The scale of the problem is considerable. In 2022, an estimated 1,300,196 new cancer cases were reported globally among adolescents and young adults, with an annual incidence of 40.3 per 100,000 individuals. Beyond the immediate medical threat, this group faces a distinctive constellation of age-specific challenges: late effects of treatment, infertility, financial strain and psychosocial disruption to education, career progression, romantic relationships and family planning. Yet the same population also reports positive transformations, including strengthened relationships, personal growth and a greater appreciation for life. Existing conceptual models, the researchers argue, have failed to capture this complexity, largely because they were developed within single countries, narrow age bands or specialised subpopulations.

To build a more culturally sensitive framework, the team at the University of Southampton, working with collaborators including Youth Cancer Europe and the Netherlands Cancer Institute, carried out 52 semi-structured online interviews. Twenty-seven were with young people with lived experience of cancer, diagnosed between the ages of 15 and 36 and drawn from 16 different nationalities; 25 were with healthcare, allied health and other professionals from 12 nationalities, with clinical experience ranging from six months to 34 years. Interviews lasted up to 120 minutes for patients and 30 minutes for professionals, were conducted between April and June 2023, and were analysed thematically using a combination of inductive and deductive coding, following COREQ reporting guidelines and ethical approval from the University of Southampton.

Four major themes emerged from the analysis. The first concerned healthcare delivery, access, quality and navigation. Participants described how differences in the organisation of health services across European settings directly shaped their ability to reach and move through cancer care. Geographic, financial and systemic barriers were recurrent: one young woman recalled finding a surgeon in a hospital 200 kilometres from her hometown, while another described going private because her insurance would not cover treatment, only to confront long waiting lists in the public system. Transitions between paediatric and adult care were a particular flashpoint, with one surgical oncologist explaining that young adults are dropped into an adult system with far more patients and far less individual dedication.

Access to information emerged as a critical mediator of these experiences. One young man diagnosed at 18 noted that his ability to read English gave him access to information that would otherwise have been unavailable in his local language, leaving his haematologist as his only source, one he described as paternalistic and unwilling to explain things. Mental health support was another glaring gap. A consultant oncologist described patients with significant psychological problems who did not meet referral criteria, waited years on waiting lists and were eventually signposted to online resources. The researchers also documented how cultural norms compound these barriers, with young men sometimes avoiding mental health care and young women in small, conservative communities delaying intimate examinations out of shame.

The second theme, economic instability and hardship in the cost of survival, revealed how financial toxicity permeates every stage of care. Participants reported out-of-pocket expenses, coverage gaps for additional medications, and reliance on family savings or community crowdfunding. One woman diagnosed at 36 turned to Facebook fundraising when she realised she could no longer work; an 18-year-old man paid roughly €200 for tests his insurance did not cover, an amount close to an entire salary in his family. Professionals observed that some young adults return to work too early to avoid financial damage, risking deterioration of their health. Even in systems that subsidise treatment, fertility preservation costs frequently fell on patients, and employment discrimination hampered social reintegration after recovery.

The third theme exposed stark inequalities in fertility counselling and preservation. Male patients generally gain rapid access to sperm banking, often same-day, whereas female patients face harvesting procedures that can take four to eight weeks, potentially delaying cancer treatment itself. Several participants learned about preservation options too late, or by accident rather than from a clinician. Long-term storage costs added further inequity, with one man reporting that he must pay every two years to keep frozen sperm. The analysis also noted societal and cultural pressures surrounding parenthood that shape how young survivors approach family planning.

The fourth theme, described as liminal survivorship, captures the precarious in-between state after treatment ends. Many participants found support services withdrawn prematurely; one young man was removed from a charity’s care list while still on maintenance therapy, describing himself as devastated because he needed support more than ever. Others were denied disability-based accommodations at university because they were not officially classified as disabled during recovery. Legal and financial protections proved decisive: in some countries, the so-called Right to be Forgotten allows survivors to obtain mortgages, insurance or adoption rights without their cancer history counting against them, while in others survivors fear they may never secure life insurance or a home loan because of their medical records.

