The drug in question is adalimumab, a fully human monoclonal antibody that neutralises tumour necrosis factor alpha, or TNFα, a cytokine central to the inflammatory cascade that drives retinal tissue damage in uveitis by promoting mononuclear cell trafficking into ocular tissues. Adalimumab is the only non-corticosteroid drug licensed for immune-mediated non-infectious uveitis, and its efficacy was established in the landmark VISUAL I and VISUAL II trials, which showed that it delayed treatment failure in patients with active disease and in those with inactive but steroid-dependent disease. Yet those trials had design limitations with real-world consequences: they did not test adalimumab as an add-on to standard therapy, the most common clinical scenario when first-line agents fail; they excluded cystoid macular oedema, the leading cause of uveitis-related visual loss, from both inclusion and treatment-failure criteria; and they required high levels of vitreous inflammation for enrolment. Those restrictions translated into restrictive access criteria worldwide, including a recommendation from the UK’s National Institute for Health and Care Excellence that bars patients with unilateral disease without systemic inflammation, steroid-dependent inactive uveitis, or those new to standard immunosuppressive therapy.

The new study, called ASTUTE, was designed specifically to close those evidence gaps. Conducted at 18 NHS tertiary ophthalmology departments across the United Kingdom, it was a multi-centre, parallel-group, placebo-controlled randomised trial with an unusual twist: a 16-week treatment run-in before randomisation. All participants first received open-label adalimumab, delivered as fortnightly subcutaneous 40-milligram injections of a licensed biosimilar, Imraldi, while their oral prednisolone was tapered toward 5 milligrams per day or less, with the rate of reduction guided by disease activity and clinician judgement rather than a fixed schedule. Only participants who responded, defined as achieving prednisolone at or below 5 milligrams per day with no sign of active disease, were then randomised to continue adalimumab or switch to an identical placebo pen containing all excipients but no active drug. This enriched enrolment, randomised withdrawal design, more often seen in industry-sponsored analgesia trials, served two purposes: it ensured the drug was only continued in patients in whom it was demonstrably working, and it provided an objective, biomarker-free criterion for stopping treatment, something absent from routine clinical practice.

The trial’s eligibility criteria were deliberately broad. Patients aged 18 or over could enter the run-in if they had active sight-threatening uveitis, marked by inflammatory chorioretinal lesions, central macular thickness above 320 microns, retinal vasculitis or vitreous haze greater than 0.5, and were taking more than 5 milligrams per day of oral prednisolone; or if they had controlled disease that still required more than 5 milligrams per day of prednisolone alongside conventional immunomodulatory drugs. Concomitant immunosuppression such as mycophenolate mofetil, methotrexate, azathioprine, ciclosporin or tacrolimus was permitted and recorded throughout. The result was a study population far more diverse than those in previous trials, spanning intermediate, posterior and panuveitis, with birdshot retinochoroiditis the single most common diagnosis, and including patients with cystoid macular oedema and mild to moderate vitreous inflammation who would previously have been turned away.

Between July 2021 and June 2024, 323 patients were screened, 202 enrolled in the treatment run-in, and 119, or 65.4 percent, responded well enough to be randomised. After four withdrawals, 115 participants entered the randomised phase, 60 receiving adalimumab and 55 placebo. The primary outcome was time to the first treatment failure, a composite endpoint capturing any of six signs of reactivated disease: a drop of 15 or more letters in best-corrected visual acuity on ETDRS charts, new active chorioretinal lesions, a greater than 20 percent increase in central macular thickness on optical coherence tomography, onset or worsening of retinal vasculitis on fluorescein angiography, a two-step worsening of vitreous haze on the Standardization of Uveitis Nomenclature scale, or the need for prednisolone above 5 milligrams per day to maintain disease control. Topical steroids were allowed for anterior chamber activity, a pragmatic choice reflecting that isolated flare manageable with drops would not prompt withdrawal of a biologic in practice, while escalation to oral prednisolone was counted as failure.

The results were striking. At 12 months, an estimated 31.4 percent of participants on adalimumab had experienced treatment failure, compared with 59.0 percent on placebo, a hazard ratio of 0.42 with a 95 percent confidence interval of 0.23 to 0.75 and a p-value of 0.003. In other words, adalimumab cut the hazard of treatment failure by more than half. The most frequently observed component of failure was the prescription of higher-dose prednisolone to maintain remission, underscoring how much of the drug’s benefit lies in enabling genuine corticosteroid sparing, an outcome patients strongly prefer given the substantial toxicity of long-term steroids. Adherence was high, with more than 90 percent of scheduled injections received in both groups, and a per-protocol sensitivity analysis produced nearly identical results.

Perhaps the most consequential finding came from a pre-specified subgroup analysis. Of the 115 randomised participants, 98, roughly 85 percent, would have been ineligible for adalimumab under the current NICE technology appraisal, whether because of steroid-dependent inactive disease, unilateral isolated uveitis, mild to moderate vitreous inflammation, macular oedema alone, or being new to standard immunosuppression. Yet these patients benefited at least as much as those who met the eligibility criteria: 30.0 percent of the NICE-ineligible group on adalimumab experienced treatment failure at 12 months versus 60.0 percent on placebo, a hazard ratio of 0.37. The benefit seen in this broader population was also of a similar magnitude to that reported in the VISUAL trials, and the biosimilar used reflects a class effect of adalimumab rather than a property of any single branded product.

