Serodifferent relationships, in which one partner lives with HIV and the other does not, represent a critical frontier in HIV prevention. The HIV-negative partner in such a couple faces ongoing exposure risk, and oral PrEP has emerged as one of the most effective biomedical tools for reducing that risk when taken consistently. The study, led by Aromorach Vivian Calo of Makerere University’s Department of Clinical Epidemiology and Biostatistics, together with colleagues from Makerere and the Infectious Diseases Institute in Kampala, set out to quantify how well HIV-negative partners in Gulu District actually adhere to their PrEP regimens and to identify the factors that shape adherence behavior.

The research team recruited participants consecutively from Gulu Regional Referral Hospital, the TASO Gulu Clinic, Awach Health Centre IV, and Layibi Techo Health Centre III between July 1 and August 31, 2024, with proportional allocation across the four sites. Data were collected using a modified AIDS Clinical Trials Group adherence questionnaire, a validated instrument widely used in antiretroviral research. Because the observed prevalence of good adherence was high, the investigators applied modified Poisson regression with robust standard errors, a statistical approach designed for common outcomes that conventional logistic regression can distort, reporting adjusted prevalence ratios with 95 percent confidence intervals.

The headline figure of 92.2 percent good adherence, with a 95 percent confidence interval of 89.4 to 94.3, is notable in a field where self-reported adherence to daily oral PrEP often falls well below optimal levels. Oral TDF/FTC requires consistent daily intake to achieve the protective drug concentrations in mucosal tissues that block HIV acquisition, and real-world adherence has been a persistent challenge across many settings. The Gulu findings suggest that, at least among this population of serodifferent couples attending regular HIV care facilities, oral PrEP is being taken reliably by the vast majority of users.

Three factors emerged as significantly associated with good adherence. The strongest was high perceived risk of HIV acquisition, which increased the prevalence of good adherence by 54 percent (adjusted prevalence ratio 1.54; 95 percent CI 1.23 to 1.95; p less than 0.001). Participants who believed they were at substantial risk of contracting HIV from their partner were markedly more likely to take their pills as directed. Tertiary education was also linked to better adherence (aPR 1.04; 95 percent CI 1.00 to 1.05; p equals 0.03), and the absence of fear of side effects raised adherence prevalence by 19 percent (aPR 1.19; 95 percent CI 1.02 to 1.39; p equals 0.025).

Two factors worked in the opposite direction. The absence of family support was associated with a 9 percent lower prevalence of good adherence (aPR 0.91; 95 percent CI 0.85 to 0.97; p equals 0.012), while forgetfulness reduced adherence prevalence by 5 percent (aPR 0.95; 95 percent CI 0.92 to 0.99; p equals 0.018). These findings point to the social and cognitive architecture underlying pill-taking behavior: people who have supportive family environments and who can remember their daily dose are more likely to maintain the regimen over time.

The emphasis on perceived risk is particularly instructive from a public health standpoint. PrEP works only if people who need it take it, and people take it only if they believe they need it. In serodifferent relationships, the HIV-negative partner’s risk perception can fade over time, especially when the HIV-positive partner is on effective antiretroviral therapy and has a suppressed viral load. The study’s authors suggest that enhanced counseling to sustain risk perception is a key intervention target, alongside strategies to manage side effects and adherence support tools such as reminders.

The finding on family support adds a social dimension to adherence science that is often overlooked in biomedical framing. In Ugandan communities, where extended family networks play a central role in daily life, the presence or absence of supportive family members can influence whether an individual maintains a daily medication routine. The authors call for greater family involvement in PrEP adherence support, an approach that could involve educating family members about PrEP and encouraging them to help remind and motivate the HIV-negative partner.

Forgetfulness, the other negative factor, is a well-documented barrier to daily oral PrEP adherence globally. The study’s recommendation of adherence support tools such as reminders aligns with a growing body of evidence that digital and behavioral reminder systems can improve medication adherence across many chronic disease contexts. Simple interventions, from mobile phone text messages to pill organizers, could help address this cognitive barrier in resource-limited settings.

The study arrives at a moment of rapid expansion in the HIV prevention landscape. While oral TDF/FTC remains an important PrEP option, the field now includes the dapivirine vaginal ring and long-acting injectable cabotegravir, and in 2025 the World Health Organization recommended twice-yearly injectable lenacapavir as an additional PrEP option. The authors note that strengthening adherence monitoring systems and preparing for the integration of these long-acting options will be critical for sustaining HIV prevention efforts in Uganda. For now, the Gulu findings offer a measure of reassurance that oral PrEP, when embedded in supportive clinical and family environments, can achieve high adherence among the people who need it most.

Subject of Research: HIV Prevention

Article Title: Prevalence and factors associated with pre-exposure prophylaxis adherence among HIV-negative partners in serodifferent relationships attending selected health facilities in Gulu District, Uganda

Article References: Calo, A. V., Ngabirano, T., Mujugira, A., Kalyango, J., & Okwero, M. (2026). Prevalence and factors associated with pre-exposure prophylaxis adherence among HIV-negative partners in serodifferent relationships attending selected health facilities in Gulu District, Uganda. BMC Public Health. https://doi.org/10.1186/s12889-026-29931-4

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29931-4

Keywords: PrEP adherence, serodifferent couples, HIV prevention, Uganda, oral TDF/FTC, prevalence, modified Poisson regression, family support, perceived HIV risk, side effects, Gulu District, antiretroviral prophylaxis

The core problem is one of scale and scarcity. Regulatory agencies collect millions of reports of adverse drug events, but the vast majority of possible drug pairs have never been systematically tested together, and most potential side effects of those pairs remain unknown. Clinical trials typically involve a restricted number of participants and a limited set of tested drug combinations, so rare or unexpected interactions slip through undetected until the drugs reach the market and are used by large, diverse populations. Patients frequently take five or more medications simultaneously, particularly the elderly and those with chronic conditions, which means the number of relevant drug combinations grows combinatorially. Exhaustive experimental testing of all these pairs is practically impossible, which is precisely why computational prediction has become such an active area of research.

