The study was conducted at Ankara Bilkent City Hospital, a tertiary referral center in Türkiye. Eligible patients were adults with histologically confirmed malignancy and a first abscess episode documented by contrast-enhanced cross-sectional imaging and a clinical diagnosis by the treating team. Percutaneous aspiration was performed in 177 patients, or 96.2% of the cohort; the other seven were diagnosed from imaging and clinical records. Culture positivity was not required, and only each patient’s first eligible episode was analyzed. Median age was 62 years, 107 participants, or 58.2%, were men, and 99, or 53.8%, had stage IV malignancy. Colorectal cancer was the most common diagnosis, followed by pancreatic cancer.

Abscesses occurred across multiple anatomical sites. The liver was involved in 73 patients, or 39.7% of the cohort, while 66 patients had non-hepatic intra-abdominal collections, 20 had retroperitoneal abscesses, 11 had thoracic or musculoskeletal collections, 10 had pelvic abscesses and four had abdominal-wall abscesses. Forty-two patients, or 22.8%, had multiple collections. The median recorded abscess diameter was 7.0 cm, with an interquartile range of 4.6–10.0 cm among 176 patients with available measurements. Therapeutic aspiration or drainage was performed in 173 patients, and 52 underwent two or more sessions. Median antibiotic duration was 16 days and median hospital stay was 32 days, although procedure outcomes and the adequacy of source control were not recorded.

An organism was recorded for 124 patients, or 67.4% of the cohort, with all isolates coming from percutaneous aspirates. Two clinicians independently reviewed every culture report and adjudicated possible contamination using the specimen type, organism identity, report wording and clinical record. After adjudication, 108 patients, or 58.7% of the full cohort, had at least one probable pathogen. Among these 108, 85, or 78.7%, had monomicrobial gram-negative growth; 16 had gram-positive pathogens, six had Candida albicans and one had a polymicrobial culture. Escherichia coli was identified in 43 patients and Klebsiella species in 29, including 27 K. pneumoniae isolates. Enterococci accounted for 10 probable pathogens and were proportionally more frequent in retroperitoneal and pelvic collections, while E. coli and K. pneumoniae together represented 74.5% of probable pathogens in hepatic abscesses.

Resistance was frequent among the 85 gram-negative isolates. Phenotypic extended-spectrum beta-lactamase, or ESBL, production was recorded in 38 of 80 tested Enterobacterales isolates, or 47.5%. It occurred in 28 of 43 E. coli isolates and eight of 27 K. pneumoniae isolates. Among the gram-negative pathogens, 55 were multidrug resistant and 14 extensively drug resistant, together accounting for 81.2%, or 69 of 85 isolates. Resistance exceeded 80% for ceftriaxone, ceftazidime, cefepime and amoxicillin–clavulanate. It was 65.7% for piperacillin–tazobactam, 62.3% for ciprofloxacin and 37.3% for ertapenem; resistance to at least one tested carbapenem was 35.2%. Amikacin, colistin and tigecycline retained activity against 84.6%, 86.4% and 100% of isolates tested, respectively.

These susceptibility results were descriptive and did not establish a preferred empirical antibiotic regimen. The researchers did not assess AmpC or carbapenemase production or confirm resistance mechanisms genetically, and the automated ESBL phenotype may misclassify some AmpC- or carbapenemase-producing isolates. The number of deaths also limited comparisons between resistance phenotypes. Moreover, the cohort came from one tertiary referral center, most cases were identified through oncology records and interventional radiology documentation, and patients managed conservatively or surgically may have been incompletely captured. The findings therefore describe resistance patterns in a selected local population rather than the full spectrum of cancer-associated abscesses.

Albumin was the strongest measured correlate of death during the first 90 days. The cohort’s median albumin was 3.0 g/dL, with an interquartile range of 2.7–3.5 g/dL. Values were obtained within the two days preceding the index date, defined as aspiration for 177 patients and first documentation for the seven who were not aspirated. In a Cox model adjusted for age and cancer stage and stratified by single versus multiple abscesses, a 0.5 g/dL higher albumin concentration was associated with a hazard ratio of 0.37 for 90-day mortality (95% CI, 0.26–0.53). The corresponding estimate was 0.14 per 1 g/dL. Stage IV disease and age were not statistically associated with mortality during this period. Model-based standardized 90-day mortality was estimated at 39.4% for albumin of 2.5 g/dL, 17.2% at 3.0 g/dL and 6.8% at 3.5 g/dL. The model-based risk difference between 2.5 and 3.5 g/dL was 32.5 percentage points, with a bootstrap 95% CI of 22.7–40.4.