Secondary outcomes painted a broadly favourable but nuanced picture. Patient-reported visual function improved more with adalimumab than placebo, with a statistically significant difference emerging after week 48; at week 80 the geometric mean ratio was 0.56, where a lower score indicates better vision-related quality of life. Other measures, including the EQ-5D utility score, visual acuity and central macular thickness, showed point estimates favouring adalimumab that did not reach statistical significance, a difficulty the investigators attribute to the fact that participants were not followed in the trial after treatment failure, so between-group comparisons drew predominantly on patients still in remission. Safety data were reassuring across two years: adverse event rates were actually lower on adalimumab than placebo for both infections and infestations, at 47.9 versus 73.2 per 100 person-years, and skin disorders, at 8.5 versus 11.7 per 100 person-years, though neither difference was statistically significant. There were no deaths and no suspected unexpected serious adverse reactions.

The economics also favour wider access. From randomisation to first treatment failure, participants on adalimumab incurred £1,117 in additional NHS costs, driven mainly by the drug itself, and the incremental cost-effectiveness ratio worked out at £2,320 per treatment failure avoided. Set against the economic burden of a treatment failure, which typically demands intensive therapy with additional imaging, high-dose steroids, immunomodulatory drugs, intravitreal injections and frequent monitoring, alongside productivity losses and informal care, the investigators argue this represents good value for money. A recent systematic review reported annual direct medical costs of non-infectious uveitis ranging from $16,428 to $134,135 per patient in 2023 US dollars, rising more than fourfold in those who become blind, providing context for the economic weight of preventing relapse. A full cost-utility analysis incorporating quality-adjusted life years is still to come from the trial.

The trial was not without limitations. Recruitment fell short of the target of 174 randomised participants, largely because the COVID-19 pandemic delayed the start and hospitals later prioritised clinical backlogs over research, and recruitment had to stop early when the placebo pens neared expiry on 31 January 2025 and the manufacturer could not extend their shelf life or supply new stock. Participants then transitioned into an ongoing open-label extension, which should establish the durability of response and longer-term safety. Even so, the message is clear: adalimumab can be used effectively and safely as an add-on therapy for a substantially broader population of patients with non-infectious uveitis than current guidance recognises, and the authors argue the NICE recommendation should now be reviewed. Roughly 30 percent of treated patients still fail within a year, however, highlighting a continuing unmet need for additional and complementary treatments for those whose disease remains refractory even to TNFα blockade.

Subject of Research: Efficacy and cost-effectiveness of adalimumab as add-on therapy for non-infectious uveitis in the ASTUTE randomised controlled trial

Article Title: Adalimumab versus placebo as add-on to standard therapy for non-infectious uveitis in the United Kingdom: the ASTUTE randomised controlled trial with treatment run-in

Article References: Dick, A. D., Reeves, B. C., Beare, N. A., Lui, M., Pike, K., Baos, S., Culliford, L., Zhu, X., Wordsworth, S., Folkard, A., Lanyon, K., Koleva-Kolarova, R., Brierley, R. C., Peto, T., Madhusudhan, S., Sharma, S. M., Epps, S., Naruganahalli, K., Clark, N., … Hamill, B. (2026). Adalimumab versus placebo as add-on to standard therapy for non-infectious uveitis in the United Kingdom: the ASTUTE randomised controlled trial with treatment run-in. eClinicalMedicine, 101, Article 104240. https://doi.org/10.1016/j.eclinm.2026.104240

Image Credits: AI Generated

DOI: Not provided

Keywords: adalimumab, non-infectious uveitis, randomised controlled trial, TNF-alpha inhibitor, biosimilar, corticosteroid sparing, NICE guidelines, treatment failure, macular oedema, ophthalmology, cost-effectiveness, immunomodulatory therapy

The study, led by Farzad Pourghazi and colleagues at the Mayo Clinic together with international collaborators, searched MEDLINE and Scopus through July 2025 and identified 27 cohort studies encompassing 1,399 patients with culture-negative periprosthetic joint infection. The infections spanned 584 hips, 809 knees, five shoulders, and a single elbow. Using random-effect statistical models designed to handle the considerable variability between studies, the team calculated a pooled treatment success rate of 83.6 percent, a pooled failure rate of 17.2 percent, and a pooled mortality of just 2.4 percent. For a condition in which both antibiotic selection and surgical planning are typically empirical, those numbers suggest that the absence of a named pathogen does not, by itself, doom a patient to a poor outcome.

The scale of the diagnostic problem is worth appreciating. Periprosthetic joint infection complicates roughly 2 percent of the millions of total joint replacements performed worldwide each year, and culture-negative cases account for somewhere between 5 and 42 percent of all such infections, depending on the population and the diagnostic methods used. Modern laboratories have an arsenal of tricks to coax hidden organisms into view, including prolonged incubation, sonication of the removed implant to dislodge adherent biofilm bacteria, and molecular assays such as next-generation sequencing that can detect bacterial DNA without requiring growth in culture. Yet even with these tools, a subset of infections yields no organism, and clinicians must treat blind.