DCSE approaches this challenge by learning what the authors call latent signatures for drugs, drug pairs, and side effects. In machine learning terms, a latent signature is a compact numerical representation, a vector of learned features that captures the essential characteristics of an entity without requiring humans to define those characteristics manually. For a drug, such a signature might implicitly encode information about its molecular structure, its known targets in the body, and the side effects it has already been observed to cause. For a drug pair, the signature encodes the interaction profile of the combination, and for a side effect, it encodes the biological circumstances under which that effect tends to appear. The model then combines these learned representations to estimate the probability that a particular side effect will occur with a particular drug combination. This probabilistic framing is important, because it allows the system to rank predictions by confidence rather than issuing binary verdicts.

Architecturally, the method belongs to the family of embedding based approaches that have transformed recommendation systems and knowledge graph completion. Just as a streaming service learns vector representations of viewers and films to predict what someone will watch next, DCSE learns vector representations of drugs and adverse events to predict which adverse events belong with which drug pairs. The training signal comes from databases of known polypharmacy side effects, where each recorded association between a drug pair and a side effect nudges the learned signatures so that the model’s predicted probability for that association rises. Associations that are absent from the data push predicted probabilities down, teaching the model not only what tends to co-occur but also what tends not to. Over many training iterations, the geometry of the latent space comes to reflect the structure of drug interaction biology, with drugs that behave similarly in combinations positioned near one another and side effects with overlapping mechanisms clustered together.

Where the paper makes its most consequential contribution, however, is not in the model itself but in the evaluation protocol. The authors first ran DCSE through the experimental settings commonly adopted in the literature, and it performed well. But they then identified a fundamental flaw in those settings: they rely on balanced testing datasets and sampled negative examples. In a balanced dataset, the number of known positive associations and artificially constructed negative examples is roughly equal, which makes classification dramatically easier. In reality, the space of unknown side effects is overwhelmingly larger than the space of known ones, and the unknowns are not randomly distributed; they are structured by drug classes, shared mechanisms, and reporting biases. A model that looks impressive on balanced, randomly sampled benchmarks may collapse when confronted with the true, highly imbalanced distribution of uncertainty that clinicians and regulators actually face.

To address this, the researchers designed prospective evaluations that simulate how a prediction system would be used in practice. They trained the model exclusively on data available before a fixed cutoff and then asked it to predict side effects that were only reported between 2009 and 2014, years after the training data ends. This temporal split is a much more honest test, because it mimics genuine forecasting rather than retrospective pattern matching. Within this prospective framework, the authors distinguished two scenarios of practical importance. In warm-start scenarios, some side effects are already known for a given drug pair, and the model must predict the additional effects that have not yet been documented. In cold-start scenarios, the model faces drug pairs for which no side effect information exists at all, the hardest and most clinically valuable case, since these are exactly the combinations about which physicians currently know nothing.

The results of these realistic tests were striking. Across both warm-start and cold-start settings, DCSE consistently outperformed state-of-the-art methods, demonstrating robustness and efficacy under conditions that more closely resemble real-world deployment. The consistency across scenarios matters as much as the raw performance figures. A model that excels only when it has partial information about a drug pair is of limited use, because the pairs that matter most urgently are often the ones with no prior data. By maintaining its advantage even in cold-start conditions, DCSE showed that its learned drug signatures generalize beyond the specific combinations seen during training, presumably because those signatures capture transferable properties of the drugs themselves rather than memorizing pair-specific associations.

The methodological critique embedded in this work has implications well beyond a single model. If the field’s standard benchmarks systematically overstate performance by using balanced data and sampled negatives, then many published claims about side effect prediction may not survive contact with reality. Prospective, temporally split evaluation is common in fields like weather forecasting and financial modeling, but it has been slower to take hold in biomedical machine learning, where the convenience of static, randomly split datasets has encouraged a kind of optimistic self-assessment. By demonstrating both the problem and a workable alternative, the authors provide a template that other groups can adopt, potentially raising the bar for the entire subfield of polypharmacy side effect prediction.

The clinical potential of reliable combination side effect prediction is considerable. Prescribing physicians could consult such a system before adding a new medication to a patient’s regimen, receiving ranked warnings about the adverse effects most likely to emerge from the combination. Drug safety teams could prioritize which emerging reports to investigate, and trial designers could use predictions to decide which combinations warrant formal testing. None of this replaces clinical judgment or pharmacovigilance, and the authors are careful to frame DCSE as a decision support tool rather than an oracle. But in a healthcare system where adverse drug reactions impose a substantial burden of hospitalizations and deaths, even a modest improvement in anticipating interactions could translate into meaningful patient benefit.

The study also illustrates a broader lesson about building machine learning for medicine: the evaluation is as important as the algorithm. A model is only as trustworthy as the conditions under which it has been tested, and benchmarks that flatter a method in the laboratory can conceal weaknesses that appear the moment it meets unstructured, real-world data. By training on the past and predicting the future, Jimenez and Paccanaro have given the field both a stronger tool and a more honest yardstick. If subsequent work adopts their prospective framework, the next generation of side effect predictors may earn the confidence that earlier benchmarks granted too easily, and patients taking multiple medications may be the ultimate beneficiaries of that rigor.

