The investigators retrospectively collected real-world clinical data from 3,002 patients treated with intravenous voriconazole between May 2018 and January 2025 at Quanzhou First Hospital. After applying predefined eligibility criteria, the model-development analysis included 96 patients and 240 trough concentration measurements. The cohort was predominantly male, with 74 of 96 patients male, and had a median age of 64 years, weight of 55 kg and creatinine clearance of 85.16 mL/min.

Patients initially received weight-based intravenous dosing, consisting of a 6 mg/kg loading dose every 12 hours followed by 4 mg/kg every 12 hours for maintenance. Trough samples were collected after steady state was reached, generally before the fifth dose. The researchers used nonlinear mixed-effects modeling with NONMEM to develop the population model. Because only trough concentrations were available, a two-compartment model produced substantial fitting errors, so the investigators selected a one-compartment model with linear elimination. An exponential model represented between-subject variability, while a combined additive and proportional model represented residual variability.

The final model estimated typical voriconazole clearance at 4.2 L/h and volume of distribution at 75.9 L. Stepwise covariate modeling identified creatinine clearance and C-reactive protein as significant predictors of clearance and concomitant omeprazole use as a significant predictor of volume of distribution. At a constant creatinine clearance, the model indicated that a twofold rise in C-reactive protein was associated with approximately 10% lower clearance. Body weight and several conventional markers of hepatic function, including ALT, AST, alkaline phosphatase and bilirubin, did not remain significant in the final model.

Model performance was evaluated with goodness-of-fit and visual predictive checks. In bootstrap resampling, 9,644 of 10,000 datasets converged, giving a convergence rate of 96.44%; parameter estimates from the final model were close to bootstrap medians and within their 95% confidence intervals. Visual predictive checks showed that the median observations fell within the 90% prediction interval and that the interval covered most observations. These results supported internal model stability, although the authors noted that sparse sampling and the exclusive availability of trough data limited precise estimation of volume of distribution.

The researchers then used Monte Carlo simulation to derive a trough-based dose-adjustment protocol. The simulation generated a virtual population of 4,000 subjects, with covariate distributions resampled from the study cohort, and evaluated possible maintenance-dose changes in 5-percentage-point increments, rounding the selected changes to achievable 50 mg increments. The modeled target was a steady-state trough concentration of 1.0–5.5 mg/L after up to three adjustments. For concentrations below 0.5 mg/L, the protocol proposed a 95% maintenance-dose increase; for concentrations of 0.5 to less than 1.0 mg/L, it proposed a 65% increase.

For concentrations above the target range, the simulation proposed reducing the maintenance dose by 35% when the trough was above 5.5 to 7.5 mg/L, by 55% when it was above 7.5 to 10 mg/L, and by 75% when it exceeded 10 mg/L. The simulated probability of reaching the target range after up to three adjustments was 89.2% for troughs below 0.5 mg/L, 90.3% for troughs from 0.5 to less than 1.0 mg/L, 97.9% for troughs above 5.5 to 7.5 mg/L, 98.9% for troughs above 7.5 to 10 mg/L and 99.0% for troughs above 10 mg/L. These were model-based predictions rather than observed outcomes in every dosing category.

Clinical validation included 67 hospitalized patients who received intravenous voriconazole and underwent therapeutic drug monitoring between January 2025 and January 2026. Clinical pharmacists calculated model-informed adjustments, physicians implemented the revised regimens and concentrations were measured again at steady state. Under the standard prescribing-information regimen, 31 of 67 patients, or 46.3%, reached the target range. After up to three model-guided adjustments, 61 patients, or 91.04%, had steady-state plasma concentrations within the range. Six patients remained outside the target: three were transferred to other hospitals because their underlying illnesses worsened, preventing further monitoring, and three remained below the range despite adjustment and were switched to alternative antimicrobial agents.

The findings are limited by the retrospective single-center design, a predominantly older population with limited demographic and clinical diversity, sparse trough-only sampling, and inclusion limited to intravenous dosing. Child-Pugh scores could not be incorporated because routine coagulation monitoring was unavailable, and CYP2C19 genotyping was not performed, preventing quantification of that genetic contribution. The validation was also conducted within the same hospital, and the safety profile of the model-guided regimen was not evaluated. Accordingly, the observed increase in target-range attainment supports the regimen within this retrospective setting but does not establish broader effectiveness or safety across other populations and clinical settings.

Subject of Research: Population pharmacokinetic modeling and clinical validation of trough-based intravenous voriconazole dose adjustment

Article Title: Population pharmacokinetic modeling of voriconazole based on real-world data for dosing regimen optimization

Article References: Jiang, L., & Lin, Z. (2026). Population pharmacokinetic modeling of voriconazole based on real-world data for dosing regimen optimization. Frontiers in Pharmacology, 17, Article 1979373. https://doi.org/10.3389/fphar.2026.1979373

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1979373

Keywords: voriconazole, population pharmacokinetics, therapeutic drug monitoring, dose adjustment, Monte Carlo simulation, intravenous dosing, real-world data, clinical validation, Population, pharmacokinetic, modeling, based

The study, conducted by a team from Shenzhen University Medical School and Shenzhen Kangning Hospital and published in BMC Geriatrics, took a deliberately practical approach. Rather than recruiting a carefully curated research cohort, the investigators turned to the everyday records of 118 hospitalized patients with clinically diagnosed Alzheimer’s disease. Their goal was to capture the heterogeneity that clinicians actually confront at the bedside, where a patient’s ability to feed themselves or walk to the toilet may tell a very different story from the agitation, apathy, or hallucinations documented in nursing observations.

To measure functional ability, the team used the Barthel Index, a widely employed clinical instrument that scores a patient’s independence across fundamental activities of daily living, including feeding, bathing, grooming, dressing, continence, toileting, transferring, mobility, and stair climbing. For behavioral symptoms, they used the Nurses’ Observation Scale for Inpatient Evaluation, known as NOSIE, a rating tool completed by nursing staff that captures the spectrum of ward-observed behaviors, from irritability and psychotic symptoms to social withdrawal and psychomotor slowing. Because these assessments are embedded in routine inpatient care, the resulting dataset reflects the textured, day-to-day clinical picture rather than a snapshot from a standardized research visit.

The first striking finding was statistical. When the researchers correlated overall functional scores with overall behavioral symptom burden using Spearman rank correlation, they found only a weak inverse relationship, with a correlation coefficient of −0.30. In plain terms, knowing how functionally impaired a patient is provides only limited information about how severe their behavioral symptoms are. Some patients with severe physical dependence show relatively modest behavioral disturbance, while others with middling function experience pronounced psychiatric symptoms. This statistical looseness is exactly what clinicians have long suspected: functional decline and behavioral symptomatology in Alzheimer’s disease are related, but they are not two faces of a single underlying dial.

That weak correlation set the stage for the study’s central move. Instead of treating Alzheimer’s patients as interchangeable points on a severity scale, the investigators applied k-means clustering, an unsupervised machine learning algorithm that partitions data points into groups based on similarity across multiple measured dimensions simultaneously. The algorithm does not know anything about diagnosis or prognosis; it simply finds natural groupings in the numbers. When fed the combined functional and behavioral data from the 118 patients, it returned three clusters, each representing a coherent clinical persona that the researchers labeled exploratory functional–behavioral profiles.

Profile 1 was defined by severe functional impairment accompanied by increased psychomotor retardation. Patients in this group were markedly dependent in basic self-care, and their behavioral signature leaned toward slowing, withdrawal, and reduced engagement rather than outward agitation. Profile 2, by contrast, represented the relatively favorable end of the spectrum: these patients retained comparatively preserved functional ability and carried a lower overall behavioral symptom burden. Profile 3 occupied an intermediate position on function but stood apart behaviorally, with more prominent psychotic symptoms such as hallucinations and delusions. In other words, the study found that a patient’s physical disability and their psychiatric symptom profile can dissociate, and that this dissociation creates recognizable, recurring clinical archetypes on the ward.

The differences between profiles were not confined to headline scores. The researchers observed significant differences across specific functional domains, including feeding, toileting, and walking, suggesting that the profiles capture granular differences in how independence erodes, not just how much. Behavioral dimensions also separated the groups meaningfully. What did not separate them was equally interesting: demographic characteristics such as age and sex, and comorbid medical conditions, showed no significant differences between the clusters. The profiles were not simply artifacts of who was older or sicker in a general medical sense; they reflected genuine variation in the neuropsychiatric expression of Alzheimer’s disease itself.