The albumin association remained in sensitivity analyses that changed the time origin to first imaging or hospital admission, used a 30-day endpoint, excluded selected groups, or separately added variables related to severity, exposure, abscess site and microbiology. Adding intensive-care admission, vasopressor use, mechanical ventilation, C-reactive protein or procalcitonin did not eliminate the association. Albumin is a negative acute-phase reactant influenced by nutrition, tumor burden and systemic inflammation. Because it was measured during an established infection, the result most likely marked acute illness severity as well as nutritional and cancer-related factors rather than baseline host reserve alone. Performance status, formal sepsis scores, antimicrobial timing, source-control delay and detailed tumor burden were unavailable, leaving scope for residual confounding. The exploratory hemoglobin–albumin–lymphocyte–platelet score was not associated with 90-day mortality and discriminated outcomes less well than albumin in this cohort.

Longer follow-up indicated that early and later mortality had different correlates. Survival was 68.4% at 180 days and 54.3% at one year. The albumin association weakened as follow-up extended, with the hazard ratio per 1 g/dL changing from 0.14 during the first 90 days to 0.20 at 180 days, 0.25 at one year and 0.31 over all available follow-up. In a period-specific analysis, the corresponding hazard ratio after day 90 was 0.48. Stage IV disease followed the opposite pattern: its early hazard ratio was 0.70, with a wide confidence interval, while mortality after day 90 was more than twice as high among patients with stage IV disease, with a hazard ratio of 2.65 (95% CI, 1.45–4.84). One-year survival was 48.9% for stage IV patients and 60.9% for those with stage I–III disease. The early stage IV estimate should not be interpreted as protective because selecting patients who survived to abscess diagnosis may have introduced selection bias.

The findings remain provisional because the study was observational, retrospective and confined to one center, with heterogeneous abscess sites and no cancer control group without abscess. Cases were not formally adjudicated by radiologists, and imaging reports used the term “abscess” explicitly in only 35.5% of patients, leaving uncertainty around some collections. Ten patients were censored alive before day 90, antibiotic duration was incompletely documented for 28, and the database did not distinguish culture-negative episodes from those in which culture was never obtained. Death before aspiration could not enter the cohort, making survival estimates from earlier time origins susceptible to immortal time bias. The models were internally bootstrapped but not externally validated. Prospective multicenter studies with standardized imaging review, confirmatory resistance testing, detailed measures of sepsis severity and source control, and cause-specific mortality are needed to determine how broadly the albumin association and resistance patterns apply.

Subject of Research: Clinical, microbiological and prognostic features of imaging-confirmed abscesses in adults with malignancy

Article Title: Clinical and microbiological characteristics of abscesses in patients with malignancy and factors associated with 90-day mortality

Article References: Gencer, H., Kos, F. T., Guven, M., Calıskan, M., Ozberk, U., Bal, O., & Akinci, M. B. (2026). Clinical and microbiological characteristics of abscesses in patients with malignancy and factors associated with 90-day mortality. Frontiers in Cellular and Infection Microbiology, 16, 1969735. https://doi.org/10.3389/fcimb.2026.1969735

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1969735

Keywords: cancer-associated abscesses, 90-day mortality, serum albumin, antimicrobial resistance, gram-negative bacteria, ESBL production, multidrug resistance, percutaneous aspiration, retrospective cohort, Clinical, microbiological, characteristics

The PRECIOUS trial enrolled 1,493 patients aged 66 or older with acute ischaemic stroke or intracerebral haemorrhage and a National Institutes of Health Stroke Scale score of at least 6, across 68 sites in nine European countries: Estonia, Germany, Greece, Hungary, Italy, the Netherlands, Norway, Poland, and the United Kingdom. Patients were randomised within 24 hours of symptom onset to receive preventive metoclopramide, ceftriaxone, paracetamol, or combinations thereof. After excluding patients lost to follow-up, those who withdrew consent, and those ultimately diagnosed with a stroke mimic, 1,451 patients remained for this analysis. The mean age was 79.7 years, 50.6 percent were male, and 86.1 percent had ischaemic stroke while 13.9 percent had intracerebral haemorrhage.