Subject of Research: Machine learning prediction of polypharmacy drug combination side effects

Article Title: Robust prediction of drug combination side effects in realistic settings

Article References: Jimenez, R., & Paccanaro, A. (2026). Robust prediction of drug combination side effects in realistic settings. PLOS Computational Biology, 22(10), e1013619. https://doi.org/10.1371/journal.pcbi.1013619

Image Credits: AI Generated

DOI: 10.1371/journal.pcbi.1013619

Keywords: drug combinations, side effects, polypharmacy, machine learning, DCSE, latent signatures, pharmacovigilance, cold-start prediction, prospective evaluation, PLOS Computational Biology, drug safety, adverse drug events

The research team, led by Simge Ünal Işık of the Department of Dermatology and Venereology at Uşak University Faculty of Medicine, together with colleagues from Kyrenia Dr. Akçiçek State Hospital and Uşak Training and Research Hospital, recruited 312 patients diagnosed with moderate-to-severe acne vulgaris between June and December 2025. Participants completed a five-part questionnaire that included the newly developed 11-item Isotretinoin Phobia Scale. The scale demonstrated strong internal consistency, with a Cronbach’s alpha of 0.86, a statistical measure indicating that the items reliably capture the same underlying construct of fear and avoidance related to the drug.

The findings reveal that apprehension about isotretinoin is nearly universal among this patient population, even if it reaches clinically significant intensity in a smaller subgroup. Of the 312 participants, who had a mean age of 20.16 years with a standard deviation of 3.96, 45.2 percent recorded low phobia scores, 50.3 percent showed moderate levels, and 4.5 percent exhibited high phobia. Using receiver operating characteristic analysis, the researchers established a cut-off score of 21.5 for clinically significant phobia, an approach that yielded an area under the curve of 0.806, indicating good discriminatory ability in distinguishing patients whose fears are severe enough to influence clinical decisions.

One of the more counterintuitive results concerns education. Higher educational attainment correlated positively with phobia scores, with a Spearman correlation coefficient of 0.191 and a p-value below 0.001. Rather than reassuring patients, greater formal knowledge appears to be associated with heightened concern, a pattern the authors’ data suggest may reflect exposure to alarming information about the drug’s side-effect profile rather than a balanced appraisal of risk. This finding echoes earlier research on medication phobias, including studies of corticosteroid phobia in atopic dermatitis, where fear frequently outpaces documented danger.

Experience, by contrast, seems to be an antidote. Patients who had previously used isotretinoin showed significantly lower phobia scores than those who had never taken the drug, with the difference reaching statistical significance at p = 0.003. This inverse relationship between direct exposure and fear suggests that much of the anxiety surrounding isotretinoin is anticipatory, built on secondhand accounts and online narratives rather than lived experience. It also implies that first-time patients represent the group most in need of structured counseling, since their fears have not yet been tested against reality.

The clinical consequences of significant phobia were substantial. Patients scoring above the 21.5 cut-off were more likely to refuse treatment altogether, reported impaired quality of life, and displayed increased healthcare-seeking behavior. Notably, however, phobia was not associated with treatment discontinuation among those who did start therapy. In other words, fear acts primarily as a gatekeeping barrier: it prevents patients from beginning isotretinoin, but once treatment is underway, the same fears rarely cause them to stop. This distinction matters for clinicians, because the window of opportunity for intervention lies before the prescription is written.

The study also identified which specific concerns dominate the phobic response. Behavioral hesitation, social influence, and fear of side effects emerged as the most common drivers of anxiety. Social influence is particularly noteworthy in the era of social media, where anecdotal reports of adverse effects can spread rapidly and shape patient perceptions before any medical consultation takes place. Prior work has even explored social media platforms as surveillance tools for monitoring isotretinoin side effects, underscoring how deeply online discourse has penetrated the therapeutic relationship. Related studies have documented similar concerns among parents of adolescents and among university students in Switzerland, indicating that isotretinoin hesitancy is a cross-cultural phenomenon.

To identify the factors predicting clinically significant phobia, the researchers employed logistic regression alongside chi-square tests and correlation analyses. The statistical framework allowed them to separate demographic and experiential variables that independently predict high fear from those that do not. The resulting model, anchored by the validated cut-off score, gives dermatologists a straightforward triage tool: a patient scoring above 21.5 on the 11-item scale can be flagged for intensified counseling, while those below the threshold may proceed with standard education about the drug’s risk profile.

The broader significance of the study lies in reframing treatment refusal not as irrational stubbornness but as a measurable, predictable, and modifiable psychological barrier. Isotretinoin’s adverse-effect profile, which includes mucocutaneous dryness, lipid changes, and stringent teratogenicity precautions, is well documented, and legitimate monitoring requirements exist. Yet the gap between perceived and actual risk has long frustrated dermatologists, particularly because untreated moderate-to-severe acne can leave permanent scarring and lasting psychological damage. Earlier research has shown that acne relapse and the need for retrial with isotretinoin remain common clinical challenges, making first-line adherence all the more valuable.

The authors conclude that isotretinoin phobia represents a significant barrier to effective acne management and that early recognition is crucial. By deploying the Isotretinoin Phobia Scale at the point of care, clinicians could detect at-risk patients before refusal hardens into a decision, then direct targeted counseling toward the specific concerns, whether behavioral hesitation, social pressure, or fear of side effects, that drive each patient’s anxiety. Given that previous users report markedly less fear, structured education that sets accurate expectations may convert the most anxious first-time patients into confident ones, improving adherence, protecting quality of life, and ultimately preventing the long-term complications of untreated severe acne.