The implications for care could be substantial. Ward teams currently tailor much of Alzheimer’s care around global severity staging, which lumps together functional and cognitive decline. A three-way functional–behavioral typology invites a different logic. A patient in the severe-dependency, psychomotor-retardation group may need fundamentally different nursing attention, rehabilitation goals, and environmental design than a patient whose main challenge is managing psychotic symptoms while retaining considerable physical function. Behavioral symptoms such as agitation and psychosis are among the most burdensome aspects of dementia care, driving caregiver distress, hospitalization, and pharmacological treatment with its attendant risks. Grouping patients by combined profiles could help staff anticipate problems, communicate more precisely during handovers, and design individualized care plans that address the specific symptom cluster a patient actually carries.

The study’s method also carries broader lessons for how real-world clinical data can be mined. Unsupervised clustering does not require pre-existing labels or hypotheses about disease subtypes; it lets the data reveal structure. Applied to routinely collected instruments like the Barthel Index and NOSIE, it can extract clinically interpretable groupings from information that hospitals already generate, without additional testing burden on patients. This fits a growing movement in geriatric medicine toward data-driven phenotyping, in which statistical patterns in electronic records are used to organize patient heterogeneity that traditional diagnostic categories obscure. For a disease as heterogeneous as Alzheimer’s, such approaches may eventually inform risk stratification, resource allocation, and even trial design, by ensuring that patients with similar profiles are analyzed together.

Caution is warranted, however, and the authors are explicit about the limits of their work. This was a retrospective, cross-sectional analysis of 118 patients at a single hospital, and the profiles are described as exploratory. Clustering algorithms will always return groupings, and it remains to be shown whether these particular profiles are stable across time, reproducible in other institutions and health systems, and genuinely predictive of outcomes such as length of stay, discharge destination, readmission, or mortality. Longitudinal studies are needed to determine whether patients migrate between profiles as their disease evolves, and whether profile membership at admission can meaningfully guide treatment decisions. The researchers themselves state that external and longitudinal validation is required before these profiles can be considered a clinical tool rather than a descriptive insight.

Even so, the study adds a valuable piece to the emerging picture of Alzheimer’s disease as a syndrome of many faces. The weak correlation between function and behavior is a quiet but important statistic: it tells clinicians and caregivers that observing one dimension tells them little about the other, and that both must be assessed in their own right. The three profiles offer a vocabulary for that dual assessment, turning what looks like random variation into organized, discussable clinical patterns. As populations age and dementia-related hospitalizations climb, tools that convert messy ward data into actionable patient groupings will only grow in value. This analysis is an early step, but it points toward a more individualized future for dementia care, in which the label Alzheimer’s disease is only the beginning of the story, and the functional–behavioral profile helps determine how that story is managed, ward by ward, patient by patient.

Subject of Research: Functional–behavioral heterogeneity and clustering profiles in hospitalized patients with Alzheimer's disease

Article Title: Functional–behavioral profiles and clinical characteristics among hospitalized patients with Alzheimer’s disease: a real-world clustering analysis

Article References: Jie, W., Yu, X., Wu, J., Xu, Y., He, C., Fan, Y., Wang, Y., & Wu, H. (2026). Functional–behavioral profiles and clinical characteristics among hospitalized patients with Alzheimer’s disease: a real-world clustering analysis. BMC Geriatrics. https://doi.org/10.1186/s12877-026-08444-0

Image Credits: AI Generated

DOI: 10.1186/s12877-026-08444-0

Keywords: Alzheimer's disease, k-means clustering, Barthel Index, NOSIE, behavioral symptoms, functional ability, psychotic symptoms, hospitalized patients, individualized care, geriatric psychiatry, real-world data, dementia care

The research team, led by scientists at the International Centre for Recurrent Head and Neck Cancer based at the Royal Marsden Hospital and the London School of Hygiene & Tropical Medicine, assembled an extraordinary data infrastructure to answer a deceptively simple question: how long do head and neck cancer patients actually wait from first referral to the start of treatment, and which factors push some patients into dangerously long delays? To do this, they linked the national Cancer Registry with the Cancer Waiting Times dataset, Hospital Episodes Statistics, and national Radiotherapy and Chemotherapy datasets, creating a real-world picture of the entire diagnostic and treatment pathway that no single dataset could provide on its own.

The headline numbers are stark. Of the 30,506 patients analysed, 43 percent experienced what the study defines as treatment delay, meaning more than 62 days elapsed between referral and the start of first treatment, whether that was surgery or chemoradiation. The picture becomes even more concerning when the researchers examined the most extreme waits. Among patients whose primary treatment was surgery, 11 percent waited more than 90 days. For those treated primarily with radiotherapy or chemoradiation, the figure rose to 20 percent, one in five patients waiting more than three months from referral to the beginning of a treatment course on which their survival may depend.

Why does timing matter so much in this particular cancer? Head and neck tumours, which arise in the mouth, throat, voice box, nose and salivary glands, are among the most time-sensitive malignancies in oncology. Unlike some slow-growing cancers where a few extra weeks make little biological difference, head and neck tumours can progress rapidly, and a substantial body of evidence links even short treatment delays to worse outcomes. A landmark systematic review and meta-analysis published in the BMJ in 2020 estimated that delays in cancer treatment translate directly into excess mortality, with the risk of death rising measurably for every additional week of waiting. Studies of head and neck cancer specifically, including analyses of the United States National Cancer Database and systematic reviews published in JAMA Otolaryngology, have consistently shown that increasing time to treatment initiation is associated with poorer survival, larger tumours at treatment, and in some cases pathological upstaging, where the tumour is found to have advanced further than initial assessments suggested.

Against this biological backdrop, the new study’s identification of the strongest predictor of delay carries real weight. Patients who entered the cancer pathway through a non-urgent referral route, rather than the urgent suspected cancer referral pathway designed to fast-track people with red-flag symptoms, faced dramatically higher odds of long delay. After adjusting for other factors in multivariable logistic regression models, non-urgent referral was associated with an adjusted odds ratio of 4.07 for long delay, with a confidence interval of 3.54 to 4.68. In practical terms, patients referred through routine channels were roughly four times more likely to experience a treatment delay exceeding 62 days than those referred urgently. Notably, non-urgent referrals accounted for 30 percent of all patients in the cohort, meaning nearly a third of people who ultimately turned out to have head and neck cancer were not flagged as urgent when they first entered the system.

This finding points to a fundamental diagnostic challenge. The early symptoms of head and neck cancer, including persistent sore throat, hoarseness, mouth ulcers that fail to heal, difficulty swallowing, and unexplained lumps in the neck, overlap heavily with far more common and benign conditions. Distinguishing a cancer that will kill from a viral infection that will resolve within days is one of the hardest triage tasks in primary care, and the study’s authors argue that improving this recognition is where interventions should focus. Public and primary care awareness of head and neck cancer symptoms, they suggest, is a critical lever for shifting more patients onto the urgent referral pathway where rapid investigation follows.

But referral route was not the only determinant of delay. The researchers found that patients whose initial assessment took place at non-specialist hospitals, referred to in the study as spoke hospitals rather than the specialist hub centres where head and neck cancer expertise is concentrated, faced higher odds of delay. This reflects a broader tension in modern NHS organisation: the centralisation of complex cancer surgery into fewer, higher-volume specialist centres improves outcomes and quality, but it can add steps, transfers and logistical friction to the patient journey, particularly for those who first present at a district general hospital and must then be referred onward for specialist assessment. The study also found that patients with cancers arising outside the oropharynx, the part of the throat that includes the tonsils and base of the tongue, experienced longer waits, likely reflecting the more indistinct symptom profiles of tumours in sites such as the voice box, nasal cavity and salivary glands. Higher socio-economic deprivation emerged as a further risk factor, adding to a long record of evidence that cancer waits in England are systematically worse for patients from poorer communities, who may present later, navigate the system with less support, and face additional barriers at every stage of the pathway.

One of the study’s most technically interesting contributions is its demonstration that the drivers of delay differ fundamentally between the two main treatment modalities. Patients destined for primary surgery and patients destined for primary chemoradiation follow different pathways, and the factors that slow each group down are not identical. Surgical patients typically require pre-operative assessment, anaesthetic review and theatre capacity, while radiotherapy patients depend on planning processes, mould fabrication, and linear accelerator capacity, and those receiving concurrent chemotherapy face additional pharmaceutical and supportive care requirements. The finding that long delays were twice as common in the chemoradiation group, 20 percent versus 11 percent, suggests that the non-surgical pathway carries a distinct and heavier burden of system-level bottlenecks. The authors argue that any strategy to reduce waiting times must acknowledge and address these modality-specific drivers rather than treating the cancer pathway as a single undifferentiated queue.