Researchers prospectively recorded whether any treatment restriction was installed in the first 24 hours after randomisation and during the first seven days, using a structured electronic case report form. Options included do-not-resuscitate orders, do-not-intubate directives, withholding other life-prolonging treatments, withholding food or fluids, and palliation with morphine or benzodiazepines. Overall, 276 patients (19.0 percent) had a restriction within the first 24 hours, and 360 (24.8 percent) within the first week. Do-not-resuscitate orders were the most common type. Restrictions were more frequent in patients with intracerebral haemorrhage (25.4 percent) than ischaemic stroke (18.0 percent), and affected patients were older, more often female, had greater stroke severity, and more pre-stroke dependency.

The variation across countries was striking. Early treatment restrictions ranged from zero percent in Estonia, Poland, and Hungary, to just 0.4 percent in Italy and 2.1 percent in Greece, up to 17.8 percent in Germany, 19.4 percent in the United Kingdom, and 48.1 percent in the Netherlands. This means nearly one in two Dutch patients in the trial had a restriction placed within the first day, while patients in Eastern and Southern European countries almost never did. Such wide practice variation, despite all countries following the same European Stroke Organisation guidelines for stroke management, points to deep cultural differences in end-of-life decision-making rather than differences in clinical standards.

To test whether early restrictions actually cause death, the researchers used two complementary approaches. First, they applied inverse probability of treatment weighting, a propensity score method that reweights patients to balance observed confounders such as age, sex, diabetes, atrial fibrillation, pre-stroke dependency, stroke severity, stroke type, time to randomisation, country, and trial treatment allocation. After adjustment, early restrictions remained strongly associated with death within 90 days (39.1 percent versus 15.5 percent; weighted odds ratio 3.39, 95% CI 2.71–4.25), death within seven days (16.7 percent versus 6.2 percent; weighted odds ratio 3.44), and worse functional outcome on the modified Rankin Scale at 90 days. E-value sensitivity analyses indicated that only very strong unmeasured confounders could fully explain these associations.

Second, the team exploited the natural experiment created by cross-country practice variation. They divided patients into those enrolled in countries where early restrictions were common (the Netherlands, Germany, Norway, and the United Kingdom; 914 patients) and those in countries where they were rare (Greece, Estonia, Hungary, Italy, and Poland; 537 patients). If restrictions truly shortened lives, death rates should be higher where they are used more. Yet 90-day mortality was comparable between the two groups: 21.3 percent in the restriction group versus 23.5 percent in the no-restriction group, with an adjusted odds ratio of 1.39 (95% CI 0.81–2.39). Similar null findings emerged for ischaemic stroke and intracerebral haemorrhage subgroups, and for functional outcome and independence in activities of daily living.

The authors argue that this pattern is most consistent with confounding by indication: restrictions are placed on patients already expected to do poorly, so the association with death reflects prognosis rather than harm. If restrictions primarily prevented prolonged survival with severe disability, one might expect fewer patients surviving in a minimally conscious or severely disabled state (modified Rankin Scale score of 5) in countries that restrict more often, but no such pattern was observed. Hospital-level funnel plots showed no consistent relationship between expected and observed restriction rates, further suggesting institutional variation rather than a systematic causal effect.

These findings challenge current guideline recommendations. European Stroke Organisation guidance on intracerebral haemorrhage advises against general policies of do-not-resuscitate orders within the first 24 hours, and American Heart Association guidelines suggest postponing new such orders until at least the second full day of hospitalisation. Care-bundle trial protocols, such as the ongoing I-CATCHER study, embed avoidance of early restrictions as a quality measure. The PRECIOUS analysis suggests such recommendations may be premature, as they are based on retrospective observational data with high risk of residual confounding, and the new evidence indicates restrictions are more a marker of poor prognosis than a cause of it.

Limitations remain. This was a post hoc, non-randomised analysis, so unmeasured confounding cannot be excluded despite the dual approach. Data on restrictions and care were only available for the first week and at 90 days, with no information on later decisions or causes of death. Trial eligibility criteria excluded patients under 66, those with severe pre-stroke disability, and those with imminent death at screening, so the cohort may underrepresent the most severely affected patients in whom restrictions are most common. Between-country comparisons also lack detailed data on healthcare system differences, and it remains unknown whether patients without restrictions received more aggressive care overall.

Nevertheless, the study offers a sobering message for clinicians, guideline developers, and families navigating the hardest decisions in acute stroke care. Rather than treating early treatment restrictions as inherently harmful interventions to be banned, the evidence suggests they should be understood as signals of severity and as expressions of culturally shaped attitudes toward end-of-life care. Whether restricting or continuing aggressive treatment changes outcomes in severely affected patients remains an open question, one that only carefully designed prospective research, sensitive to cultural context and patient values, can ultimately answer.