Subject of Research: Isotretinoin phobia and its clinical impact in moderate-to-severe acne vulgaris

Article Title: Fear of isotretinoin in moderate-to-severe acne: clinical impact and a new assessment approach

Article References: Ünal Işık, S., Yüksek, T., Koç Yıldırım, S., & Işık, L. (2026). Fear of isotretinoin in moderate-to-severe acne: clinical impact and a new assessment approach. Archives of Dermatological Research, 318(1), Article 441. https://doi.org/10.1007/s00403-026-04905-z

Image Credits: AI Generated

DOI: 10.1007/s00403-026-04905-z

Keywords: isotretinoin, acne vulgaris, phobia, dermatology, treatment adherence, quality of life, patient counseling, Isotretinoin Phobia Scale, side effects, medication hesitancy, cross-sectional study, skin health

The study, led by Sarah Boafowaa Owusu with colleagues Konlan Kennedy Dodam and Lydia Aziato, used an exploratory descriptive qualitative design. The participants were women diagnosed with breast cancer who had undergone radiotherapy at Komfo Anokye Teaching Hospital in Kumasi, one of Ghana’s major referral centres for cancer care. Each woman took part in an in-depth interview guided by a piloted interview guide, meaning the questions had been tested and refined before the main data collection began. The interviews were audio-recorded, transcribed word for word, and then analysed using thematic analysis with the assistance of NVivo 10, a software package designed to help researchers organise and code large volumes of qualitative text.

The analysis revealed two overarching themes: physical experiences and social experiences. Within the physical domain, the researchers identified four sub-themes: pain, breast and skin changes, physical exhaustion, and gastrointestinal symptoms. These are not minor inconveniences. The title of the paper itself, drawn from a participant’s words, captures the daily reality many of these women described: after radiation sessions, they sometimes felt weak and dizzy. Such systemic symptoms, layered on top of localised effects at the treatment site, shaped how the women judged the entire treatment experience and whether they would encourage others to follow the same path.

The physical findings carry particular weight in the Ghanaian context. Breast cancer remains one of the leading causes of illness and death among women worldwide, and radiotherapy, alongside surgery and chemotherapy, is a well-established treatment option. Yet the authors note that many women in low-resource settings refuse these allopathic treatments precisely because of adverse effects that remain poorly identified and poorly managed. In other words, the problem is not only that side effects occur, but that health systems often fail to detect, acknowledge, and treat them effectively. When a woman’s suffering goes unrecognised, her experience becomes a cautionary tale told to sisters, friends, and neighbours, and the cycle of treatment refusal deepens.

The biology of these side effects is well understood in radiation oncology. Ionising radiation damages cancer cells by breaking their DNA, but it also injures surrounding healthy tissue. Skin in the treated field can become red, dry, itchy, and in more severe cases blistered, a condition known as radiation dermatitis. Fatigue is among the most commonly reported effects of radiotherapy across all cancer types, and dizziness and weakness, as the study participants described, can compound the exhaustion. Gastrointestinal symptoms such as nausea can arise when radiation fields affect or lie near the upper abdomen, and radiation-induced nausea and vomiting are recognised clinical entities in the pharmacological literature. Pain at the treatment site and changes in the appearance and texture of the breast add a further layer of physical and emotional burden.

What makes the new study distinctive is not the catalogue of symptoms itself but the way those symptoms ripple outward into women’s decisions and their communities’ attitudes. The researchers found that several of the physical experiences reported after radiation therapy deter women diagnosed with breast cancer from consenting to radiotherapy, or from recommending it as a treatment option to others in Ghana. In a health system where late presentation of breast cancer is already a documented problem, and where studies from across Sub-Saharan Africa have shown that women are often diagnosed and treated late, any factor that discourages acceptance of effective treatment has serious consequences for survival.

Yet the study is not only a story of suffering. The second major theme, social experiences, contained three sub-themes: family and partner support, support from employers, churches, and friends, and sexual life challenges. The women’s social networks emerged as a genuine resource. Family members and partners who provided practical and emotional support helped women cope with the demands of daily treatment trips and the physical aftermath. Employers, churches, and friends also played roles in sustaining women through the treatment period, offering encouragement, material help, and a sense that they were not facing the disease alone. The authors argue that these networks could serve as useful motivators in promoting radiotherapy, turning social capital into a clinical asset.

At the same time, the study does not shy away from the difficulties embedded in social life after treatment. Sexual life challenges formed one of the identified sub-themes, reflecting a dimension of post-treatment experience that is frequently overlooked in busy oncology clinics but that shapes intimate relationships and self-image. Research elsewhere in Africa has documented how changes in body image following breast cancer treatment affect women’s sense of identity and their relationships, and the Ghanaian findings align with that broader picture. The authors also build on their own earlier work in Kumasi, which examined the psycho-spiritual wellbeing and coping strategies of women after radiation therapy, indicating a sustained research programme into how Ghanaian women live with and through this treatment.

The recommendations that flow from the study are directed at two levels of the health system. First, the authors call on nurses, midwives, and other health professionals to handle the negative physical symptoms experienced after radiation therapy with tact and professionalism through supportive care services, with the aim of improving acceptance of this treatment option. Supportive care, the branch of oncology concerned with symptom management and quality of life, is thus framed not as an optional extra but as a strategic necessity: if side effects are managed well, women will be more willing to start and complete radiotherapy, and more willing to speak positively about it to others. Second, the researchers recommend that the Ghana Health Service ensure nurses at breast cancer care units utilise the social networks of women undergoing radiation therapy as social capital to reduce the psychological stress associated with the treatment.