The study’s methodology deserves attention because it exemplifies a growing trend in oncology research: the use of real-world data at national scale to answer questions that clinical trials, with their selected populations and controlled conditions, cannot address. By linking five national datasets covering registration, waiting times, hospital admissions, radiotherapy and chemotherapy, the researchers captured the full continuum of care for essentially an entire national population of head and neck cancer patients over the study period. The analyses were conducted in RStudio, and the work was funded by the International Centre for Recurrent Head and Neck Cancer, with ethics approval from the London School of Hygiene and Tropical Medicine Ethics Committee. The data were provided under licence by NHS Digital’s National Disease Registration Service, and the underlying data cannot be shared publicly, reflecting the governance constraints that accompany such sensitive linked health records.

The implications reach well beyond the borders of England. Health systems worldwide are grappling with post-pandemic backlogs, rising cancer incidence and constrained diagnostic capacity, and head and neck cancer services have been repeatedly identified as among those with the longest treatment waits. The study’s conclusions offer a roadmap: optimise referral pathways so that more patients with suspicious symptoms enter the urgent route, invest in symptom awareness among both the public and general practitioners, design diagnostic pathways that minimise transfers between spoke and hub hospitals, and build separate solutions for the surgical and radiotherapy arms of the service. The authors also point to emerging research programmes, such as the NIHR-funded EVEREST-HN initiative, which is developing a patient-reported symptom-based risk stratification system to redesign the suspected head and neck cancer referral pathway, as the kind of innovation the findings demand. For the one in five patients whose radiotherapy begins more than three months after their first referral, and for the 43 percent overall who wait beyond the 62-day standard, the message of this study is unambiguous: the delay is measurable, its causes are identifiable, and its consequences for survival are too well documented to ignore.

Subject of Research: Waiting times and determinants of treatment delay for head and neck cancer patients in England

Article Title: Waiting times and determinants of delay for head and neck cancer in England—a population cohort study using real world data

Article References: Jashek-Ahmed, F., Almilaji, O., Cowling, T. E., Hamilton, D. W., Walker, K., & Paleri, V. (2026). Waiting times and determinants of delay for head and neck cancer in England—a population cohort study using real world data. British Journal of Cancer. https://doi.org/10.1038/s41416-026-03630-8

Image Credits: AI Generated

DOI: 10.1038/s41416-026-03630-8

Keywords: head and neck cancer, treatment delay, waiting times, NHS England, cancer referral, real-world data, radiotherapy, cancer surgery, health inequalities, cancer epidemiology, diagnostic pathway, British Journal of Cancer

The analysis, published in Frontiers in Drug Safety and Regulation, was conducted by researchers in Safety Surveillance Research at Pfizer and led by Stacy Chen and Sarah C. MacDonald. The team queried the Heads of Medicines Agency and European Medicines Agency Catalogue of Real-world Data Studies on 12 June 2025, filtering for non-interventional post-approval safety studies in pregnant people that were required by a regulator, covered a human medicinal product, and had a publicly available protocol with prespecified safety outcomes. Of roughly 3,100 studies in the catalogue at the time of the query, only 30 met all of the eligibility criteria, a reminder of how narrow this slice of the pharmacovigilance landscape really is.

Across those 30 studies, the most frequently evaluated outcomes were stillbirth, appearing in 25 studies; congenital malformations, major or minor, in 24 studies; spontaneous abortion in 22; small for gestational age or fetal growth restriction in 21; and preterm birth or labor in 21. Elective termination appeared in 15 studies and preeclampsia in 13. Yet despite this apparent overlap, the researchers found no outcome that was measured in all of the studies, and none of the studies evaluated every outcome recommended by either the United States Food and Drug Administration or the European Medicines Agency. In other words, even the most commonly tracked pregnancy outcomes are not tracked universally.

The authors attribute much of this inconsistency to the guidance documents themselves. The FDA’s 2019 draft guidance on postapproval pregnancy safety studies provides a list of recommended data collection elements, covering the outcome of pregnancy, gestational complications, and infant and neonatal outcomes, but it does not explicitly state which outcomes a study must assess. The EMA’s guidance similarly asks sponsors to justify their chosen pregnancy and infant outcomes without supplying a comprehensive required list. Sponsors are therefore left to propose outcome sets on their own, often receiving feedback only when a post-marketing requirement letter is issued or during protocol review, a process the authors describe as inefficient and one that limits comparability across studies.

Study characteristics clearly shaped which outcomes were measured. Among the 14 studies committed to the FDA, stillbirth was the most studied outcome, appearing in 13 of 14 studies, and the average number of evaluated outcomes was approximately 12 per study, with a range of 5 to 25. Among the 16 EMA-committed studies, stillbirth and congenital malformations tied as the most common outcomes, each in 14 of 16 studies, but the average number of outcomes per study was only about seven, with a range of 2 to 11. Eclampsia, notably, was not evaluated in any EMA-committed study, while it appeared in half of the FDA-committed studies. Preeclampsia and small for gestational age or fetal growth restriction were also more frequently captured in FDA-committed studies.

Product type mattered as well. All six vaccine studies evaluated stillbirth, and five of six measured small for gestational age or fetal growth restriction and preterm birth or labor. Medication studies, of which there were 24, more often evaluated outcomes that require early pregnancy exposure: congenital malformations appeared in 21 of 24 medicinal studies compared with half of the vaccine studies, and elective termination appeared in 62.5 percent of medicinal studies and none of the vaccine studies. The authors suggest this pattern likely reflects when the vaccines under study, such as the RSVpreF vaccine, are recommended for administration during pregnancy, mostly late in pregnancy, which reduces the relevance of early-exposure outcomes. They also note that some outcomes, such as Guillain-Barré syndrome, anaphylaxis, and myocarditis, may be more relevant to vaccine products, yet existing guidance does not differentiate recommended outcomes by vaccine versus medicinal exposure.

Data collection method produced another split. Fourteen studies used primary data collection, typically pregnancy registries, while 16 relied on secondary data sources such as claims databases and medical records. Spontaneous abortion was the most frequent outcome in primary data studies, appearing in 12 of 14, while stillbirth dominated secondary data studies, appearing in 15 of 16. Two outcomes stood out as much harder to capture in secondary data: infant growth and development, measured in 79 percent of primary studies and none of the secondary ones, and elective termination, measured in 71 percent of primary studies versus 31 percent of secondary ones. The authors point out that growth metrics like weight, height, and head circumference are rarely included in secondary databases, and that elective terminations may not appear in claims data if insurance does not cover them or if women pay out of pocket. The FDA guidance itself acknowledges that direct access to mothers allows specialized physical examinations and developmental follow-up of offspring.

Perhaps the most sobering finding concerns statistical power. Finalized studies had a median of only 110 exposed participants, well below what is needed to detect meaningful risks. To detect a risk ratio of at least 2.0 for major congenital malformations, assuming 80 percent power and a 4 percent prevalence rate in the United States, an exposed cohort of approximately 600 is required; only 6 of 13 finalized studies examining that outcome met this threshold. For stillbirth, detecting a risk ratio of at least 3.0 requires roughly 1,700 exposed individuals, and only 4 of 11 finalized studies met that bar, with none of the planned or ongoing studies doing so. These figures echo earlier systematic reviews, which found the median enrollment in pregnancy registries to be just 36 pregnancies and that no registries were designed with sufficient power to assess specific malformations.

There were also signs of movement over time. Of the protocols dated after the FDA guidance was implemented in May 2019, nine outcomes saw an increase in the percentage of studies evaluating them, including preeclampsia, which jumped from zero percent in pre-2019 FDA-committed studies to 90 percent afterward, and eclampsia, which rose from zero to 70 percent. Meanwhile, some outcomes declined in frequency, including congenital malformations, spontaneous abortion, and elective termination, though the authors caution that only four FDA-committed studies predated the guidance, making these comparisons fragile.

The authors argue that harmonized regulatory guidance specifying a required set of outcomes, and ideally target effect sizes, would allow sponsors to generate protocols more efficiently, join existing disease-specific registries rather than building product-specific ones, and permit meaningful comparison of findings across studies. They acknowledge limitations, including the small number of studies analyzed, the reliance on publicly available protocols, the exclusion of studies relying only on spontaneous case reports, and the fact that the FDA does not require its post-approval studies to be registered in the European catalogue, meaning some FDA-committed studies may be missing. Still, the core message stands: without a common outcome framework, the pregnancy safety data that regulators, clinicians, and patients depend on remain fragmented, and the opportunity to pool scarce evidence is being lost study by study.