From these themes, the team constructed a conceptual model with four interacting layers. Structural and contextual factors, including healthcare system design, insurance coverage, policies, cultural norms and legal protections, form the outer environment. Mediators such as access to information, social support, financial coverage and system responsiveness determine how those conditions are actually experienced. These give rise to processes, including treatment delays, fragmented care, reliance on family and missed support, which in turn shape individual outcomes spanning mental and physical health, identity, autonomy, relationships, education and work trajectories and trust in healthcare. The researchers argue this cascade explains why two clinically similar patients in different countries can experience radically divergent outcomes.

The model represents a deliberate departure from earlier frameworks. Where previous AYA models, such as Fern and colleagues’ UK-based model for 13 to 25 year olds or Gray and colleagues’ survivorship model drawn from a relatively homogeneous online sample, centred on individual and healthcare-level factors, the new framework situates the young patient within a broader structural and societal environment. Its practical implications are concrete: it will guide the development of a Core Outcome Set for adolescent and young adult cancer care, inform patient-reported outcome and experience measures, and support policy reforms ranging from expanded Right to be Forgotten legislation to standardised fertility preservation reimbursement. The authors acknowledge limitations, including limited ethnic diversity, reliance on English-speaking participants recruited largely through a single charity, and under-representation of brain tumour survivors and some European healthcare systems. They call for testing and refining the model in more diverse populations, but the central message stands: improving outcomes for young people with cancer demands coordinated, multi-level action on the systems and structures surrounding them, not just the tumours within them.

Subject of Research: Development of a conceptual model of cancer care experiences among adolescents and young adults with cancer in Europe

Article Title: A conceptual model of cancer care and lived experiences in adolescents and young adults with cancer: results from the multi-national European STRONG AYA project

Article References: Collaço, N., Sodergren, S., Way, K., Košir, U., Cairns, C., van der Graaf, W., Husson, O., & Darlington, A.-S. (2026). A conceptual model of cancer care and lived experiences in adolescents and young adults with cancer: results from the multi-national European STRONG AYA project. Supportive Care in Cancer, 34(10), Article 1017. https://doi.org/10.1007/s00520-026-11170-3

Image Credits: AI Generated

DOI: 10.1007/s00520-026-11170-3

Keywords: adolescent and young adult cancer, STRONG AYA project, conceptual model, cancer care navigation, fertility preservation, financial toxicity, survivorship, qualitative research, healthcare disparities, Right to be Forgotten, psychosocial support, Europe

The research, led by Sayanta Adhikari, Vishnuprasadh Kumaravelu, and P. K. Srijith of the Bayesian Reasoning and Inference Lab at the Indian Institute of Technology Hyderabad, introduces a framework called UnCLe, short for a Hypernetwork Framework for Data-Free Unlearning and Continual Learning. The core insight is that machine unlearning—the deliberate removal of a task’s influence from a trained model—has almost always been designed with offline training in mind, where engineers retain full access to the original dataset. In continual learning, that assumption collapses. Data arrives task by task and is typically discarded after use, so when an unlearning request arrives, the original examples may simply no longer exist.

The team identified two failure modes that emerge when conventional unlearning is naively applied to continual learning environments. The first is catastrophic forgetting of retained tasks, a well-known pathology in which updating a network to remove one capability wipes out unrelated capabilities it was supposed to keep. The second is subtler and, in some ways, more troubling: catastrophic remembering, in which tasks that were supposedly unlearned resurface when the model later absorbs new information. A model that appears to have forgotten sensitive data can effectively relapse, undermining the very privacy guarantees that unlearning is meant to provide.

UnCLe attacks both problems by restructuring how the model’s parameters are produced in the first place. Instead of training a single monolithic network, the framework employs a hypernetwork—a network that generates the weights of another network—conditioned on compact task embeddings. Each task the system encounters is represented by its own embedding vector, and the hypernetwork maps that embedding to a full set of task-specific parameters. Learning a new task therefore means learning or refining an embedding, while the shared hypernetwork machinery remains stable across the entire sequence of operations.

Unlearning under this scheme becomes elegantly simple at the task level. To remove a task, the framework optimizes the hypernetwork so that, for that task’s embedding, it generates parameters that behave like noise. The generated network produces uniform, maximum-entropy outputs on the forgotten task—in other words, the model becomes maximally uncertain, exactly as if it had never seen the task at all. Crucially, this procedure does not require the original training data, which is precisely what makes it data-free. The optimization is guided by a mean squared error objective that pulls the generated parameters toward freshly sampled Gaussian noise, combined with a regularization term that anchors the hypernetwork’s outputs for all previously retained tasks, preventing collateral damage.