Subject of Research: Treatment restrictions after acute stroke and their association with 90-day mortality

Article Title: Treatment restrictions after acute stroke across nine European countries and their association with 90-day mortality: a post hoc analysis of the PRECIOUS randomised controlled trial

Article References: Sluis, W. M., van Amsterdam, W. A., de Jonge, J. C., Reinink, H., Ntaios, G., Kõrv, J., Aamodt, A. H., Thomalla, G., Csiba, L., Ciccone, A., Kurkowska-Jastrzebska, I., Macleod, M. R., Bath, P. M., van der Worp, H. B., Reinink, H., de Jonge, J. C., Bart van der Worp, H., Bath, P. M., van de Beek, D., … Thomalla, G. (2026). Treatment restrictions after acute stroke across nine European countries and their association with 90-day mortality: a post hoc analysis of the PRECIOUS randomised controlled trial. The Lancet Regional Health – Europe, 71, Article 101894. https://doi.org/10.1016/j.lanepe.2026.101894

Image Credits: AI Generated

DOI: 10.1016/j.lanepe.2026.101894

Keywords: stroke, treatment restrictions, do-not-resuscitate, PRECIOUS trial, intracerebral haemorrhage, ischaemic stroke, 90-day mortality, end-of-life care, propensity score, confounding, European practice variation, guidelines

The study included 132 patients treated between 2021 and 2024, all of whom had at least two brain metastases and underwent microsurgical resection of one or more selected lesions. In this population, surgery is rarely curative in the conventional sense. Surgeons operate either to relieve symptoms caused by a mass pressing on critical brain structures, or to obtain tissue for histopathologic diagnosis when the primary cancer or its molecular profile remains uncertain. Because most of the tumor burden remains elsewhere in the brain, the operation is a calculated intervention whose value depends entirely on whether the patient lives long enough, and recovers well enough, to benefit from it.

The headline finding is stark: the ninety-day mortality rate was 28 percent, with 37 of the 132 patients dying within three months of surgery. That figure alone reframes the conversation around surgery for multiple metastases, a field where much of the historical literature has focused on carefully selected single-lesion patients. The median age of the cohort was 63 years, roughly 41 percent were women, and about two-thirds of patients had between two and four metastases while nearly a third had five or more. The median resected tumor volume was 15.4 cubic centimeters, quantified through volumetric assessment of magnetic resonance imaging rather than the crude diameter measurements used in many older studies.

What elevates the work beyond a simple outcome audit is its systematic search for preoperative predictors. Four factors emerged as significantly associated with increased ninety-day mortality: advanced age, a lower preoperative Karnofsky Performance Status, the absence of adjuvant treatment after surgery, and a higher score on the five-item modified frailty index. Each of these captures a different dimension of physiological reserve. Age reflects cumulative biological wear; the Karnofsky score, a scale from 0 to 100 that grades a patient’s functional independence, measures how much the disease has already eroded daily functioning; the frailty index aggregates comorbidities such as diabetes, cardiac disease and functional dependence into a compact vulnerability score; and adjuvant therapy, whether radiotherapy or systemic treatment, addresses the residual disease that surgery cannot touch.

The statistical relationships were strong. Age and lack of adjuvant treatment each reached p-values below 0.001, meaning the probability that these associations arose by chance is vanishingly small. The frailty index was associated with mortality at p = 0.008, and Karnofsky status at p = 0.048. Notably, the sheer number of metastases and the volume of the resected tumor did not predict early death. This is a counterintuitive and clinically important result: it suggests that the decision to operate should hinge less on how many lesions a patient carries, and more on who the patient is physiologically when they arrive in the operating room.

Survival was not the only endpoint. The researchers also examined where patients went after discharge, how long they stayed in the hospital, and whether they were allocated to adjuvant treatment. Here, two variables dominated. Patients with lower preoperative Karnofsky scores and higher frailty index scores were significantly more likely to be discharged to institutional care rather than to their own homes, with both associations reaching p-values below 0.001. For patients and families weighing surgery against symptom-directed care, this is a concrete and often underappreciated consideration: the operation may relieve pressure and clarify diagnosis, but in frail or functionally compromised patients it may also mark the transition from independent living to a nursing facility.