The broader significance of the study lies in its method and its setting. Qualitative research of this kind cannot measure how common each symptom is across the population, and the authors are careful to present the findings as the experiences of seventeen women at a single teaching hospital. But what such research can do, and does here, is reveal the texture of lived experience that surveys miss: the weakness and dizziness after a session, the fear of a changing breast, the strain on intimate life, and the quiet power of a supportive church community or an understanding employer. In resource-constrained settings, where oncology services are concentrated in a handful of teaching hospitals and follow-up care can be fragmented, understanding these experiences is a prerequisite for designing services that women will actually use. The study suggests a practical path forward: manage the physical side effects with skill and compassion, and mobilise the families, faith communities, and friendships that already surround these women. If those two strands are woven together, the authors argue, radiotherapy in Ghana could move from being a feared ordeal to an accepted and better tolerated part of breast cancer care.

Subject of Research: Physical and social experiences of Ghanaian women with breast cancer after radiotherapy

Article Title: “After radiation sessions, I sometimes feel weak and dizzy’’: Experiences of Ghanaian women with breast cancer post radiotherapy

Article References: Owusu, S. B., Dodam, K. K., & Aziato, L. (2026). “After radiation sessions, I sometimes feel weak and dizzy’’: Experiences of Ghanaian women with breast cancer post radiotherapy. Supportive Care in Cancer, 34(10), Article 1064. https://doi.org/10.1007/s00520-026-11304-7

Image Credits: AI Generated

DOI: 10.1007/s00520-026-11304-7

Keywords: breast cancer, radiotherapy, Ghana, qualitative research, supportive care, side effects, nursing, social support, Sub-Saharan Africa, cancer treatment, radiation side effects, women's health

The research team, led by Sarah Jane Smith of the Centre for Dementia Research at Leeds Beckett University, recorded consultations involving ten clinicians, twenty-nine patients and thirty-two relatives or supporters. This is a technically demanding kind of study to pull off. Recording real clinical encounters requires ethical approval, patient consent and careful handling of sensitive moments in people’s lives, and the study was approved by the National Research Ethics Service committee for Yorkshire and The Humber in March 2021, with adaptations for verbal and audio-recorded consent during the COVID-19 pandemic. Once recorded, the consultations were subjected to content analysis, a systematic method in which researchers code the talk against predefined categories to identify patterns in what is communicated, how often, and in what form. The approach allows researchers to move beyond anecdote and quantify precisely how treatment information is being delivered in routine practice.

The most consistent finding concerned how benefits were presented. In twenty-six of the twenty-nine consultations, clinicians framed the purpose of dementia medication as symptom management, and within that framing the dominant emphasis was on memory and cognition. That makes clinical sense, because the pharmacological treatments available for dementia, such as the cholinesterase inhibitors commonly prescribed for Alzheimer’s disease, are licensed to treat symptoms rather than to halt the underlying neurodegenerative disease. Yet the consistency ended there. In only thirteen of the twenty-nine consultations did clinicians explicitly state that the treatment was not curative, meaning that in more than half of the encounters patients may have left the room without a clear verbal statement that the drug could not stop or reverse their condition.

More surprising still was the language used to describe what the drugs might achieve beyond symptoms. In roughly one third of consultations, clinicians at some point suggested that treatment might be disease modifying, a claim that overstates the established evidence base for currently licensed dementia medications. Around half of the consultations referred to symptom reduction or delay. The researchers characterise this overall pattern as positive framing of treatment benefits, and the technical concern underlying it is well recognised in health communication research: when clinicians use leading or optimistic language about a therapy’s effects, patients can develop expectations that outstrip what the evidence supports, which in turn complicates decisions about continuing or stopping medication when benefits fail to materialise.

Side effects, by contrast, were discussed almost universally, appearing in twenty-six of the twenty-nine consultations. But the manner of that discussion revealed its own patterns. Clinicians frequently deployed verbal qualifiers, small linguistic hedges that signal the certainty or likelihood of an effect, such as indicating that something happens only occasionally or might occur. Qualifiers are not inherently problematic; they are a natural way of conveying probability in everyday speech. The difficulty arises when they are applied inconsistently, because a patient hearing that a side effect is possible in one consultation and unlikely in another receives materially different information about the same drug.

Perhaps the most consequential finding was how side effects were characterised when they were raised. The study found that the side effect profile was commonly described in minimising terms, presented as not serious or downplayed relative to the potential benefits. For patients and families weighing whether to start a medication, this framing matters enormously. Nausea, loss of appetite, vivid dreams and other recognised effects of dementia drugs can be distressing enough to prompt discontinuation, and a patient who has been told the effects are minor may feel misled when they experience them. The authors also note that explicit discussions weighing risks against benefits, the kind of balanced deliberation that shared decision-making ideally requires, occurred in only half of the consultations.

There was at least one area where communication fell short in a more mechanical way. In thirteen consultations, clinicians explained how the treatments actually work, the mechanism of action, and in eight of those the explanation was delivered in a patient-friendly way. That means fewer than a third of all consultations included an accessible account of the drug’s mechanism. For a population in which cognitive impairment is the defining feature, understanding how a medication is supposed to act is not a luxury; it is foundational to adherence, to monitoring for effects, and to making sense of follow-up appointments where treatment continuation is decided.