Why do cultures fail in these cases? The most likely explanations, long discussed in the literature, include prior antibiotic exposure before samples were taken, slow-growing or fastidious organisms that resist standard culture conditions, and bacteria hiding within biofilms on the implant surface in forms that do not grow on routine media. Prior antimicrobial therapy has been identified as a risk factor for culture negativity, which is one reason some experts advocate holding antibiotics before surgery whenever clinically safe. Whatever the cause, the practical consequence is the same: the surgical team must choose between treatment strategies and pick an antibiotic regimen without knowing exactly what they are treating.

The surgical options fall into three broad categories. Debridement, antibiotics, and implant retention, known as DAIR, involves opening the joint, washing out infected tissue, and keeping the original prosthesis in place; it is generally reserved for acute infections with a well-fixed implant and no sinus tract. One-stage exchange arthroplasty removes and replaces the implant in a single operation, while two-stage exchange, often considered the standard of care for chronic infection, removes the prosthesis, places a temporary antibiotic-loaded spacer, and reimplants a new device only after a course of antibiotics. The new meta-analysis compared these strategies directly and found significant differences.

The numbers were striking. DAIR achieved a pooled success rate of only 73.2 percent, with failure in 29.4 percent of cases, whereas two-stage exchange reached 87.4 percent success with 14.0 percent failure, and one-stage exchange achieved 86.2 percent success with 13.8 percent failure. In other words, patients treated with exchange arthroplasty fared substantially better than those whose implants were retained. The authors are careful to note that this does not mean culture status alone should dictate surgical strategy; the choice between DAIR and exchange depends heavily on infection chronicity, implant stability, the patient’s overall health, and soft-tissue conditions. But the pattern is consistent with the broader understanding that DAIR works best in a narrow, carefully selected group of early infections.

An intriguing secondary finding emerged when the researchers split the data by joint. Hip infections showed a pooled success rate of 94.2 percent, while knee infections lagged at 75.8 percent, a statistically significant difference. The authors urge caution here, noting that this gap may reflect differences in patient selection, infection characteristics, or treatment practices between the hip and knee populations rather than any true biological effect of the joint itself. Still, the finding adds to a long-running debate in the orthopedic literature about whether knee infections are intrinsically harder to eradicate.

On the antimicrobial side, the picture was far messier. Antibiotic regimens across the 27 studies were so heterogeneous and so incompletely reported that the authors could not pool them statistically and instead summarized them descriptively. Vancomycin-based postoperative regimens were the most commonly reported, either alone or combined with other agents, reflecting the need to cover staphylococci, the most frequent causes of prosthetic joint infection, including resistant strains. Cephalosporins such as cefazolin, ceftriaxone, ceftazidime, and cefepime appeared frequently, and fluoroquinolones like ciprofloxacin and levofloxacin were common choices for oral step-down therapy after an initial intravenous phase. Treatment durations varied widely. This individualized, protocol-free approach likely reflects the thin evidence base for empirical therapy, but it also raises concerns: broad-spectrum regimens carry risks of toxicity and adverse events, and experts have called for better diagnostics and stronger antimicrobial stewardship in this space.

The study is not without caveats, and the authors are candid about them. Twenty-six of the 27 included studies were retrospective, exposing the analysis to selection bias, confounding, and incomplete reporting. Definitions of treatment success, failure, and reinfection varied from institution to institution despite repeated attempts by professional societies to standardize them, and this inconsistency fueled substantial statistical heterogeneity. More troubling, funnel plot asymmetry and Egger’s regression test suggested possible publication bias: trim-and-fill analysis, which imputes hypothetically missing studies, lowered the adjusted success estimate to 74.0 percent and raised the adjusted failure estimate to 25.6 percent. The true picture, in other words, may be somewhat less rosy than the raw pooled numbers suggest, though still consistent with a generally favorable prognosis.

What emerges from this work is a dual message. On one hand, culture-negative periprosthetic joint infection, long viewed with apprehension precisely because the pathogen is unknown, appears to carry outcomes comparable to, and in some analyses better than, culture-positive disease, particularly when managed with exchange arthroplasty guided by established surgical principles and sensible empirical antibiotics. On the other hand, the evidence remains observational, heterogeneous, and vulnerable to bias. The authors call for prospective, multi-center studies with standardized definitions of culture-negative infection and uniform reporting of outcomes, surgical selection criteria, and antimicrobial regimens. Until such studies arrive, clinicians treating an infected joint replacement with negative cultures can take some comfort: treating blind, it turns out, is not the same as treating badly.

Subject of Research: Treatment strategies and outcomes of culture-negative periprosthetic joint infection

Article Title: Management and outcomes of culture-negative periprosthetic joint infection: a systematic review and meta-analysis

Article References: Pourghazi, F., Alavi, S. M. A., Matsuo, T., Baaklini, G., Abdel, M. P., Tande, A. J., & Berbari, E. F. (2026). Management and outcomes of culture-negative periprosthetic joint infection: a systematic review and meta-analysis. Journal of Bone and Joint Infection, 11(4), 413-430. https://doi.org/10.5194/jbji-11-413-2026

Image Credits: AI Generated

DOI: 10.5194/jbji-11-413-2026

Keywords: periprosthetic joint infection, culture-negative infection, total joint arthroplasty, meta-analysis, DAIR, exchange arthroplasty, antibiotic therapy, orthopedic surgery, vancomycin, treatment failure, systematic review, implant infection

The study, led by Renato Pascale and Sara Tedeschi of the University of Bologna together with colleagues at the IRCCS Istituto Ortopedico Rizzoli and Maria Cecilia Hospital, followed 161 adults diagnosed with deep surgical-site infections following spinal instrumentation between January 2018 and June 2022. These infections, abbreviated DSII, involve tissues beneath the fascia surrounding implanted hardware, and their management is uniquely difficult. The implants themselves provide a surface on which bacteria can form biofilms, those resilient, glue-like microbial communities that shield pathogens from both antibiotics and immune cells. Standard care therefore combines surgical debridement, the physical removal of infected tissue, with prolonged antibiotic therapy, and the choice between retaining, replacing, or removing the hardware is individualized according to spinal stability and the severity of infection.