The choice of a noise-matching objective, rather than a direct norm penalty, turns out to matter a great deal. The authors show mathematically that averaging the squared distance to random Gaussian samples converges to the squared L2 norm of the parameters plus a constant, meaning the MSE objective implicitly drives the forgotten task’s parameters toward zero and its outputs toward a uniform distribution. But applying the L2 norm directly would, over repeated unlearning operations, drag the hypernetwork’s own shared weights toward zero and destabilize the whole system. By contrast, sampling a fresh noise target at each optimization step constrains the forget task only in distribution, acting as an implicit regularizer that preserves the shared representation. The researchers compared alternatives—including fixed noise targets, pure norm reduction, and simply discarding the task embedding—and found their approach achieved the best balance between erasing the target task and protecting retained performance.

Scaling a hypernetwork to generate all the weights of a modern convolutional backbone such as ResNet18 or ResNet50 presents its own engineering challenge, since the hypernetwork’s output layer would otherwise balloon to an impractical size. The team’s solution is chunked generation: the main network’s parameters are partitioned into roughly 200 chunks, each produced by a dedicated head of the hypernetwork conditioned on a unique chunk embedding concatenated with the task embedding. These chunk embeddings are frozen after the first task to guard against forgetting, and the final layer is split into specialized heads for weights, batch normalization parameters, and residual connection parameters, reducing redundancy and computational overhead.

The practical consequences of this design are striking. Because unlearning operates entirely in parameter space, its computational cost is dominated by the hypernetwork’s forward and backward passes and is essentially independent of dataset size and class count. The only term that grows over a sequence is the regularization over retained tasks, and it grows linearly in the number of tasks, not data points. In conventional replay-based unlearning adapted to continual settings, by contrast, the cost scales with a replay buffer whose size is difficult to budget in advance. The researchers also introduced an annealing strategy that shrinks the burn-in phase of each unlearning operation by ten percent per operation, exploiting forward transfer to cut unlearning time without degrading quality.

Empirical evaluations across sequential vision benchmarks—including Permuted-MNIST, a five-dataset suite combining MNIST, Fashion-MNIST, KMNIST, notMNIST and SVHN, CIFAR-100, and TinyImageNet—showed that UnCLe can perform long interleaved sequences of learning and unlearning requests, up to 30 operations on TinyImageNet, with minimal disruption to previously acquired knowledge. Measured against baselines including fine-tuning, retraining from scratch, and hypernetwork variants that rely on natural catastrophic forgetting, UnCLe performed on par or better across metrics, and strictly better on three of five measures with a ResNet-18 backbone. It also achieved membership inference attack accuracy closest to the ideal fifty percent, a key indicator that forgotten data is genuinely indistinguishable from never-seen data—a central concern for privacy.

Perhaps most importantly for real-world deployment, UnCLe prevented the relapse phenomenon that plagues conventional approaches: tasks unlearned by prior methods tend to creep back once new learning occurs, whereas UnCLe’s unlearned tasks stayed forgotten even as subsequent tasks were absorbed. The authors argue this has broad implications for responsible AI governance, from honoring the right to be forgotten under data protection law to stripping biased or harmful behaviors from deployed models without full retraining. At the same time, they caution that the very possibility of relapse under weaker methods underscores the need for robust verification mechanisms. With code released publicly, the framework offers a template for AI systems that can keep learning throughout their operational life while remaining accountable to demands that they forget on command.

Subject of Research: A hypernetwork framework enabling data-free machine unlearning within continual learning settings

Article Title: A Hypernetwork Framework for Data-Free Unlearning and Continual Learning

Article References: A Hypernetwork Framework for Data-Free Unlearning and Continual Learning. (n.d.). https://doi.org/10.1007/s10994-026-07125-8

Image Credits: AI Generated

DOI: 10.1007/s10994-026-07125-8

Keywords: machine unlearning, continual learning, hypernetworks, data privacy, catastrophic forgetting, task embeddings, neural networks, AI safety, right to be forgotten, ResNet, membership inference, parameter generation