The variability the researchers documented is not simply a matter of individual clinician style. Because the consultations took place across two NHS Trusts with ten different clinicians, the inconsistency suggests an absence of shared standards for how dementia treatment information should be conveyed. The authors respond with three concrete recommendations. First, clinicians should exercise caution about using leading language that frames benefits optimistically without evidential warrant, including suggestions that drugs may modify the disease. Second, communication about both benefits and side effects should be consistent and transparent, so that patients receive comparable information regardless of which clinician they see. Third, evidence-based patient decision aids should be employed where appropriate, giving patients structured, balanced written or visual support to complement the spoken consultation.

The stakes of getting this right are considerable. Decisions about whether to take dementia medication are made at a moment when patients and families are absorbing a life-changing diagnosis, and the study’s authors point out that these discussions have a direct impact on the decision to take treatments, on understanding of the medications, and on adherence over time. A patient who believes a drug will slow their disease may persist with it despite troublesome side effects and no perceived benefit; a patient who was never told the treatment is not curative may face that realisation later without clinical support. Communication in the consultation is, in effect, the first dose of the treatment.

The research, funded by the National Institute for Health and Research through its Research for Patient Benefit programme, was strengthened by involvement from a patient and public involvement panel at the Centre for Dementia at Leeds Beckett. Its publication comes at a moment of intense public interest in dementia therapeutics, as new disease-modifying therapies enter clinical discussions worldwide and the pressure on memory clinics to communicate complex evidence grows. What this study demonstrates is that the infrastructure for those conversations, the everyday language of benefit and risk exchanged between doctor and patient, remains uneven. Standardising it, the authors argue, through guidance and decision aids, is a practical and achievable step toward ensuring that every patient with dementia receives the same honest, balanced account of what treatment can and cannot do.

Subject of Research: How clinicians communicate the benefits and side effects of dementia medications to patients in memory assessment clinics

Article Title: How do doctors discuss dementia treatments with patients? Content analysis of the presentation of treatment information, benefits and side effects

Article References: Smith, S. J., Griffiths, A. W., Knapp, P., Dickinson, R., Dooley, J., Russell, G., & Kelley, R. (2026). How do doctors discuss dementia treatments with patients? Content analysis of the presentation of treatment information, benefits and side effects. BMC Health Services Research. https://doi.org/10.1186/s12913-026-15657-x

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15657-x

Keywords: dementia, health communication, shared decision-making, pharmacotherapy, memory clinics, content analysis, side effects, patient outcomes, geriatric care, NHS, informed consent, clinical consultation

The MAUDE database is the FDA’s primary passive surveillance system for medical device problems. Manufacturers, importers, and user facilities are required to submit reports of device-related deaths, serious injuries, and malfunctions, while the public and health professionals may file voluntary reports. Because it aggregates real-world experience across thousands of dental sleep medicine practices, it offers a view of device safety that clinical trials, which typically follow small cohorts for limited periods, cannot match. Glick and Richardson analyzed this publicly available database to identify adverse event reports associated with oral appliances used for sleep-disordered breathing, and to examine how the categories of reported harm have trended across the reporting period.

Oral appliance therapy, in its most common form, uses a mandibular advancement device: a two-arch dental appliance that holds the lower jaw forward during sleep, tightening the upper airway and reducing collapse that drives apneic events. The approach is endorsed in clinical practice guidelines from the American Academy of Sleep Medicine and the American Academy of Dental Sleep Medicine as a legitimate therapy for patients with obstructive sleep apnea, particularly those who are intolerant of CPAP or prefer an alternative. Recent studies cited in the letter underscore why the therapy has gained ground. One 2025 analysis found that oral appliance therapy is highly efficacious at reducing sleep apnea-specific hypoxic burden, a metric predictive of cardiovascular morbidity and mortality. Another 2026 study comparing long-term outcomes of mandibular advancement devices with CPAP found meaningful therapeutic efficacy in patients with obstructive sleep apnea, reinforcing the devices’ clinical standing.

Yet the same mechanism that makes these appliances effective, mechanically protruding the mandible night after night for years, is also the source of their characteristic side effects. The dental sleep medicine literature has long catalogued a spectrum of harms: temporomandibular joint discomfort, myofascial pain, dental pain, occlusal changes in which the bite shifts as teeth move under sustained force, dry mouth, hypersalivation, and gum irritation. A 2025 systematic review and meta-analysis comparing side effects across different mandibular advancement devices found that these effects are common and vary by device design, a point that matters because the market now includes a wide range of custom, semi-custom, and over-the-counter products. The American Academy of Dental Sleep Medicine has published recommendations for managing these side effects, emphasizing monitoring and follow-up as integral parts of therapy rather than optional extras.

Against that backdrop, the rising volume of adverse event reports in MAUDE is a signal that cannot be dismissed as statistical noise, even though passive reporting systems carry well-known caveats. Report counts in such databases are influenced by many factors beyond true incidence: the size of the exposed population, growing awareness among clinicians and patients that reporting is possible, expanded marketing of oral appliances, and the rise of teledentistry pathways that deliver dental sleep medicine with fewer in-person touchpoints. The letter’s authors note the growth of teledentistry in dental sleep medicine as an innovation that requires safeguards, a nod to the possibility that remote care models, while expanding access, may change how side effects are detected, managed, and reported.