The patient population reflected the demanding reality of complex spine surgery. Participants had a median age of 62 years, a median Charlson comorbidity index of 3, and more than half were men. Degenerative disease and scoliosis each accounted for about 29 percent of the original operations, but a striking 19.3 percent had undergone instrumentation because of cancer. Infections typically declared themselves early: 70.2 percent were diagnosed within the first three months after surgery, with a median interval of 27 days between the primary operation and the diagnosis of infection. Wound dehiscence, the reopening of the surgical incision, was the most common warning sign, occurring in 59 percent of cases. Polymicrobial infections, in which several bacterial species are present at once, made up just over a quarter of the cohort.

Despite meticulous care at two referral centers for bone and joint infection, treatment failure was common. Fifty-one patients, or 31.7 percent, ultimately failed treatment, an outcome defined by wound reopening, the appearance or recurrence of a fistula, new pain or functional decline after apparent resolution, or death attributable to the infection. Among the failures, wound reopening dominated, occurring in nearly two-thirds of cases, while 11 patients developed fistulas and three died. The figures are a sobering reminder that even in high-volume centers with dedicated infectious disease consultants, roughly one in three patients with an infected spinal construct will not be cured by the first combined medical and surgical assault.

When the researchers dug into what separated the failures from the cures, one microbial distinction stood out with statistical force. Gram-positive bacteria, the traditional culprits in implant infections, caused 60.3 percent of cases, but the 33.8 percent of infections driven by Gram-negative rods, including Pseudomonas aeruginosa and Escherichia coli, were disproportionately linked to bad outcomes. Gram-negative etiology proved to be an independent risk factor for treatment failure in two separate multivariable analyses: the odds of failure were roughly four times higher for these infections whether the treatment window was framed as eight weeks (odds ratio 4.20) or twelve weeks (odds ratio 4.46). This aligns with a growing body of evidence from prosthetic joint infections suggesting that Gram-negative pathogens are inherently harder to eradicate from foreign material, a fact that current understanding of spinal implant infections, still surprisingly thin, has only begun to absorb.

The second headline finding concerns quinolones, the antibiotic class that includes ciprofloxacin and levofloxacin. Patients whose regimens included a quinolone had dramatically lower odds of treatment failure, with odds ratios of 0.11 in both the eight-week and twelve-week analyses, meaning the odds of failure were reduced by nearly 90 percent in this observational dataset. The biological rationale is plausible: quinolones penetrate bone well, are active against many Gram-negative rods, and can be administered orally for months. Yet the investigators themselves urge caution. Because quinolones were likely reserved for patients with susceptible pathogens and optimized targeted therapy, the association may reflect careful antimicrobial selection within a higher-risk subgroup rather than a genuine causal benefit. Residual confounding, they acknowledge, cannot be excluded in a retrospective design.

Perhaps the most consequential result is the one that did not appear. When the team compared treatment schedules of eight versus twelve weeks, excluding patients who failed early during therapy to isolate the effect of a completed course, they found no significant association between duration and outcome. To guard against a subtle statistical trap known as immortal time bias, in which patients must survive long enough to complete a long regimen before being counted as such, the researchers also performed prespecified landmark sensitivity analyses at eight and twelve weeks, restricting comparisons to patients still at risk at each time point. Those analyses, conducted on 90 and 75 patients respectively, confirmed the primary result: longer was not demonstrably better. The median treatment duration in the cohort was 82 days, and the question of whether some patients could safely stop earlier remains, as the authors stress, unresolved rather than settled.

The study’s methodology reveals both its strengths and its limits. The investigators applied a rigorous Delphi consensus definition of postoperative spine infection, using hierarchical criteria spanning microbiology, imaging, laboratory values, intraoperative findings, and histology, and a dedicated infectious disease team made every diagnosis. Firth penalized logistic regression was used to temper the small-sample bias that arises when only 51 events are distributed across multiple covariates. Yet the same constraints shape the conclusions. The study spans two hospitals in the same Italian region, limiting generalizability; treatment duration was chosen by individual physicians rather than randomized, inviting confounding by indication; the rate of culture-negative infections exceeded that reported in prior literature; and detailed surgical variables such as the number of instrumented levels were not consistently available for adjustment. The authors explicitly label their duration findings exploratory.

What emerges is a reframing of the clinical conversation around one of orthopedic surgery’s costliest complications. Rather than asking only how long to treat, the DeFEndO study suggests clinicians should ask what they are treating. Cancer patients, whose immunosuppression, chemotherapy-induced mucosal damage, and malnutrition impair infection control, fared worse, and so did those harboring Gram-negative rods. Rifampin, long championed for biofilm infections of prosthetic joints, showed no protective signal here, echoing the mixed results of randomized trials elsewhere. The authors call for prospective studies, particularly to define optimal therapy for Gram-negative spinal infections, where evidence remains scarcest. Until then, the message for surgeons and infectious disease physicians is quiet but profound: the calendar may matter less than the microbe, and curing a spine seeded with infection begins with knowing exactly which organism is hiding on the hardware.