Subject of Research: Post-approval non-interventional safety studies in pregnancy and the standardization of evaluated outcomes

Article Title: Outcomes of post-approval non-interventional safety studies in pregnancy: wide variation demonstrates need for further standardization

Article References: Chen, S., & MacDonald, S. C. (2026). Outcomes of post-approval non-interventional safety studies in pregnancy: wide variation demonstrates need for further standardization. Frontiers in Drug Safety and Regulation, 6, Article 1805759. https://doi.org/10.3389/fdsfr.2026.1805759

Image Credits: AI Generated

DOI: 10.3389/fdsfr.2026.1805759

Keywords: pregnancy safety studies, post-approval studies, pharmacovigilance, FDA, EMA, congenital malformations, stillbirth, pregnancy registries, real-world data, regulatory guidance, sample size, vaccine safety

The study, published in the journal Pharmacia, is notable for what it did not use: the polished, idealized data of randomized clinical trials. Instead, researchers built their model on local real-world evidence drawn from 69 Bulgarian patients treated at the Specialized Hospital for Active Treatment of Hematological Diseases. Forty-nine patients received venetoclax plus azacitidine in a prospective cohort followed from 2022 to 2024, after venetoclax entered the national reimbursement lists, while 20 patients treated with azacitidine alone between 2018 and 2022 served as the comparator. By anchoring the azacitidine cohort in the pre-venetoclax era, the investigators deliberately avoided calendar-related bias, ensuring that both groups were treated within comparable modern clinical contexts.

All patients had newly diagnosed acute myeloid leukemia and were deemed ineligible for intensive chemotherapy for concrete clinical reasons: age of 75 or older, significant pulmonary or cardiac disease, liver cirrhosis, active resistant infection, poor functional status, kidney failure, or similar conditions. This is precisely the population in which treatment decisions carry the highest stakes and the least margin for error, and in which the balance between extending life and preserving its quality becomes most delicate.

Methodologically, the team constructed a one-year state-occupancy model with monthly cycles, dividing each patient’s trajectory into three mutually exclusive states: event-free survival, progressed or relapsed disease, and death. Rather than a full Markov transition model, which would have demanded stronger assumptions than the sparse progression data could justify, the researchers used a partitioned-survival-style approach in which monthly state proportions were derived directly from observed overall survival and reconstructed event-free survival. Progressed disease occupancy was calculated as the difference between the two survival curves, and death as one minus overall survival, with internal consistency checks ensuring that event-free survival never exceeded overall survival and that no state ever turned negative.

Quality of life entered the model through monthly patient responses to the EORTC quality-of-life questionnaire, mapped onto utility values using a published regression algorithm. Event-free patients carried a mean utility of 0.75, while those in progressed or relapsed disease averaged 0.63. Costs were tallied from the perspective of the Bulgarian National Health Insurance Fund and included drug acquisition, hospitalization, monitoring, diagnostic investigations, and supportive and palliative care, all priced at 2024 list prices and converted to euros at the fixed national exchange rate. Venetoclax acquisition alone cost roughly €9,760 per month, while azacitidine monotherapy ran about €2,802 monthly.

The base-case results were stark. Total yearly costs reached €99,532.86 per patient on the combination versus €44,344.09 on azacitidine alone, an incremental cost of €55,188.77. Against that expenditure, the combination delivered 0.51 quality-adjusted life-years versus 0.32, and 0.68 life-years versus 0.45, yielding incremental gains of 0.19 QALYs and 0.23 life-years. The resulting cost-effectiveness ratios were €289,309.59 per QALY gained and €235,736.28 per life-year gained, both far above Bulgaria’s accepted threshold of roughly €54,815, which corresponds to three times gross domestic product per capita.

Probabilistic sensitivity analysis using Monte Carlo simulation underscored how fragile the value proposition is at current prices. Only 4.4 percent of simulated cost-per-QALY estimates fell below the national threshold, and for cost per life-year gained the figure was 6.7 percent. The analysis suggested that the combination would only become acceptably cost-effective at a willingness-to-pay threshold of around €300,000, equivalent to more than twelve times GDP per capita, with results plateauing near €400,000. The model also faithfully captured the darker side of the therapy: some patients experienced severe adverse events during induction, and in certain simulations the combination produced negative quality-adjusted or even negative life-year gains, reflecting early treatment-related mortality.

One-way sensitivity analyses revealed that the results were most sensitive to the effectiveness of the combination itself. If venetoclax plus azacitidine delivered 30 percent lower QALYs or life-years than assumed, the cost per QALY soared past €1.5 million and the cost per life-year beyond €1.9 million. Yet the dispersion of probabilistic results clustered tightly around the base case for survival outcomes, suggesting that the survival benefit itself was relatively consistent, even as quality-of-life impacts varied more widely across patients.

Perhaps the most intriguing finding is how closely the Bulgarian real-world result mirrors trial-based economics from wealthier countries. The exploratory cost-utility ratio of roughly €289,000 per QALY aligns remarkably well with a United States estimate of $260,343 per QALY, lending incidental validation to the quality-of-life mapping algorithm the Bulgarian team employed. Meanwhile, first-year survival in the local cohort was numerically higher on the combination, and event-free survival was significantly better, indicating improved disease control even within a single year of follow-up. The authors contrast this with a real-world study in older, frailer patients where the combination was poorly tolerated, noting that their cohort’s median age of 71, close to Bulgarian life expectancy, may partly explain the difference.

The researchers are careful to frame their work as hypothesis-generating rather than reimbursement-grade. The sample of 69 patients is small, progression ascertainment was incomplete, the one-year horizon misses longer-term survival and late costs, and no external validation against larger multicenter datasets was yet possible. Longer time horizons in other analyses have pushed cost-effectiveness ratios down dramatically, to figures as low as $74,141 per life-year gained, suggesting that the true value of the combination may be substantially better than a one-year snapshot can capture. Still, the study delivers something Bulgaria has lacked: a transparent, locally grounded account of what this breakthrough therapy actually costs and delivers within its own health system. As venetoclax and similar targeted agents spread across Central and Eastern Europe, models like this one will become essential instruments for deciding which medical miracles a national payer can realistically afford, and at what price the arithmetic of hope must ultimately be settled.

Subject of Research: Cost-effectiveness of venetoclax plus azacitidine versus azacitidine monotherapy in Bulgarian patients with acute myeloid leukemia ineligible for intensive chemotherapy

Article Title: Exploratory cost-utility and cost-effectiveness analysis of venetoclax plus azacitidine vs. azacitidine monotherapy in Bulgarian patients with AML ineligible for intensive chemotherapy: a real-world-data-influenced state-occupancy model

Article References: Donchev, M., Davidkova, Y., Tachkov, K., & Petrova, G. (2026). Exploratory cost-utility and cost-effectiveness analysis of venetoclax plus azacitidine vs. azacitidine monotherapy in Bulgarian patients with AML ineligible for intensive chemotherapy: a real-world-data-influenced state-occupancy model. Pharmacia, 73, Article e197277. https://doi.org/10.3897/pharmacia.73.e197277

Image Credits: AI Generated

DOI: 10.3897/pharmacia.73.e197277

Keywords: acute myeloid leukemia, venetoclax, azacitidine, cost-effectiveness, cost-utility analysis, QALY, real-world data, state-occupancy model, Bulgaria, health economics, pharmacoeconomics, National Health Insurance Fund

The study drew on 2,270 health-examination records generated in 2025 through China’s National Basic Public Health Service program, which provides annual checkups for residents aged 65 and older. The records came from a rural primary-care program in Wangkui County, Heilongjiang Province, an agricultural region where 20 examination sites contributed data. The researchers, led by Kun Xu of Henan University’s Institute of Gerontology, set out to answer a deceptively simple question: could the routine main-medication field in these records be trusted as a measure of medication exposure in secondary research?

The answer turned on a comparison that initially looked compelling. Records were split by the exercise-frequency item into those with recorded physical activity, meaning daily, weekly, or occasional exercise, and those recording no exercise at all. Among the 2,226 records with usable activity information, nonblank medication fields appeared in 56.4 percent of the active group versus 42.8 percent of the inactive group, a crude difference of 13.6 percentage points. A modified Poisson regression, chosen because the outcome was common, produced a prevalence ratio of 1.317 with a confidence interval comfortably excluding the null. On its face, older adults who exercised appeared far more likely to be taking medications.

But that interpretation collapses under scrutiny of the data-generating process. Under the national health-service specification, the main-medication item was never designed as a population-wide drug inventory. It was intended to document principal medications used during the preceding year among people receiving long-term treatment for chronic disease, specifically within hypertension and diabetes management programs. When the researchers cross-tabulated field presence against chronic-disease management registration, the pattern was stark: 96.3 percent of records with hypertension and/or diabetes registration had nonblank medication fields, compared with just 2.4 percent of records with neither registration. The field was essentially a switch flipped by enrollment in a disease-management pathway.