Length of postoperative hospital stay told a similar story through correlation analysis. Stay duration correlated inversely with Karnofsky status, with a Spearman rank correlation coefficient of rs = -0.29 and p = 0.001, meaning weaker patients stayed longer. It correlated positively with the frailty index at rs = 0.26 and p = 0.003. Once again, age, tumor volume and the number of lesions showed no significant relationship with recovery time. The pattern is consistent and mechanistically plausible: functional reserve and frailty govern how quickly a brain recovers from surgery, how well wounds heal, and how quickly complications such as confusion, weakness or infection can be overcome, whereas the anatomical extent of metastatic disease matters less for the perioperative trajectory.

The findings arrive at a moment when the treatment landscape for brain metastases is shifting rapidly. Stereotactic radiosurgery, which delivers precisely focused radiation to multiple lesions in a single session without opening the skull, has become the default for many patients with limited metastatic burden, supported by large prospective studies such as the JLGK0901 trial. Meanwhile, targeted therapies and immunotherapies increasingly penetrate the blood-brain barrier, allowing systemic treatment to control intracranial disease in ways that were unimaginable two decades ago. Surgery has retained a defined but narrower role: large symptomatic lesions threatening herniation or severe neurological deficit, and diagnostic uncertainty that blocks access to molecularly matched treatment. Within that narrowed role, knowing who benefits and who is harmed becomes paramount, which is precisely the gap this study addresses.

The authors are careful about the implications. They do not argue that frail, elderly or functionally impaired patients should be denied surgery outright; rather, they state that resection for symptom relief or tissue diagnosis in patients with multiple metastases may be performed with these precautions in mind. In practice, that means the four risk factors should be part of the preoperative conversation. A 78-year-old with a Karnofsky score of 50, a frailty index of three out of five, and no realistic prospect of adjuvant therapy carries a fundamentally different risk profile from a 58-year-old who is functionally independent, robust, and eligible for radiosurgery or targeted drugs afterward. Quantifying that difference before consent is obtained is the essence of risk stratification, and it converts a vague warning about being a poor candidate into a structured, evidence-based assessment.

There are limits worth keeping in view. The study is retrospective and bicentric, drawing on two academic centers in Germany and Austria, so selection biases in who was offered surgery cannot be fully excluded, and the ninety-day window captures early perioperative harm but not longer-term survival or quality of life. The frailty index, while practical, is a coarse instrument compared with comprehensive geriatric assessment. Still, the message for oncology teams is actionable and timely: in the era of powerful non-surgical options for multiple brain metastases, the bar for operative intervention is set not by the tumor map alone, but by age, frailty, functional status and the availability of subsequent therapy. Patients with multiple brain metastases and their families should expect those four factors to be weighed explicitly, and documented, before anyone is wheeled into the operating theater.

Subject of Research: Risk stratification for surgical outcomes in patients with multiple brain metastases

Article Title: Risk stratification for unfavorable outcome after resection of selected brain metastases in patients with multiple brain metastases

Article References: Eibl, T., Mercea, P. A., Liebert, A., Ritter, L., Rossmann, J., Rezai, A., Moskwa, M., Hurtig, H., Stadler, C., Wunderlich, S., Feldheim, J., Liman, J., Schwartz, C., Griessenauer, C. J., & Schebesch, K.-M. (2026). Risk stratification for unfavorable outcome after resection of selected brain metastases in patients with multiple brain metastases. Journal of Neuro-Oncology, 180(1), Article 9. https://doi.org/10.1007/s11060-026-05819-6

Image Credits: AI Generated

DOI: 10.1007/s11060-026-05819-6

Keywords: brain metastases, neurosurgery, risk stratification, Karnofsky Performance Status, frailty index, 90-day mortality, tumor resection, adjuvant therapy, stereotactic radiosurgery, Journal of Neuro-Oncology, prognostic factors, cancer surgery

Albumin is not simply a structural protein. It is the principal determinant of oncotic pressure, the osmotic force that keeps fluid inside blood vessels rather than allowing it to leak into surrounding tissues. It also serves as a carrier molecule for hormones, fatty acids, drugs, and bilirubin, and it has antioxidant and anti-inflammatory properties that have intrigued intensive care researchers for decades. When a preterm infant’s albumin level drops after major surgery, a logical instinct follows: replace the missing protein and the body’s fluid balance and healing capacity should improve. That instinct has made albumin infusion one of the most commonly used blood products in neonatal units worldwide, despite a long-running and often contentious debate about whether it actually improves outcomes.