Subject of Research: Risk factors for treatment failure of deep surgical-site infections following spinal instrumentation

Article Title: Risk factors for treatment failure of surgical site deep infections following spinal instrumentation: multicentre observational study (DeFEndO study)

Article References: Pascale, R., Gallo, M., Regge Gianas, N., Fornaro, G., Grassi, I., Zagarrigo, M., Grechi, A., Zuppiroli, A., Santoro, D., Tenti, E., Boriani, L., Zamparini, E., Gasbarrini, A., Viale, P., Giannella, M., Curti, S., & Tedeschi, S. (2026). Risk factors for treatment failure of surgical site deep infections following spinal instrumentation: multicentre observational study (DeFEndO study). Journal of Bone and Joint Infection, 11(5), 601-611. https://doi.org/10.5194/jbji-11-601-2026

Image Credits: AI Generated

DOI: 10.5194/jbji-11-601-2026

Keywords: spinal instrumentation, surgical site infection, deep infection, treatment failure, antibiotic duration, quinolones, Gram-negative bacteria, biofilm, rifampin, spine surgery, multicentre observational study, infectious diseases

The research team, led by Pauline Gesson and Clémence Houizot of Nantes University Hospital together with corresponding author Adrien Evin, examined all patients who underwent continuous deep sedation until death in their palliative care unit between 2020 and 2024. Of 1,376 patients admitted during the study window, 287 received continuous deep sedation with midazolam as the first-line agent. The cohort had a median age of 69 years, with an interquartile range of 59 to 77, and just over half of the patients, 52.3 percent, were male. Cancer was the underlying diagnosis in 74.2 percent of cases, reflecting the typical caseload of a dedicated palliative care unit. The researchers collected demographic, clinical, pharmacological, and organisational data from medical records and then applied univariable and multivariable logistic regression to identify factors associated with what they defined as failure of midazolam monotherapy: the need to introduce additional sedative drugs.

The headline finding is stark. Midazolam monotherapy failed in 211 of the 287 patients, a failure rate of 73.5 percent. In other words, roughly three out of every four patients who entered continuous deep sedation on midazolam alone ultimately required at least one additional sedative to achieve or maintain adequate sedation until death. Escalation most commonly involved chlorpromazine, a first-generation antipsychotic with sedative properties, which was added alone in 134 of the 211 escalated patients, or 63.5 percent. Propofol, a short-acting intravenous anaesthetic agent, was used alone in 20 patients, or 9.5 percent, while both agents were required in 57 patients, or 27 percent. These proportions sketch a practical hierarchy at the bedside: when midazolam is insufficient, French palliative teams reach first for chlorpromazine, reserving propofol, and combinations of the two, for the most resistant cases.

The statistical analysis uncovered several independent determinants of failure, and the most striking of them was polypharmacy. Patients receiving between five and nine concomitant medications had roughly three times the odds of midazolam failure compared with those on fewer drugs, with an odds ratio of 3.06 and a 95 percent confidence interval of 1.60 to 5.91. Remarkably, the risk did not continue to climb with further drug burden: patients on ten or more medications had an almost identical odds ratio of 3.05, with a confidence interval of 1.22 to 8.11. The effect appears to be a threshold phenomenon rather than a dose-response relationship. The authors suggest that the pharmacological milieu created by multiple co-administered drugs, including potential interactions at the level of cytochrome P450 metabolism, protein binding, or receptor dynamics at the GABA-A receptor where midazolam acts, may blunt the drug’s effectiveness in ways that are difficult to predict from any single interaction pair.

The clinical context in which sedation was initiated also mattered profoundly. When continuous deep sedation was performed in the setting of withdrawal of life-sustaining treatment, the odds of midazolam monotherapy failure more than tripled, with an odds ratio of 3.51 and a confidence interval of 1.46 to 9.88. This finding carries particular weight for intensive care and palliative teams who manage the transition when ventilatory support or other organ support is withdrawn. The physiological state of a patient undergoing such withdrawal, including accumulating carbon dioxide, hypoxaemia, agitation, and the pharmacodynamic turbulence of the dying process itself, may create symptom burdens that a benzodiazepine alone cannot reliably control. The result suggests that teams should anticipate the likely need for adjunctive sedation from the outset in these situations rather than treating escalation as an unexpected event.

Perhaps the most scientifically intriguing result concerns age. Rather than treating age as a simple linear variable, the researchers modelled it with restricted cubic splines, a technique that allows the data to reveal non-linear patterns. Age showed a significant non-linear association with sedation failure, with a p-value below 0.001. Although the abstract does not spell out the exact shape of the curve, the very existence of a non-linear effect challenges the common assumption that older patients are uniformly more sensitive to sedatives and therefore easier to sedate. End-of-life pharmacology is shaped by shifting volumes of distribution, changing albumin levels, altered hepatic blood flow, renal decline, and receptor sensitivity changes that do not map neatly onto chronological age. The finding underscores that dosing strategies anchored solely on age-based rules risk being wrong in both directions across the lifespan.