Because exercise status was itself correlated with registration, the apparent activity contrast was a mixture of two things: any real relationship between activity and medication use, and the fact that active records were more likely to belong to the registered group where medication fields were almost always filled. The researchers built a sequence of models to watch the contrast dissolve. After adjusting for age, sex, and examination site, the prevalence ratio actually rose slightly to 1.346. But adding a four-category hypertension-diabetes registration variable, which allowed each combination of registrations its own effect, dropped the ratio to 1.001, with a confidence interval from 0.960 to 1.043. Further adjustment for body mass index, smoking, and alcohol use left it essentially unchanged at 1.002.

The absolute scale told the same story. Marginal standardization of a logistic model, which sets every record’s activity status in turn while holding all other covariates at observed values, yielded standardized field-presence proportions of 51.26 percent for the active group and 51.14 percent for the inactive group, a difference of just 0.12 percentage points with a confidence interval spanning negative 2.07 to 2.30. Within registration strata, the differences were tiny: negative 1.41 percentage points among unregistered records, 0.44 among hypertension-only records, and negative 0.56 among dual-registered records. Only the sparse diabetes-only stratum, with just 37 records, showed a large descriptive difference, and the authors explicitly declined to generalize from it.

The team did not stop at field presence. They audited the medication text itself, splitting 1,141 nonblank fields into 1,368 retained text entries after removing exact duplicates. A version-controlled rule-based parser identified medication names, doses, units, and duration information, and a random sample of 200 entries was independently reviewed by two pharmacists masked to the parser’s output. Inter-reviewer agreement on broad therapeutic class reached 99.5 percent, with a kappa of 0.982, and parser-to-manual concordance on that domain was 99.0 percent. Archived labels classified 75.9 percent of entries as antihypertensive and 14.5 percent as glucose-lowering, with 7.4 percent unresolved. Yet the authors are careful to note that this high concordance validates only the interpretation of recorded text, not actual medication exposure, because no linked prescriptions, dispensing records, or medication reconciliation data existed.

The audit also exposed structural ambiguities that would plague any secondary analysis. Nearly half of the entries, 45.0 percent, contained a bare number in the duration field with no time unit, and another 15.5 percent held timing or frequency information rather than a treatment duration. Standardized medication naming showed more variability than direct text extraction, with parser-to-reference concordance of 77.5 percent on exact normalized matching, reflecting differences in formulation qualifiers, salt wording, and dictionary granularity. A blank field, meanwhile, could mean no long-term medication, inapplicability, the question never asked, unknown status, refusal, or simple omission, all compressed into an undifferentiated empty cell.

The implications reach well beyond one county in Heilongjiang. Routine health data are products of healthcare and administrative processes rather than neutral research measurements, and informative field presence can generate associations that reflect documentation architecture rather than clinical reality. The authors sketch the specific risks: coding blank fields as medication nonuse would produce differential exposure misclassification patterned by disease registration; restricting analyses to records with nonblank fields would introduce selection bias by preferentially retaining records from chronic-disease pathways; and the intended one-year reference period may misalign with outcomes measured at a single visit. They propose a tiered approach, reserving the field for documentation audits and narrowly defined descriptive work within its intended population, while requiring validation against prescriptions, dispensing data, or structured medication reconciliation for any study of drug utilization, safety, effectiveness, or medication-outcome associations.

The study also offers concrete fixes for primary-care information systems: separating disease-management prompts from comprehensive medication lists, storing identity, indication, dose, route, frequency, and status in distinct structured elements, adopting standardized terminologies such as RxNorm and the ATC/DDD classification, and replacing undifferentiated blanks with explicit categories distinguishing nonuse, inapplicability, and missingness. The limitations are acknowledged candidly, including the single-center setting, the cross-sectional design, the absence of stable person-level identifiers, and reliance on only 20 site clusters. But the central message stands with unusual clarity: before any routinely collected variable is reused as research data, its intended population, recording workflow, and reference period must be understood. A field fit for its original operational purpose can be quietly unfit for the next question asked of it, and the ghost of that purpose can masquerade as a scientific finding.

Subject of Research: Secondary-use suitability of a routine primary-care medication field and its association with recorded physical activity among rural older adults

Article Title: Among rural older adults, an apparent physical-activity contrast in nonblank medication-field presence was markedly attenuated when chronic-disease management registration was incorporated into the analysis

Article References: Xu, K., Zhang, Y., Zhao, Y., Xi, X., Hou, G., Guo, Z., Liu, L., Wang, Y., & Zhang, D. (2026). Among rural older adults, an apparent physical-activity contrast in nonblank medication-field presence was markedly attenuated when chronic-disease management registration was incorporated into the analysis. Frontiers in Pharmacology, 17, Article 1965382. https://doi.org/10.3389/fphar.2026.1965382

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1965382

Keywords: pharmacoepidemiology, real-world data, medication data quality, electronic health records, older adults, physical activity, chronic disease management, exposure misclassification, primary care, rural China, data provenance, selection bias

The rationale for repeating critical results before reporting traces back to guidelines issued decades ago, when haematology analysers were mechanically simpler and flagging systems less sophisticated. At that time, a second measurement served as one of the few available checks against instrument drift, sample aspiration problems, or transient electrical noise. As analytical platforms evolved, manufacturers built layered verification into the instruments themselves: hydrodynamic focusing, repeated cell counting within a single aspiration, internal quality-control materials, and real-time decision rules that flag suspicious distributions of cell counts or optical signals. The question facing laboratories today is not whether analytical quality matters, but whether a blanket repetition policy still adds meaningful protection or simply adds cost, delay, and workload without changing clinical decisions.

The Portuguese data speak directly to that question. Across nearly half a million haemograms reviewed over five years, the overwhelming majority of critical haemoglobin repeats returned values essentially identical to the first measurement. Almost half of the paired readings were exactly equal, and fewer than two in every hundred differed by more than two grams per litre. When discrepancies did exceed the clinically relevant threshold of four grams per litre, investigation revealed pre-analytical culprits rather than analytical ones: clotted samples, fibrin strands, or mislabelled tubes that required recollection regardless of any repeat policy. This pattern is important because it locates the residual risk where laboratory medicine has long suspected it resides, in the journey of the specimen before it reaches the analyser, not in the analytical measurement itself.

The distinction between analytical and pre-analytical error carries practical consequences. A repeat measurement on the same tube can detect instrument malfunction, but it cannot detect a clot that formed in the tube before analysis, a sample drawn from the wrong patient, or a label applied to the wrong container. Those failures demand different countermeasures: better phlebotomy practice, tube integrity checks, positive patient identification, and vigilant review of sample quality indicators. When a laboratory discovers, as this one did, that every large discrepancy traced back to pre-analytical problems, the case for automatic repetition weakens further, because repetition is the wrong tool for the failure modes that actually occur.

The study also illustrates a meaningful difference between intra-analyser and inter-analyser agreement. When repeats ran on the same module, nearly ninety-eight percent of pairs agreed within one gram per litre, and the coefficient of variation sat comfortably below the manufacturer’s specification. When repeats ran on a different module, agreement loosened slightly, with roughly eighty percent within one gram per litre and wider limits of agreement. This is expected behaviour for any multi-instrument platform: each analyser carries its own calibration, optical characteristics, and reagent lot interactions, and small between-module differences accumulate. The observed inter-analyser bias of less than a quarter of a gram per litre, and a coefficient of variation well within the manufacturer’s tolerance, nonetheless indicate that the laboratory’s harmonisation practices, including calibration alignment and shared maintenance schedules, were functioning as intended across the fleet.

The small negative bias observed in both comparison types deserves comment. Repeat measurements tended to read marginally lower than initial ones, by roughly a tenth to a fifth of a gram per litre. A bias of this magnitude is far below any threshold of clinical significance for haemoglobin, and it may reflect subtle sample-level changes such as settling of cells within the tube between aspirations, minor evaporation, or slight dilution effects from mixing. Its consistency across hundreds of paired measurements, however, demonstrates the power of large duplicate datasets: effects too small to notice in individual cases become statistically visible and quantifiable when aggregated over years of routine operation.

Methodologically, the study occupies a deliberate middle ground. The authors are careful to state that their retrospective analysis does not replace formal precision protocols such as those published by the Clinical and Laboratory Standards Institute, which prescribe controlled experiments with defined materials, replication schemes, and statistical treatment. Those protocols remain the foundation for verifying manufacturer claims and for regulatory compliance. What the retrospective approach adds is a complementary layer: evidence of how instruments behave under the messier conditions of real practice, with heterogeneous patients, variable sample quality, rotating operators across day, evening, and night shifts, and the full range of clinical demand that a twenty-four-hour hospital laboratory experiences. A analyser that performs beautifully on control material but degrades under routine load would be exposed by longitudinal duplicate data in a way that a bench experiment might not reveal.