The controversy is not new. A landmark randomized trial published in the New England Journal of Medicine in 2004 compared albumin with saline for fluid resuscitation in critically ill adults and found no survival advantage for the protein solution. A Cochrane review examining albumin infusion specifically for low serum albumin in preterm newborns found insufficient evidence of benefit. Yet clinical practice has been slow to change, particularly in the postoperative setting, where hypoalbuminemia is common and appears to correlate with poor wound healing, edema, and mortality. The problem, as researchers increasingly recognize, is that low albumin may be a marker of illness severity rather than a cause of it, and correcting the number on a laboratory report may not correct the underlying disease.

Into this uncertainty steps a study published in Pediatric Research by Xiaofei Ma, Xiaolan Li, Yongjiang Jiang, Yunzhi Liu, and Yueju Cai of Guangzhou Women and Children’s Medical Center and its Liuzhou hospital. The team conducted a two-center retrospective cohort study of 102 preterm infants who underwent intestinal resection in China between 2018 and 2024. Their question was deceptively simple: does the duration of albumin therapy after surgery make any difference to survival? Infants who received albumin for five days or fewer were classified as the short-course group, comprising 69 babies, while the 33 infants treated for more than five days formed the long-course group. The primary outcome was all-cause mortality within 90 days of the operation.

The raw numbers were striking. Crude 90-day mortality was 15.2 percent in the long-course group compared with 4.3 percent in the short-course group, a numerical difference that would immediately catch the eye of any neonatologist. But the investigators, adhering to rigorous epidemiological practice, did not stop at the crude comparison. They recognized that the two groups were not randomly assigned. Instead, treating physicians decided how long to continue albumin based on each infant’s clinical condition, meaning that babies who stayed sicker for longer were more likely to receive prolonged therapy. This phenomenon, known as confounding by indication, is the central challenge of all observational treatment research, and it means the raw mortality gap cannot be interpreted at face value.

To address this, the researchers employed multivariable regression analyses in which treatment duration was modeled as a continuous variable, adjusting for measured baseline characteristics. In this adjusted model, each additional day of albumin therapy was associated with an odds ratio for mortality of 1.36, with a 95 percent confidence interval of 0.98 to 1.97 and a P value of 0.063. In plain terms, the point estimate suggested a possible harm signal, with longer courses trending toward higher mortality, but the confidence interval crossed the threshold of no effect and the result did not reach conventional statistical significance. The team also examined secondary outcomes and repeated their analyses using alternative duration thresholds, finding no statistically significant differences in any of them. Notably, the measured baseline characteristics of the two groups showed no statistically significant differences, which lends some credibility to the comparisons even though unmeasured factors may still differ.

The authors are careful, appropriately so, about what these findings do and do not prove. In their conclusions and impact statements, they emphasize that no statistically significant differences in mortality or major postoperative outcomes were detected between the treatment groups, but they explicitly warn that clinician-determined treatment duration and residual confounding by indication preclude any conclusion about equivalence or about the clinical benefit of prolonged therapy. This is a crucial distinction that is often lost when observational studies circulate on social media and in clinical discussion groups. Absence of statistical significance in a small, confounded cohort is not evidence that two treatments are the same. It is evidence that this particular study, with its 102 patients and its non-randomized design, cannot answer the question definitively.

Why does this matter so much for preterm infants specifically? Neonates, especially those born very early, have physiology that differs profoundly from adults. Their capillary permeability is higher, their plasma volumes are smaller, their livers synthesize albumin at lower rates, and their fluid management windows are extraordinarily narrow. Restricted versus liberal water intake has itself been the subject of decades of Cochrane evidence synthesis in this population, underscoring how sensitive preterm babies are to fluid shifts. Albumin infusions in such patients can expand intravascular volume rapidly, and concerns about volume overload, patent ductus arteriosus, and pulmonary edema are ever-present in the neonatal intensive care unit. Meanwhile, albumin is an expensive, donor-derived blood product, and stewardship of limited resources is a legitimate consideration alongside clinical efficacy. Studies of albumin use in adult and pediatric intensive care have documented substantial practice variation and cost implications, suggesting that clarifying the appropriate duration of therapy could have both clinical and economic consequences.

The study also highlights a broader methodological lesson for neonatology. Many routine practices in newborn intensive care, from fluid strategies to transfusion thresholds to nutritional supplementation, have never been tested in randomized trials, and observational studies like this one serve as essential signals that such trials are needed. The researchers followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines and conducted sensitivity and subgroup analyses to probe the robustness of their findings, reflecting a maturing standard of rigor in retrospective neonatal research. They also note that their data, while not publicly shareable due to ethical and privacy restrictions, are available to qualified researchers upon reasonable request and institutional approval, an approach that supports future collaborative analyses and meta-analyses.