One factor moved in the opposite direction. Patients who had been receiving oxygen therapy in the 24 hours preceding the induction of sedation had a substantially reduced risk of midazolam failure, with an odds ratio of 0.35 and a confidence interval of 0.18 to 0.65. The authors do not over-interpret this protective association, and it may well reflect confounding by indication: patients already on oxygen may represent a different physiological subgroup, or the oxygen may itself alleviate distressing dyspnoea that would otherwise compete with sedation. Nevertheless, the signal is strong enough to warrant attention, and it hints that pre-sedation respiratory status and symptom control in the final day of life may shape how the patient responds to sedative drugs.

The sheer frequency of escalation documented in this single-centre cohort raises questions about whether midazolam monotherapy should remain the default expectation in guidelines and in clinical practice. The authors are careful in their framing: the findings do not argue against midazolam as first-line therapy, which remains supported by its favourable pharmacokinetic profile, titratability, and extensive clinical experience. Instead, they argue for individualised sedation strategies and for early anticipation of combination therapy in patients who carry the identified risk factors. In practical terms, a patient on multiple medications, or a patient whose sedation accompanies the withdrawal of life-sustaining treatment, might benefit from a plan that pre-specifies the dose and timing of an adjunctive agent such as chlorpromazine, rather than leaving the team to improvise an escalation in the middle of the night.

The study’s limitations are those inherent to its design. It was retrospective and conducted at a single palliative care unit, which means the results reflect the protocols, patient mix, and escalation habits of one French university hospital team. The definition of failure, the need for second- or third-line agents, is operational rather than physiological, and it captures both true pharmacological resistance and clinician prescribing behaviour. With 287 sedation episodes, the confidence intervals around some estimates, particularly for the highest polypharmacy category, are wide. Still, the internal consistency of the findings, the use of spline modelling for age, and the adjustment for multiple variables in the multivariable analysis give the results credibility, and the near-universal relevance of the clinical scenario makes external validation studies an obvious next step.

For the growing number of countries legislating or debating continuous deep sedation at the end of life, from France with its Claeys-Leonetti framework to jurisdictions across Europe and beyond, this study delivers a sobering empirical message. The most widely used sedative in the world fails as a single agent in nearly three quarters of the patients to whom it is given for this purpose, and the failure is not random. It clusters around measurable, identifiable characteristics: the burden of concurrent medications, the circumstances of treatment withdrawal, the patient’s position in a complex age-related pharmacological landscape, and the respiratory interventions of the final day. Translating that knowledge into anticipatory prescribing protocols could spare dying patients the distress of inadequate sedation and spare their families the sight of it, while giving clinicians a firmer evidence base for one of medicine’s most delicate decisions.

Subject of Research: Determinants of failure of midazolam monotherapy during continuous deep sedation until death in palliative care

Article Title: Determinants of failure of midazolam monotherapy during continuous deep sedation until death: evidence from a French cohort in a palliative care unit

Article References: Gesson, P., Houizot, C., Coudol, S., Huon, J.-F., Nguyen, O., Prampart‐Fauvet, S., & Evin, A. (2026). Determinants of failure of midazolam monotherapy during continuous deep sedation until death: evidence from a French cohort in a palliative care unit. BMC Medicine. https://doi.org/10.1186/s12916-026-05282-0

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05282-0

Keywords: continuous deep sedation, palliative sedation, midazolam, chlorpromazine, propofol, treatment failure, end-of-life care, polypharmacy, drug interactions, palliative care, refractory symptoms, BMC Medicine

The distinction between resistance and tolerance is central to the argument. Resistant bacteria grow even in the presence of an antibiotic, and the mutations that confer resistance are readily detected because they allow measurable growth under drug pressure. Tolerant bacteria, by contrast, survive antibiotic exposure without necessarily growing, and they do so in ways that can be temporary and hard to distinguish from ordinary physiological stasis. That has historically made tolerance difficult to study: if a phenotype is transient and non-heritable, it leaves little trace in the kinds of experiments that classical microbiology was built to perform. The result has been a research landscape in which resistance mechanisms are well catalogued while tolerance mechanisms remain comparatively obscure.

Kobras’s review contends that this obscurity is no longer tenable, because the tools now exist to interrogate tolerance at the level of the genome. Genome-wide analyses, combined with laboratory evolution experiments and functional genetic screens, have begun to reveal the pathways that allow bacteria to survive antibiotic exposure beyond the familiar model of dormant, metabolically quiescent cells. Laboratory evolution, in which bacterial populations are propagated under antibiotic stress over many generations, allows tolerance-associated variants to accumulate and be identified through whole-genome sequencing of the evolved lineages. Functional screens, including genome-scale knockout and overexpression libraries, allow researchers to ask systematically which genes, when disrupted or amplified, change how long cells survive a drug.

The technical power of these approaches lies in their convergence. Sequencing an evolved population identifies candidate mutations, but it cannot by itself distinguish causal changes from hitchhikers. Functional screens provide the complementary evidence: if a mutation identified by evolution also produces a tolerance phenotype when introduced deliberately, the case for causality strengthens. Combining the two, alongside precise reconstruction of variants in clean genetic backgrounds, allows the genetic architecture of tolerance to be resolved, including cases where multiple loci interact. This is where the concept of epistasis becomes important. The review highlights that the effects of tolerance variants may depend on the presence of other variants, meaning that the same mutation can produce different survival outcomes in different genetic backgrounds. Untangling these interactions is essential if tolerance is to be predicted from sequence data alone.