The intellectual lineage of this approach is older than modern automation. The idea of using naturally occurring duplicate measurements to estimate analytical imprecision echoes work from the late 1970s, when duplicate patient samples were proposed as a window into hidden analytical variability. What has changed is scale and feasibility. In the era of paper records, assembling thousands of duplicate pairs required laborious manual effort. Today, haematology analysers archive their raw measurement data automatically, and laboratory information systems can be interrogated with scripting tools to extract, pair, and analyse those records with minimal human intervention. The Portuguese team used open-source statistical software to automate sample pairing across nearly half a million records, a task that would have been impractical a generation ago.

This reframing of archived data as a quality asset has implications beyond haemoglobin. Modern analysers generate a rich stream of ancillary signals, including cell population data, flag frequencies, and instrument status parameters, much of which is stored but rarely examined systematically. Periodic mining of these archives could complement traditional internal quality control, which typically relies on control materials run at intervals and may miss intermittent faults or sample-specific problems. Duplicate-based monitoring, by contrast, draws on real patient specimens analysed under genuine operating conditions, and it costs essentially nothing beyond computational effort and analyst time. For laboratories facing budget pressure and workforce shortages, a zero-cost surveillance layer that runs continuously in the background is an attractive proposition.

The economics of reflex repetition deserve scrutiny in their own right. Each automatic repeat consumes reagent, instrument time, and technologist review, and it adds to turnaround time for a result that clinicians are waiting on in urgent clinical situations. Studies cited in the original letter have argued that routine repetition of critical haematology values wastes resources and delays clinical action without improving accuracy. When the measured repetition rate of meaningful error is as low as this study found, the marginal benefit of a universal repeat policy approaches zero, while the marginal cost is paid on every single critical result. A targeted policy, in which repetition is reserved for samples with flags, suspicious histograms, or clinical indications, would preserve the safeguard where it has value while removing it where it does not.

Any change to a long-standing standard operating procedure requires careful governance. Laboratories considering a move away from blanket repetition would need to document their own baseline performance, ideally with longitudinal duplicate analysis of the kind described here, and to establish ongoing monitoring so that any degradation in analytical performance triggers a return to more conservative practices. The override pathway used in the Portuguese laboratory, in which a small fraction of critical results were validated by a clinical pathologist without repetition based on clinical urgency, shows one model for balancing safety with efficiency: a defined, auditable process that concentrates expert judgement where it is most needed rather than applying mechanical rules indiscriminately.

The five-year span of the dataset is itself a strength. Analytical performance is not static: reagent lots change, lamps and optical components age, mechanical parts wear, and software updates alter behaviour. Demonstrating stable agreement across half a decade of continuous operation, with nearly seven thousand measurements from three interconnected modules, provides a level of assurance that a short-term validation study cannot match. It also demonstrates that the monitoring approach is sustainable, since the same pipeline that produced this analysis can be rerun periodically to detect emerging drift or module divergence before it affects patient care.

For the wider laboratory medicine community, the message is twofold. First, the evidence base for questioning universal repetition of critical haemoglobin results continues to grow, and laboratories still enforcing such policies should examine their own duplicate data to determine whether the practice is protecting patients or merely consuming resources. Second, the raw data already flowing through modern analysers represent an underused resource for quality assurance. With modest analytical effort, routinely generated duplicate measurements can be transformed into a continuous, real-world performance monitor, one that complements rather than replaces formal precision studies and that costs almost nothing to maintain. In an era when laboratories are asked to do more with less, turning existing data exhaust into actionable quality intelligence is a strategy whose time has arrived.

Subject of Research: Longitudinal agreement of repeated critical haemoglobin measurements as a measure of analytical quality in a hospital haematology laboratory

Article Title: Longitudinal agreement of repeated critical haemoglobin measurements as evidence of sustained analytical quality

Article References: Simões, V., Magalhães, C., & Eremina, Y. O. (2026). Longitudinal agreement of repeated critical haemoglobin measurements as evidence of sustained analytical quality. British Journal of Biomedical Science, 83, Article 16480. https://doi.org/10.3389/bjbs.2026.16480

Image Credits: AI Generated

DOI: 10.3389/bjbs.2026.16480

Keywords: haemoglobin, critical values, internal quality control, haematology analysers, Bland-Altman analysis, reflex testing, laboratory medicine, coefficient of variation, Sysmex XN-9100, pre-analytical errors, clinical pathology, real-world data

The scale of the problem is stark. According to the review, led by Ryan Kilpatrick and Jonathan Davis of Tufts University School of Medicine together with Rachel Greenberg of Duke University, Danielle Boyce of Johns Hopkins University, and Kanwaljit Singh of the Critical Path Institute, most medications administered in the neonatal intensive care unit lack adequate safety, efficacy, and dosing data. Newborns have been described as one of the last therapeutic orphans: drugs are routinely used off-label, at doses extrapolated from older children or adults, in patients whose developing kidneys, livers, and brains handle medications in profoundly different ways. No major therapeutic advance has improved preterm survival since the introduction of surfactant therapy in the early 1990s.

Why has this population been left behind? The review points to a convergence of ethical and practical barriers that make traditional randomized clinical trials extraordinarily difficult in ELGANs. Parents are asked to consent to research for a child they have just met, often in the midst of a medical crisis, and studies of enrollment decision-making show how heavy that burden is. The population is small relative to adult disease areas, fragmented across hundreds of intensive care units, and heterogeneous in ways that demand enormous sample sizes to detect meaningful effects. Regulatory incentives have historically done little to encourage investment in neonatal drug development, and pediatric trials more broadly suffer from high rates of early discontinuation and incomplete reporting.

And yet, paradoxically, ELGANs generate more clinical data than almost any other patient group. A single stay in the neonatal intensive care unit can stretch for months, during which continuous physiologic monitors stream heart rate, oxygen saturation, and respiratory measurements second by second. Laboratory values are drawn daily or more often. Every medication is recorded with dose, route, and timing, adjusted to a body weight measured on a daily or twice-daily schedule. Imaging studies, ventilation settings, nutrition records, and neurodevelopmental assessments accumulate in the electronic health record. The review’s authors argue that this extraordinary density and granularity makes ELGANs ideal candidates for real-world data research—if the infrastructure exists to use it.

Real-world data, in the language of regulators, refers to information relating to patient health collected outside the confines of conventional randomized trials: electronic health records, claims databases, registries, and device data. When such data are analyzed with rigorous methods, they become real-world evidence, and agencies including the United States Food and Drug Administration and the European Medicines Agency have published frameworks describing how such evidence can support regulatory decision-making. The FDA’s 2024 guidance on assessing electronic health records and claims data, and the EMA’s real-world evidence framework for regulator-led studies, signal a regulatory environment far more receptive to these approaches than a decade ago.

The technical challenge, however, is formidable. The review examines the current landscape of common data models—standardized structures such as the OMOP model developed by the Observational Health Data Sciences and Informatics collaboration, and interoperability standards like HL7’s FHIR—that allow different institutions to pool and compare their data. These models were not designed with neonatal complexity in mind. They lack native support for linking mothers and infants, a critical requirement when maternal exposures, medications, and pregnancy complications shape neonatal outcomes. They struggle with continuous physiologic waveforms, with weight-based medication dosing that changes daily, and with the long-term neurodevelopmental follow-up that defines success in this population. Extensions are emerging, including perinatal adaptations of OMOP and standardized representations of ventilation data, but the review concludes that neonates remain poorly served by existing infrastructure.

To address this, the authors propose a minimum neonatal-specific variable set: a core collection of data elements that every neonatal network should capture in a standardized way. This includes gestational age and birth weight, maternal and delivery characteristics, medication exposures with precise dosing, respiratory support parameters, laboratory values, diagnoses such as bronchopulmonary dysplasia defined using evidence-based criteria, adverse events graded with neonatal-specific severity scales, and neurodevelopmental outcomes extending well beyond discharge. Defining bronchopulmonary dysplasia, the chronic lung disease of prematurity, illustrates the difficulty: multiple competing definitions exist, and without harmonization, one center’s outcome cannot be meaningfully compared with another’s. Data quality assessment frameworks, the authors note, must be applied at every step, because a model is only as trustworthy as the messy clinical reality it encodes.