For clinicians at the bedside, the practical takeaway is one of humility rather than a new protocol. The study does not demonstrate that prolonged albumin therapy harms preterm infants after intestinal resection, nor does it show that short courses are safe and sufficient. What it does show is that the common practice of extending albumin infusions beyond five days in the sickest babies is not supported by clear evidence of benefit, and that a possible signal of harm deserves serious attention. The natural next step, which the authors’ exploratory findings implicitly call for, is a properly powered randomized controlled trial comparing short-course and long-course albumin therapy in this vulnerable population, with mortality and major morbidities such as bronchopulmonary dysplasia, wound complications, and length of intensive care stay as prespecified outcomes. Until such evidence arrives, the decision to continue albumin day after day in a recovering preterm gut will remain one of neonatal medicine’s quiet uncertainties, made infant by infant, at the bedside, in the absence of definitive data.

Subject of Research: Duration of albumin infusion therapy and postoperative outcomes in preterm infants undergoing intestinal resection

Article Title: Albumin therapy in preterm infants after intestinal resection: a two-center retrospective cohort study

Article References: Ma, X., Li, X., Jiang, Y., Liu, Y., & Cai, Y. (2026). Albumin therapy in preterm infants after intestinal resection: a two-center retrospective cohort study. Pediatric Research. https://doi.org/10.1038/s41390-026-05504-4

Image Credits: AI Generated

DOI: 10.1038/s41390-026-05504-4

Keywords: albumin therapy, preterm infants, intestinal resection, neonatal intensive care, necrotizing enterocolitis, retrospective cohort study, 90-day mortality, confounding by indication, hypoalbuminemia, Pediatric Research, fluid management, neonatal surgery

The study drew on the CandidICU cohort, a retrospective collection of all adults hospitalized between January 1, 2015 and January 1, 2023 in participating intensive care units who had at least one positive blood culture for Candida within seven days before admission or during the stay. Of 492 patients initially enrolled, the investigators narrowed the analytic population to 237 individuals. The exclusions were deliberate and methodologically consequential: 122 patients who died within the first five days were removed because they could never have been eligible for either treatment strategy in the emulated trial, 76 patients treated for more than 21 days were excluded because such prolonged courses usually signal complicated infections such as endocarditis or deep-seated dissemination, 41 patients treated for fewer than five days were dropped because such brief exposure likely reflected discontinuation for unrelated reasons, and 16 patients lacked complete data. What remained was a focused population of patients with uncomplicated candidemia who survived long enough to plausibly receive either a short or a long course.

The population was severely ill. Median age was 62 years, roughly two-thirds were men, and the median SAPS II severity score was 53, a figure that places these patients among the sickest in intensive care. Candida albicans caused about two-thirds of infections, with Candida glabrata the most common non-albicans species, and echinocandins served as initial therapy in more than 84 percent of cases. Notably, patients excluded for early death were markedly sicker than those included, with higher SAPS II scores and greater rates of cirrhosis, renal replacement therapy, and acute respiratory distress syndrome—a comparison that supported the appropriateness of the five-day exclusion while clarifying that the findings apply only to patients who survive the initial phase of infection.

The central methodological challenge in any observational study of treatment duration is that duration is not known at the moment treatment begins. Patients who receive short courses may differ systematically from those treated longer, and clinicians often stop therapy early precisely because a patient is deteriorating, or conversely because the patient is improving rapidly. Worse, so-called immortal time bias arises because a patient must survive long enough to accumulate days of therapy. To confront these distortions, the researchers turned to target trial emulation, a framework in which the analysis is explicitly designed to mimic a randomized trial. The key instrument was a technique called cloning, censoring, and weighting: every patient was duplicated into two virtual clones, one assigned to a short-course strategy and one to a long-course strategy, and each clone was censored at the moment its observed treatment deviated from the assigned strategy, using a 10-day grace period for classification.

On top of the cloning structure, the team applied inverse probability of treatment weighting based on a propensity score model that incorporated 22 baseline covariates, including age, sex, comorbidities such as cirrhosis and diabetes, immunosuppression, neutropenia, mechanical ventilation, the Candida species involved, echinocandin use, and suspected source of infection. The model achieved a C-statistic of 0.704, and weighting brought covariate balance within acceptable limits, reducing the standardized mean difference for Candida albicans status from 0.417 to 0.108. Sensitivity analyses varied the grace period to 7 and 14 days, adjusted for treatment center, and applied inverse probability of censoring weights to address potential informative censoring, all yielding results consistent with the primary analysis.

