaluating new diagnostics.

Instead, they propose clinical adjudication committees that determine whether a new test correctly identified a pathogen or resistance pattern that culture missed. This is not a minor procedural tweak. It is a regulatory and methodological inflection point. It acknowledges that the gold standard is flawed — and that new direct-from-sample tests should be allowed to be “more correct” than the legacy method.

This distinction matters enormously for the next generation of phenotypic AST methods — particularly those designed to work directly on clinical specimens without the isolation step. These methods will inevitably disagree with culture-based results in a significant fraction of cases. The Kadry data tells us why: culture misses resistance in at least 18% of patients. The question has always been whether regulators and clinicians would accept that disagreement as evidence that the new test is more sensitive, rather than less accurate. The ARLG framework opens the door — though some clinicians may still require further convincing before adopting this shift in practice.

The Implication

Direct-from-sample phenotypic AST — methods that bypass isolation entirely and measure the metabolic or physiological response of live bacteria to antibiotics in the original clinical specimen — is no longer theoretical. Multiple approaches are now in clinical validation, several with published data demonstrating MIC concordance with reference methods across diverse panels of Gram-negative and Gram-positive pathogens. Emerging data from at least one 740-patient direct-from-urine study suggests that these methods detect clinically significant resistance in approximately 35% of specimens where isolate-based testing reports susceptibility.

That number — 35% — should stop any clinical microbiologist, any infectious disease physician, any pharma medical director running a cUTI trial mid-sentence. It means that for roughly one in three patients, the susceptibility report guiding treatment may be incomplete. It means that the microbiologic eradication rates in Phase 3 cUTI trials may be artificially inflated by a reference method that does not see the full resistance picture. And it means that the empiric prescribing failures documented by Thorpe, the shadow relapses documented by Kadry, and the hidden heteroresistance documented by Van Riel are not isolated findings — they are symptoms of the same structural flaw.