The regulatory applications of well-structured real-world data are where the review’s vision becomes most consequential. External controls—comparison groups drawn from existing patient data rather than randomized within the trial—could allow new therapies to be tested in ELGANs without withholding promising treatment from a control group, an approach the FDA has formally addressed in guidance for externally controlled trials. Target trial emulation, a framework developed by epidemiologist Miguel Hernán and colleagues, allows researchers to design an observational analysis that mimics a hypothetical randomized trial, specifying eligibility, treatment strategies, and follow-up in advance to avoid the biases that plague naive analyses of registry data; emulation studies across 32 clinical trials have shown that such designs can reproduce randomized results when executed carefully. Pharmacovigilance—the ongoing surveillance of drug safety—could operate continuously across networks rather than waiting for rare events to surface in small trials. Natural history studies, mapping how diseases progress without intervention, remain foundational for a population in which the course of illness is still incompletely described.

The review is equally clear that technology alone will not deliver this transformation. Building real-world data infrastructure for ELGANs requires coordinated investment from a broad coalition of stakeholders. Neonatologists and bedside caregivers must ensure that the data entered into electronic health records are accurate and complete, often under time pressure at the cot side. Informaticians must design models and vocabularies that respect neonatal physiology, extending frameworks such as the GA4GH Phenopacket schema for computable clinical representation. Regulators must continue to clarify when real-world evidence is acceptable, and industry must see a viable path to neonatal drug development. Families, the authors emphasize, belong at the table as well: international priority-setting partnerships led by parents of the most premature infants have already identified research questions that matter most to those who live with the consequences.

The stakes extend beyond individual units. Wide practice variation across neonatal intensive care units is well documented, and outcomes for very preterm infants differ substantially across health systems even after adjustment for case mix. That variation is often treated as a problem of quality; the review reframes it as an opportunity, because differences in practice, captured in standardized data, become natural experiments that can reveal which strategies genuinely improve survival without severe morbidity. Cohorts such as the European EPICE study have shown that greater use of evidence-based practices is associated with better outcomes, but identifying those practices requires the kind of harmonized, granular data infrastructure the authors describe. If the vision is realized, the data exhaust of intensive care—currently a byproduct of survival—could become the engine of the first genuine therapeutic progress for extremely preterm infants in more than three decades.

Subject of Research: Use of neonatal real-world data to generate clinical evidence for extremely preterm infants

Article Title: Enabling evidence generation for extremely preterm neonates: the use of neonatal real-world data

Article References: Kilpatrick, R., Greenberg, R. G., Boyce, D., Singh, K., & Davis, J. (2026). Enabling evidence generation for extremely preterm neonates: the use of neonatal real-world data. Pediatric Research. https://doi.org/10.1038/s41390-026-05561-9

Image Credits: AI Generated

DOI: 10.1038/s41390-026-05561-9

Keywords: real-world data, extremely preterm neonates, neonatal intensive care, ELGAN, clinical trials, electronic health records, OMOP common data model, external controls, target trial emulation, pharmacovigilance, drug dosing, neurodevelopmental outcomes

The study, led by Ann-Sophie Stratil and colleagues including Tjalf Ziemssen, analyzed retrospective data from Brisa, a prescription-free digital health application designed to support people with multiple sclerosis throughout their patient journey. The app allows users to log symptoms on a simple emoticon-based scale ranging from one, indicating poor condition, to five, indicating excellent condition, and to complete validated patient-reported outcome measures, or PROs. These included the Modified Fatigue Impact Scale for fatigue, the Patient-Determined Disease Steps for self-assessed mobility, the five-item Perceived Deficits Questionnaire for cognitive symptoms, the Incontinence Voiding Intensity Scale for bladder symptoms, and the Pain Effects Scale for pain. Together, these instruments cover many of the symptoms that most strongly shape quality of life in multiple sclerosis but that often receive less clinical attention than relapse counts or lesion load.

The researchers set strict inclusion criteria to keep their analysis clean. Users had to be at least eighteen years old, had to have reported use of a disease-modifying therapy classified into one of three categories according to German neurological society guidelines, and had to have remained on the same therapy without switching during the observation window. Category one covered low-efficacy treatments, category two medium-efficacy treatments, and category three the high-efficacy agents that have increasingly become first-line options in modern multiple sclerosis care. Users diagnosed within the previous zero to one years were classified as newly diagnosed, allowing the team to track how treatment patterns are shifting at the very start of the disease course.

The resulting cohort painted a familiar demographic picture of the multiple sclerosis population, but in unusually fine grain. Among 4,194 users, 80.8 percent were female, consistent with the well-documented female predominance of the disease. The largest age group, 30.5 percent of users, was between twenty-six and thirty-five years old, and 77.4 percent reported a diagnosis of relapsing-remitting multiple sclerosis, the form characterized by discrete attacks followed by periods of recovery. Notably, 27.9 percent of the cohort fell into the newly diagnosed category, meaning more than one in four app users were within their first year of diagnosis, a population whose treatment decisions are among the most consequential in the entire disease course.

One of the most striking findings concerned those early treatment decisions. Among newly diagnosed users, the proportion receiving category three, high-efficacy disease-modifying therapies increased significantly between 2023 and 2024 and 2025, a difference the authors report as statistically robust with a p-value below 0.0001. This shift mirrors a broader transformation in the field, in which neurologists have moved away from the traditional escalation approach, starting with milder drugs and reserving potent agents for treatment failures, toward an early high-efficacy strategy that aims to suppress inflammatory disease activity before irreversible damage accumulates. That this trend is visible in self-reported app data, rather than in registry data curated by clinics, suggests digital platforms can capture prescribing patterns as they happen in routine care.

Engagement, the perennial Achilles heel of digital health, was treated with unusual statistical rigor. The researchers defined engagement as any screen activity within the app and disengagement as more than 180 consecutive days without any activity, a threshold long enough to distinguish genuine abandonment from ordinary lapses in use. They then applied Cox proportional hazards regression, a survival-analysis technique that estimates how different characteristics change the hazard of an event over time, to identify which users were most likely to stay active. The results challenged some common assumptions about who benefits from health apps.

Three groups showed a significantly lower risk of disengagement. Users receiving high-efficacy, category three therapies were more likely to remain engaged than those on category one treatments, possibly reflecting a greater sense of investment in monitoring their disease under aggressive treatment. Users aged thirty-six and older stayed engaged longer than those aged eighteen to twenty-five, suggesting the youngest users, often assumed to be the most digitally native, may actually be the hardest to retain. And newly diagnosed users showed lower disengagement risk than those diagnosed two or more years earlier, indicating that the period immediately after diagnosis, when patients are actively seeking information and control, is when digital tools have their strongest foothold. For developers and clinicians alike, this is a roadmap: engagement strategies should be tailored to the newly diagnosed, to older users, and to those on intensive treatment regimens.

Overall, about 76 percent of users tracked at least one symptom or patient-reported outcome, a remarkably high proportion for a voluntary health application. The longitudinal picture among users with at least one year of tracking was one of stability: symptom ratings and PRO scores remained largely unchanged over time. At first glance, stability might seem like a null result, but in the context of multiple sclerosis it is arguably the most reassuring finding possible. It suggests that the population actively using the app, many of them on modern high-efficacy therapies, was not experiencing progressive worsening of fatigue, mobility, cognition, bladder function or pain during the observation period. It also demonstrates that these instruments, designed for clinical settings, produce coherent, trackable data when completed by patients at home on their own schedule.

The technical significance of the study lies in what it validates rather than in any single clinical result. Patient-reported outcomes collected in randomized trials are often criticized as artificial, captured under conditions that bear little resemblance to daily life. Real-world digital collection solves that problem but introduces others: attrition, self-selection, inconsistent use, and the question of whether people who download a health app are representative of patients in general. By characterizing exactly who uses the app, how long they stay, and how their self-reports trend over time, the Brisa analysis provides the kind of baseline that future studies will need to interpret digital symptom data responsibly. The finding that engagement itself is patterned by treatment intensity, age and disease duration is particularly valuable, because it tells researchers which covariates must be accounted for when app data are used to infer disease activity or treatment effects.

The implications reach beyond multiple sclerosis. Digital health applications are being prescribed and deployed across chronic disease care in Germany and across Europe, yet evidence about their real-world performance, who uses them, whether they keep users, and whether the data they generate are clinically meaningful, remains thin. This study offers a template for generating that evidence: define engagement operationally, apply survival analysis to retention, anchor symptom tracking to validated instruments, and report demographic and clinical characteristics of the user base transparently. If the shift toward high-efficacy therapy visible in this data continues, and if digital tools can reliably flag symptom changes between clinic visits, the quiet data stream from a phone in a patient’s pocket may become as important to managing multiple sclerosis as the annual MRI. For now, the Brisa analysis shows that thousands of patients are willing to report how they feel, day after day, and that what they report is stable, structured and scientifically usable, a foundation on which the next generation of remote disease monitoring can be built.