The raw numbers told a dramatically different story from the adjusted ones. Among the 237 patients, 48 received short-course therapy and 189 received long courses, and crude 90-day mortality was 79.2 percent in the short-course group versus 39.7 percent in the long-course group, a difference so stark it would seem to condemn abbreviated therapy outright. But after cloning and weighting, that apparent chasm collapsed: the weighted hazard ratio for 90-day mortality was 1.14, with a 95 percent confidence interval of 0.82 to 1.58 and a p-value of 0.449—statistically indistinguishable from no effect. The authors argue that the crude difference was almost certainly driven by confounding by indication and selection effects rather than by any causal harm of shorter treatment, and they present the attenuation itself as a cautionary lesson about the magnitude of bias in conventional observational analyses of antimicrobial duration.

Secondary outcomes offered additional nuance. In weighted analyses, patients who received shorter therapy had significantly shorter hospital stays, a median of 31 days versus 46 days, and shorter intensive care unit stays, though the authors urge caution because early death shortens length of stay through competing risk. No meaningful differences emerged in ventilator duration, renal replacement therapy, vasopressor use, ventilator-associated pneumonia, or recurrent fungemia, and no cases of second fungemia occurred in the short-course group. Restricted mean survival time analyses hinted at a possible survival advantage for longer therapy at 30 days, with a p-value of 0.094, but the difference was not significant and faded at later time points.

The authors are emphatic that these findings should not change clinical practice. The confidence interval is wide, compatible with as much as a 58 percent relative increase in mortality from short courses, and the study was never designed as a formal non-inferiority trial. Residual confounding from unmeasured factors—time to blood culture clearance, microbiologic response, adequacy of source control, and the clinical reasons behind a clinician’s decision to stop therapy—cannot be excluded, and the modest residual imbalances in solid tumor status and SAPS II after weighting leave open the possibility of residual bias. With only 48 patients in the short-course group, the study had 69 percent power to detect a 25 percent absolute mortality difference, and the extreme censoring inherent to the cloning design limits the reliability of survival estimates in that arm beyond the grace period.

Nevertheless, the study fills a genuine evidence gap. Two randomized trials comparing 7 versus 14 days of therapy are registered but have not begun enrolling patients, and both exclude the sickest individuals—those in shock or requiring mechanical ventilation—precisely the population this French analysis addresses. From a stewardship perspective, the results are biologically plausible: prolonged antifungal exposure carries risks of hepatotoxicity, drug-drug interactions, and selection of resistant isolates, including echinocandin-resistant strains emerging after relatively brief exposure. The authors suggest that distinguishing uncomplicated from complicated candidemia may help clinicians and antimicrobial stewardship programs weigh abbreviated courses in selected patients who survive the early phase and show clear clinical improvement, while careful screening for ocular, cardiac, and deep-seated complications remains essential before any decision to shorten therapy. Until randomized trials specifically designed for intensive care populations deliver definitive answers, this target trial emulation offers the most rigorous glimpse yet into a question that has governed antifungal prescribing for decades without ever having been properly tested.

Subject of Research: Optimal duration of antifungal therapy for critically ill adults with candidemia in the intensive care unit

Article Title: Association of antifungal treatment duration with 90-day mortality in critically ill adults with candidemia: a multicenter target trial emulation in France

Article References: Reizine, F., Henry, J., Desmedt, L., Camus, C., Tadié, J. M., Coirier, V., Arrive, F., Burban, E., Eustache, G., Belleville, T., Marc, A., Malherbe, J., Bouju, P., Jaubert, P., Lesieur, O., Leclerc, M., Fillâtre, P., Frérou, A., Prével, R., … Gangneux, J.-P. (2026). Association of antifungal treatment duration with 90-day mortality in critically ill adults with candidemia: a multicenter target trial emulation in France. eClinicalMedicine, 99, Article 104178. https://doi.org/10.1016/j.eclinm.2026.104178

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104178

Keywords: candidemia, antifungal therapy duration, intensive care unit, target trial emulation, cloning censoring weighting, 90-day mortality, Candida albicans, echinocandins, antimicrobial stewardship, immortal time bias, propensity score, eClinicalMedicine