One of the most consequential shifts described in the review is conceptual. The dominant model of antibiotic tolerance has centred on persisters, a subpopulation of cells that enter a dormant or slow-growing state and thereby evade antibiotics that kill actively growing bacteria. Dormancy remains an important part of the picture, but the genomics-driven work summarised by Kobras reveals survival pathways that operate independently of the dormant-cell model. Bacteria can survive antibiotic exposure through active physiological responses, stress-response programmes and metabolic adjustments that are encoded in the genome and subject to evolutionary modification. Recognising this breadth matters because it expands the set of potential intervention points: if tolerance is not solely a property of dormant cells, then strategies aimed solely at waking persisters up may address only part of the problem.

The clinical stakes are considerable. Treatment failure in bacterial infections is usually attributed to resistance, but tolerance can plausibly explain cases in which patients fail therapy despite isolates that test as susceptible in the laboratory. Standard susceptibility testing measures growth inhibition at fixed drug concentrations, a framework that can miss survival phenotypes expressed over longer time scales or under the fluctuating drug concentrations that occur in treated patients. Tolerant populations also buy time for resistance to emerge, because cells that survive initial exposure constitute the population in which resistance mutations can subsequently arise. The review therefore frames tolerance not only as a direct cause of treatment failure but as a factor that shapes the evolution of resistance itself, an influence that genomic approaches are now positioned to quantify.

That connection between tolerance and resistance evolution is one of the areas the review identifies as ripe for further work. If particular tolerance genotypes accelerate the emergence of resistance, then detecting those genotypes in patient-derived isolates could inform treatment choices before resistance becomes detectable by conventional means. Genome-wide analyses of tolerance offer a route to identifying candidate genes and variants in clinical isolates, extending the insights of laboratory systems to the strains that actually cause disease. Population genomics of clinical collections could, in principle, reveal whether tolerance-associated variants recur in specific lineages, correlate with treatment outcomes, or precede the acquisition of classical resistance determinants.

Kobras is careful, however, about the limits of the current evidence base. Most of what is known about the genetics of tolerance comes from laboratory systems: reference strains, defined media, controlled drug exposures and evolution experiments designed to make phenotypes measurable. Whether the pathways identified under those conditions operate to the same extent, or in the same combinations, in the environments of actual infections is unresolved. Infection sites present bacteria with nutrient limitation, host immune pressures, biofilm structures and antibiotic concentrations that vary over space and time, and any of these factors could modify the contribution of a given tolerance variant. The review presents the translation of laboratory insights into patient-derived isolates as the key outstanding challenge, and as the step that will determine whether genomic tolerance research changes clinical practice.

The potential payoffs of meeting that challenge are spelled out in diagnostic and therapeutic terms. On the diagnostic side, a better understanding of tolerance genetics could support assays that flag tolerant genotypes in isolates, complementing conventional susceptibility testing and giving clinicians a more complete picture of how a pathogen is likely to behave during therapy. On the therapeutic side, the pathways revealed by genome-wide analyses define candidate targets for drugs that would sensitize tolerant bacteria to existing antibiotics, an attractive strategy because it would extend the useful life of drugs already in the clinic rather than requiring entirely new classes of antimicrobials. Combination therapy designed to eliminate tolerant subpopulations could also reduce the reservoir from which resistance emerges, addressing both problems simultaneously.

The review, which is open access and was published on 2 October 2026 after being received on 28 May and accepted on 24 August, arrives at a moment when antimicrobial resistance remains one of the most pressing threats to global health, and when the research community is increasingly aware that resistance alone does not explain every therapeutic failure. By consolidating the genomics-driven transformation of tolerance research, Kobras provides a roadmap for a field moving from phenomenology toward mechanism. The core message is that antibiotic tolerance has a genetics, that this genetics is now tractable with modern tools, and that resolving it, first in the laboratory and then in patient-derived isolates, offers a realistic path toward diagnostics and therapies that account for the full range of bacterial survival strategies. If that programme succeeds, the stubborn gap between a susceptible laboratory test and a failed course of treatment may finally begin to close.

Subject of Research: Genomic analysis of antibiotic tolerance and its genetic basis in bacteria

Article Title: Transforming antibiotic tolerance research through genomics

Article References: Kobras, C. M. (2026). Transforming antibiotic tolerance research through genomics. Genome Biology. https://doi.org/10.1186/s13059-026-04257-x

Image Credits: AI Generated

DOI: 10.1186/s13059-026-04257-x

Keywords: antibiotic tolerance, antimicrobial resistance, bacterial genomics, persistence, epistasis, laboratory evolution, functional genomics, treatment failure, population genomics, bacterial survival, Genome Biology, microbial genetics

The original article, which appeared online on 31 January 2026, described laboratory and animal experiments linking buparlisib, an investigational drug that targets the PI3K signaling pathway, to an unexpected mechanism of treatment resistance. According to the paper, exposure to the compound appeared to drive up levels of eukaryotic elongation factor-2, a molecular machine central to protein synthesis in cells. The authors argued that this upregulation contributed to treatment failure in lung cancer cells, a finding that, if confirmed, would have carried implications for how clinicians combine targeted therapies with other anticancer agents.