The second half of the finding, the shifting categorical trends, is arguably the more consequential one. When the composition of adverse event reports changes over time, it can indicate that the nature of device use is changing, that new device designs are introducing new failure modes, or that the patient population receiving appliances is evolving. Prior work on the MAUDE database has already explored this terrain: a 2025 analysis in Respiratory Medicine examined adverse events associated with mandibular repositioning appliances specifically, establishing that the database contains enough signal to support categorical analysis. By updating and extending that picture, Glick and Richardson provide dental sleep medicine clinicians with a current map of where reported harms are concentrating, information that can guide the follow-up questions dentists ask and the monitoring protocols they adopt.

For patients, the practical message is that oral appliance therapy is not a set-and-forget intervention. Dental sleep medicine standards published in 2025 for screening, treatment, and management of sleep-related breathing disorders in adults using oral appliance therapy stress that therapy begins with proper screening and diagnosis, continues with device fitting and titration, and depends on ongoing surveillance of both efficacy and side effects. The side effects documented in the literature, particularly the dental and skeletal changes that can accrue silently over years of use, are precisely the kind of harm that a patient sleeping alone with a remote provider may not notice until the bite has shifted. Regular in-person evaluation of the bite, the temporomandibular joints, and the airway response remains a cornerstone of safe therapy.

The study also arrives at a moment when the dental sleep medicine field is professionalizing rapidly. More dentists are seeking certification, more primary care physicians are referring patients who cannot tolerate CPAP, and more consumers are encountering oral appliances through advertising and direct-to-consumer channels. Each of these trends enlarges the population of device users, and each carries implications for the adverse event reporting stream. A database like MAUDE cannot by itself separate increased use from increased risk, which is why the authors’ framing of increasing reports alongside shifting categorical trends is careful: it documents what is being reported and how the pattern changes, rather than claiming a precise incidence rate that passive surveillance cannot deliver.

That limitation is worth stating plainly, because it defines what the study can and cannot tell the public. MAUDE reports are voluntary in part, unverified in most cases, and subject to duplicate, incomplete, and delayed submissions. The FDA itself cautions that the database cannot be used to calculate incidence. What it can do, and what this research letter does, is reveal the shape of the harm landscape: which categories of adverse events are appearing, whether reports of temporomandibular disorders, dental changes, or other complications are growing relative to one another, and whether the overall reporting trajectory is climbing. For a therapy whose user base is expanding and whose delivery models are changing, that shape is exactly the information clinicians, device designers, and regulators need to prioritize.

The authors, who analyzed the data jointly and drafted the manuscript together, report no funding for the work; Richardson discloses no conflicts of interest, while Glick discloses an ownership stake in Ember Sleep and consultancy work for the American Academy of Dental Sleep Medicine, disclosures that are relevant context for readers weighing the letter’s conclusions. The study was received in April 2026, revised in June, accepted on 1 July, and published on 22 July 2026 in the journal’s volume 22 as article number 121. Its brevity, a research letter rather than a full study, reflects the nature of the evidence base: a public database, a categorical analysis, and a clear signal that warrants prospective research. For the millions of patients who wear a mandibular advancement device every night, and for the dentists who fit them, the takeaway is not alarm but vigilance. Oral appliance therapy remains a guideline-endorsed, increasingly evidence-supported treatment for obstructive sleep apnea. The rising and shifting adverse event reports in the FDA’s own surveillance data are a reminder that its safety depends on the follow-up that happens after the appliance leaves the dental chair.

Subject of Research: Adverse event reporting trends for oral appliances used in obstructive sleep apnea therapy

Article Title: Increasing reports and shifting categorical trends in adverse events for oral appliances

Article References: Glick, A., & Richardson, S. E. (2026). Increasing reports and shifting categorical trends in adverse events for oral appliances. Journal of Clinical Sleep Medicine, 22(1), Article 121. https://doi.org/10.1007/s44470-026-00139-y

Image Credits: AI Generated

DOI: 10.1007/s44470-026-00139-y

Keywords: oral appliance therapy, obstructive sleep apnea, MAUDE database, adverse events, mandibular advancement devices, dental sleep medicine, temporomandibular disorders, FDA surveillance, teledentistry, CPAP alternatives, side effects, sleep disorders

The research, led by Tyler P. Shern, Victoria Chamberlain, Logan R. Holt and colleagues in the Division of Breast Surgery at Mass General Brigham in Boston, and published in Breast Cancer Research and Treatment, offers some of the clearest clinical evidence to date that checkpoint blockade can alter how patients experience surgical pain. The investigators retrospectively reviewed 277 consecutive patients who underwent mastectomy with immediate reconstruction after neoadjuvant chemotherapy between 2020 and 2024. Of these, 54 had received a PD-1 inhibitor as part of their neoadjuvant regimen, while 223 received chemotherapy without immune checkpoint blockade. By comparing opioid consumption measured in oral morphine equivalents, pain scores recorded in the post-anesthesia care unit, and prescription refill patterns after discharge, the team was able to quantify, for the first time in this surgical population, what laboratory scientists had suspected for years: blocking PD-1 signaling appears to blunt the effectiveness of opioid analgesia.

The biology behind this finding is as intriguing as the clinical result. PD-1 and its ligand PD-L1 are best known as immune checkpoints, but research over the past decade has revealed that the PD-1/PD-L1 axis also operates within the nervous system. PD-L1 expressed on neurons suppresses nociceptive signaling, dampening both acute and chronic pain, and this natural analgesic system appears to work in concert with opioid receptors. Crucially, PD-1 signaling has been shown to modulate the activity of the µ-opioid receptor, the molecular target of morphine and most perioperative opioids. Animal studies have demonstrated that anti-PD-1 treatment can significantly impair opioid antinociception, and a 2020 study in Science Translational Medicine found that anti-PD-1 therapy reduced opioid effectiveness not only in rodents but also in nonhuman primates. The new clinical data suggest this phenomenon translates to human surgical patients as well.