Subject of Research: Longitudinal real-world symptom tracking and patient-reported outcomes in users of a German digital health app for multiple sclerosis

Article Title: A longitudinal analysis of real world symptom tracking and patient-reported outcomes in Brisa, a German Digital Health App for multiple sclerosis

Article References: Stratil, A.-S., Papukchieva, S., Steinmetz-Späh, J., Joschko, N., Masanneck, L., Friedrich, B., & Ziemssen, T. (2026). A longitudinal analysis of real world symptom tracking and patient-reported outcomes in Brisa, a German Digital Health App for multiple sclerosis. PLOS Digital Health, 5(10), e0001751. https://doi.org/10.1371/journal.pdig.0001751

Image Credits: AI Generated

DOI: 10.1371/journal.pdig.0001751

Keywords: multiple sclerosis, digital health app, patient-reported outcomes, symptom tracking, real-world data, disease-modifying therapy, app engagement, high-efficacy treatment, newly diagnosed, longitudinal analysis, PLOS Digital Health, Cox regression

The study, led by Yuan Liu and Leng Han of the Indiana University School of Medicine together with colleagues including Eytan Ruppin, Douglas B. Johnson and Javid J. Moslehi, began with a sobering audit of the evidence base. Of the 487,688 trials listed on ClinicalTrials.gov up to March 2024, 97,254 were oncology-related and 7,274 involved immune checkpoint inhibitors. Yet only 210 of those immunotherapy trials — a mere 2.89 percent — included race information and enrolled more than 100 participants. The overwhelming majority, 86.77 percent, either lacked race information entirely or had enrolments too small to support meaningful comparisons. Most striking of all, among the 22 trials that enrolled at least 100 participants with more than 20 patients from each major racial group, not a single one reported adverse event data stratified by race.

To circumvent this data desert, the team turned to real-world evidence. They first mined Optum’s de-identified Clinformatics Data Mart Database, spanning 2007 to 2023 and encompassing 81.10 million insured individuals. Within it they identified 6.22 million patients with cancer diagnoses, of whom 91,852 received anti-PD-1 or anti-PD-L1 therapy: 1,979 Asian, 10,341 Black, 5,795 Hispanic and 65,907 white patients. Using diagnosis codes to capture immune-related adverse events arising after the start of checkpoint blockade, they documented 91,169 such events in 48,611 patients. After adjusting for age and sex with multivariable logistic regression, a clear pattern emerged: Black patients had higher odds of overall immune-related adverse events than Hispanic patients (odds ratio 1.08, 95 percent confidence interval 1.01 to 1.15), while Asian patients had lower odds than white patients (odds ratio 0.89, confidence interval 0.81 to 0.97).

Crucially, the findings did not rest on a single dataset. The team replicated the analysis in the National COVID Cohort Collaborative, a nationwide electronic health record resource containing data on 22.8 million individuals across 98 sites as of April 2024. The same directional patterns held: higher odds among Black patients compared with Hispanic patients, and lower odds among Asian patients compared with white patients. When the researchers stratified results by organ system using the MedDRA classification, aggregating adverse events into 13 system organ classes, they found that of 29 significant differences identified in the insurance database, 24 — 82.8 percent — showed consistent directional trends in the electronic health record cohort, a level of agreement that argues strongly against the results being statistical artifacts.

At the organ level, some of the differences were substantial. Black patients showed markedly higher odds of musculoskeletal and connective tissue disorders than Asian patients (odds ratio 1.44, confidence interval 1.30 to 1.60) and moderately higher odds than white patients (odds ratio 1.06). Conversely, Black patients had significantly lower odds of cardiac disorders than Asian patients (odds ratio 0.56, confidence interval 0.36 to 0.85), a finding with obvious clinical relevance given that checkpoint inhibitor myocarditis is among the most feared and fatal immune-related toxicities. When the analysis was further stratified by primary cancer site, the disparities sharpened: Black patients with stomach cancer had more than three times the odds of musculoskeletal disorders compared with Asian patients (odds ratio 3.31), white patients with liver cancer had lower odds of gastrointestinal disorders than Hispanic patients (odds ratio 0.68), and Black patients showed higher occurrence of nervous system disorders relative to Hispanic patients (odds ratio 1.42).

Drilling down to individual adverse event subtypes revealed even finer texture. Among 174 subtype-level comparisons tested in the insurance database, 60 showed differential odds across racial groups, and of the 56 that could be evaluated in the electronic health record cohort, 85.7 percent showed the same direction of effect. Arthritis stood out: Black patients had more than twice the odds of arthritis compared with Asian patients (odds ratio 2.27, confidence interval 1.92 to 2.68) and higher odds than white patients (odds ratio 1.27). Asian patients, by contrast, experienced significantly more rash than white patients (odds ratio 1.30), and in the validation cohort also showed higher rates of pruritus. In total, the researchers identified racial differences in 20 of the 29 immune-related adverse event types examined, indicating that the pattern is not confined to a handful of toxicities but pervades the entire spectrum of immunotherapy side effects.

Severity may differ as well. Because the real-world datasets do not natively record adverse event grades, the team used systemic steroid prescriptions as a proxy, following American Society of Clinical Oncology guidelines that recommend systemic steroids as first-line therapy for immune-related adverse events of at least grade 2. Among patients who experienced an adverse event, Black patients had the highest proportion of steroid exposure — 42.3 percent, compared with 39.9 percent for Hispanic and 39.2 percent for white patients — differences that reached statistical significance. The authors caution that drug claims do not specify why steroids were given, so the measure cannot pin severity to any particular toxicity, but it suggests that Black patients may shoulder a greater burden of serious, treatment-altering side effects.

The team also worked to rule out alternative explanations. A concern with any claims-based study is that differences in reporting or care patterns, rather than biology, might drive the results. To test this, they classified providers by patient composition and compared the average number of adverse events recorded per patient: providers predominantly serving white patients averaged 2.89 events per patient versus 2.88 for those predominantly serving non-white patients — no significant difference. Sensitivity analyses adjusting for comorbidities, neighborhood-level socioeconomic indicators and prior cancer treatments preserved the findings, with 89.1 percent of re-tested associations keeping the same direction of effect, a concordance significantly exceeding chance. Still, the authors are careful to note that the study is retrospective and observational, reflecting associations rather than proven causal relationships, and that residual confounding cannot be excluded.

Perhaps the most important caveat concerns what race actually represents. The authors emphasize that race is a social construct rather than a strictly biological variable, and that the observed disparities may reflect a tangle of influences: genetic variation in immune regulation, differences in tumor-immune interactions, unequal healthcare access, environmental exposures, surveillance intensity and the enduring effects of structural racism on diagnosis and treatment. They also note the limits of their data — the insurance database records race and ethnicity as a single combined variable, patients outside the four largest groups were too few to analyze, and the findings may apply most directly to patients covered by the insurer studied. Translating population-level patterns into individual clinical decisions, they warn, must be done cautiously to avoid entrenching the very disparities the work exposes.

What the study makes unmistakably clear is that the current system is flying blind. With 87.3 percent of checkpoint inhibitor trials on ClinicalTrials.gov conducted in the United States, where white patients dominate enrolment — even the landmark KEYNOTE-001 pembrolizumab trial, with 1,260 enrollees, counted only 63 Hispanic and 26 Black participants in total — safety signals in minority populations are systematically under-represented. Real-world datasets with millions of patients may be the only practical way to detect these differences. The authors call for racial diversity in both clinical trials and post-marketing surveillance, and for adverse events to be systematically recorded in a race-stratified manner. As immunotherapy reaches a growing share of the roughly 90,000 insured patients now treated annually, ensuring that its benefits and risks are understood — and managed — equitably across all populations has become not just a scientific imperative but a matter of basic fairness in cancer care.

Subject of Research: Racial disparities in immune-related adverse events during immune checkpoint inhibitor cancer therapy

Article Title: Disparities in adverse events during cancer immunotherapy

Article References: Liu, Y., Luo, M., Yang, J., Chen, Y., Zhang, P., Su, J., Huang, K., Salem, J.-E., Diao, L., Bian, J., Ruppin, E., Johnson, D. B., Moslehi, J. J., Han, L., on behalf of N3C, & Chute, C. G. (2026). Disparities in adverse events during cancer immunotherapy. Nature Health. https://doi.org/10.1038/s44360-026-00201-2

Image Credits: AI Generated

DOI: 10.1038/s44360-026-00201-2

Keywords: cancer immunotherapy, immune checkpoint inhibitors, immune-related adverse events, racial disparities, health equity, clinical trials, real-world data, PD-1 inhibitors, arthritis, pharmacovigilance, electronic health records, oncology