That conclusion is now formally withdrawn. In the retraction note, the Editor-in-Chief of the journal stated that the article has been retracted because the tumors presented in Figure 5 of the paper exceed the accepted limits for tumor burden. In animal cancer studies, tumor burden refers to the total volume or mass of tumor tissue carried by an experimental animal. Research ethics frameworks impose upper limits on how large tumors may be allowed to grow, both to protect animal welfare and to ensure that experiments remain within humane boundaries. When tumors surpass those limits, the study is considered to have breached the ethical standards under which animal research is permitted.

The editor’s decision was anchored explicitly in the ARRIVE 2.0 reporting guidelines, a widely adopted framework for the design, conduct and publication of animal research. ARRIVE, which stands for Animal Research: Reporting of In Vivo Experiments, was developed to improve the reproducibility and ethical rigor of studies involving animals. The updated 2.0 version, published in 2020 by Percie du Sert and colleagues in the journal PLoS Biology, sets out detailed expectations for how animal studies should be planned, reported and reviewed, including humane endpoints, welfare monitoring and limits on tumor growth. The retraction note states that, as a result of the excessive tumor burden shown in the paper, the study does not appear to meet the animal ethics standards described in those guidelines.

The retraction is notable for the response of the author team. According to the notice, Xu Li, speaking on behalf of all the authors, stated that they disagree with the retraction. Disagreement between authors and journals is not uncommon in retraction cases, and journals typically publish the authors’ position alongside the retraction notice to make the record transparent. In this instance, the journal has recorded the authors’ objection while still removing the article from the body of credible scientific literature, leaving readers to weigh the dispute for themselves.

The research behind the paper originated from a collaboration spanning several Chinese institutions. The author group included DanPing Wang, ChenLi Su, Xiao Zhang, Wei Mao, XiaoQiang Dai, Li Xu and XiaoMin Si. Affiliations listed for the team include the Department of Oncology at Xianyang Central Hospital in Shaanxi Province, the Clinical Medicine College at Shaanxi University of Chinese Medicine, and a key laboratory at Yangzhou University dedicated to integrated traditional Chinese and Western medicine in the control of senile diseases. DanPing Wang and ChenLi Su were identified as having contributed equally to the work, with correspondence directed to Li Xu and XiaoMin Si.

The scientific premise of the retracted study touched on a question of real clinical importance. Buparlisib is one of a family of PI3K inhibitors developed to disrupt signaling pathways that drive tumor growth and survival. Resistance to such targeted agents remains a major obstacle in oncology, and proposed mechanisms of resistance attract intense interest because they can suggest combination strategies to overcome treatment failure. Eukaryotic elongation factor-2, the molecule implicated in the paper, is an essential component of the translational apparatus that converts messenger RNA into protein, and its regulation has previously been linked to stress responses in cancer cells. A credible demonstration that a PI3K inhibitor induces this factor and thereby contributes to treatment failure would have been a meaningful addition to the resistance literature.

That potential contribution is now in doubt. Retractions serve a specific function in the scientific record: they alert readers that a published article’s findings can no longer be trusted as reliable, whether because of errors, ethical violations or misconduct. In this case, the stated reason is not fraud or data fabrication but a failure of animal ethics compliance, specifically tumor burden beyond accepted limits in the animal experiments underpinning Figure 5. The distinction matters, because it underscores that scientific publishing increasingly treats ethical compliance as inseparable from scientific validity. A result obtained outside accepted welfare boundaries is regarded not merely as ethically problematic but as unfit for the literature.

The case also illustrates the practical reach of the ARRIVE 2.0 guidelines, which have become a benchmark that editors and reviewers consult when evaluating animal studies. By citing the framework directly in the retraction notice, the journal signaled that deviations from those standards are grounds for withdrawal even after publication. For researchers working with tumor-bearing animals, the message is that monitoring tumor size against predefined humane limits is not an administrative formality but a condition of publication. Journals and their editors are prepared to retract work, years of effort and all, when the welfare thresholds written into those guidelines are crossed.

For the field of lung cancer research, the retraction removes one proposed mechanism of resistance to a PI3K-targeted therapy from the pool of evidence that other scientists might build upon. The original article remains accessible as a retracted document, clearly flagged, so that readers who encounter it can understand its status. The episode serves as a reminder that the credibility of cancer biology depends not only on the ingenuity of its hypotheses but on the rigor and ethics of the experiments that support them, and that the machinery of scientific self-correction, though sometimes contested by authors, continues to operate when standards are not met.

Subject of Research: Retraction of a lung cancer study on buparlisib-induced eukaryotic elongation factor-2 expression due to animal ethics violations

Article Title: Retraction Note: Buparlisib induces eukaryotic elongation factor-2 expression to cause treatment failure for lung cancer cells

Article References: Wang, D., Su, C., Zhang, X., Mao, W., Dai, X., Xu, L., & Si, X. (2026). Retraction Note: Buparlisib induces eukaryotic elongation factor-2 expression to cause treatment failure for lung cancer cells. Molecular Genetics and Genomics, 301(1), Article 195. https://doi.org/10.1007/s00438-026-02522-z

Image Credits: AI Generated

DOI: 10.1007/s00438-026-02522-z

Keywords: retraction, buparlisib, eukaryotic elongation factor-2, lung cancer, treatment failure, animal ethics, tumor burden, ARRIVE 2.0, PI3K inhibitor, Molecular Genetics and Genomics, research integrity, targeted therapy