The numbers from the Boston cohort are striking. Patients who had received neoadjuvant chemotherapy with a PD-1 inhibitor required an average of 3.5 oral morphine equivalents per hour in the post-anesthesia care unit, compared with 2.1 per hour in patients who received chemotherapy alone, a difference that was statistically significant. Their maximum pain scores in the immediate recovery period were also higher, averaging 6.3 on the standard numeric pain scale compared with 5.4 for the control group. The disparities did not end at discharge. Nearly 39 percent of patients in the PD-1 group required an opioid refill after leaving the hospital, versus only 17 percent of those who had not received the immunotherapy. In practical terms, more than one in three patients exposed to PD-1 blockade needed additional opioids to manage their pain, more than double the rate of their peers.

To ensure that these differences were not simply artifacts of patient selection, the researchers performed multivariable statistical analyses adjusting for a range of clinical and procedural variables. They found that PD-1-based neoadjuvant therapy independently predicted higher opioid use per hour in the recovery unit, alongside breast volume greater than 700 cubic centimeters and the use of preoperative analgesic medications. For post-discharge refills, PD-1 therapy was by far the strongest predictor, increasing the odds of needing a refill by roughly three and a half times, while axillary lymph node dissection also independently increased refill likelihood. Interestingly, the two groups were comparable in their preoperative analgesic use and intraoperative opioid administration, meaning that the divergence in pain and opioid consumption emerged precisely when patients awakened from surgery, the moment when the interplay between immune signaling and analgesic response would be expected to matter most.

The study population does introduce some complexity in interpretation. Patients in the PD-1 group were more likely to carry BRCA1 or BRCA2 mutations, with rates of nearly 39 percent compared with just over 7 percent in the control group, reflecting current trial designs that enroll germline mutation carriers into immunotherapy protocols. They were also more likely to undergo bilateral surgery, at 92.6 percent versus 67.7 percent, and sentinel lymph node biopsy, both of which are associated with greater surgical burden and pain. However, the statistical models accounted for these imbalances, and the independent association between PD-1 exposure and increased opioid requirement persisted. This careful adjustment strengthens the argument that the immunotherapy itself, rather than the biology of the tumors or the extent of surgery, drove the differences in analgesic consumption.

The findings arrive at a moment when immunotherapy is expanding rapidly across the breast cancer landscape. Landmark trials such as KEYNOTE-522 established pembrolizumab plus chemotherapy as standard care for high-risk early triple-negative breast cancer, and more recent phase 3 studies have extended the approach to estrogen receptor-positive disease, with nivolumab and pembrolizumab demonstrating event-free survival benefits in randomized trials. Additionally, novel combinations pairing PD-1 inhibitors with PARP inhibitors such as niraparib in BRCA-mutated breast cancer are advancing through clinical development. As the population of patients arriving in the operating room after PD-1 exposure grows, anesthesiologists and surgeons will increasingly encounter the analgesic challenge this study documents, often without realizing that their patient’s immunotherapy history may be reshaping their pain response.

The clinical implications extend beyond the recovery room. Enhanced recovery after surgery protocols and multimodal analgesia strategies emphasize minimizing opioid exposure, and prior work from the same surgical group has shown that mastectomy with reconstruction can be safely performed in ambulatory settings with low opioid requirements. If PD-1 inhibitors undermine the efficacy of opioid-based analgesia, perioperative teams may need to lean more heavily on non-opioid modalities, including regional nerve blocks, acetaminophen, nonsteroidal anti-inflammatory drugs, and agents such as gabapentinoids, to achieve adequate pain control without escalating opioid doses that may be pharmacologically less effective and carry their own risks. Conversely, the study also raises questions flowing in the other direction: opioids themselves have immunosuppressive properties, and preclinical research suggests morphine can blunt responses to immune checkpoint inhibitors, creating a potentially vicious cycle in which immunotherapy reduces opioid efficacy while opioids may undermine immunotherapy.

The researchers are careful to frame their results within the limitations of a retrospective, single-institution design. The PD-1 group was relatively small, at 54 patients, and the study relied on chart review rather than prospective pain assessments. Confounding by unmeasured variables, such as differences in tumor biology or reconstruction technique, cannot be entirely excluded. Nonetheless, the biological plausibility provided by the neuroimmunology literature, combined with concordant findings from other surgical populations, including increased perioperative opioid consumption reported in patients receiving neoadjuvant immunotherapy for esophageal and non-small-cell lung cancers, lends considerable weight to the conclusions. The consistency across tumor types suggests that impaired opioid analgesia may be a class effect of checkpoint blockade rather than a curiosity specific to breast surgery.

For now, the most important message is one of awareness. Patients who have received PD-1 inhibitors before breast surgery are not more fragile or less resilient, but their nervous systems may respond differently to standard analgesic protocols. Anesthesiologists planning perioperative care, surgeons counseling patients about recovery expectations, and oncologists sequencing systemic therapy all stand to benefit from knowing that checkpoint blockade can influence pain pathways in ways that extend far beyond the tumor microenvironment. As immunotherapy continues its march into earlier stages of cancer treatment, the medical community will need to integrate a more complete picture of these drugs’ physiological reach, one that encompasses not only tumor rejection and immune-related adverse events, but also the everyday, deeply human experience of postoperative pain.