Background: Retinoblastoma (RB) is a rare pediatric malignancy characterized by small sample sizes and low event rates, where conventional association-driven feature selection may lead to unstable models, overadjustment, and limited generalizability. However, existing survival prediction studies lack a careful treatment of feature selection that accounts for underlying causal structure. Objectives: To develop and validate a causality-guided machine learning model for RB survival prediction by jointly incorporating survival time and survival status as outcome variables. Methods: We analyzed 1015 RB patients from the SEER database (1975-2020). A causality-informed feature selection framework was developed to address the challenges of rare-disease data. Specifically, candidate variables were evaluated through a three-step evidence-integration process: (1) univariate Cox proportional hazards (CPH) analysis for initial statistical screening; (2) causal structure learning using the PC algorithm on the variables retained from Step 1 to construct a directed acyclic graph (DAG) and exclude structurally inappropriate variables (colliders or descendants of the outcome); and (3) LASSO-based feature screening performed independently on the full set of candidate variables. The final features were obtained by taking the intersection of the variables retained from Step 2 and Step 3. Survival models were then trained using the selected features, with model comparison performed as a secondary step. Results: The proposed framework consistently identified four structurally and prognostically robust predictors-laterality, "SEER historic stage A", "RX Summ", and sequence number-through this evidence-integration process. Compared with conventional approaches, the causality-informed framework reduced the feature set while improving model stability and interpretability. Notably, compared with LASSO-only selection, which retained a larger set of variables, the causality-informed approach yielded a more parsimonious feature set with improved predictive performance, suggesting reduced overfitting in a low-event setting. Survival models trained on this refined feature set demonstrated reliable predictive performance, with the random survival forest achieving the highest discrimination (C-index = 0.739). Importantly, the selected predictors aligned with clinically plausible pathways in the learned DAG, supporting their causal relevance. Conclusions: This study demonstrates that incorporating causal structure into feature selection provides a more reliable and interpretable foundation for survival modeling in retinoblastoma. Rather than focusing on algorithmic comparison alone, our findings highlight that careful, causality-informed feature selection is critical for improving robustness in rare-disease prediction tasks. This framework may serve as a generalizable methodological template for other rare clinical settings prone to spurious associations.

Low skeletal muscle index (SMI) has been linked to adverse outcomes in colorectal cancer, but its association with local pathological tumor extent is less clear. This study examined whether preoperative CT-derived SMI was associated with pathological T4 disease in patients undergoing colorectal cancer resection.

This retrospective single-center observational study included 147 consecutive adults who underwent colorectal resection for histologically confirmed adenocarcinoma between January 2022 and November 2025 and had suitable preoperative abdominal or abdomino-pelvic CT imaging within 90 days before surgery. Skeletal muscle area was measured on a single axial CT image at the L3 level, and SMI was calculated as muscle area/height². Patients were classified as having pathological T1-3 or T4 disease. Logistic regression assessed the association between SMI, expressed per 5 cm²/m² increase, and pathological T4 stage.

Patients with pathological T4 tumors had lower SMI than those with T1-3 disease (37.22 vs. 42.85 cm²/m², p = 0.016). Higher SMI was associated with lower odds of T4 disease in univariable analysis (OR 0.79, 95% CI 0.66-0.94; p = 0.008) and after adjustment for age and sex (OR 0.77, 95% CI 0.63-0.94; p = 0.009).

Lower preoperative SMI was associated with pathological T4 colorectal cancer in this cohort. Because of the retrospective observational design, causality cannot be inferred. The association should be interpreted as hypothesis-generating and may reflect reverse causality, shared inflammatory-nutritional pathways, or residual confounding.

Background/Objectives: Celiac disease (CD) is common, but its cancer-risk profile remains incompletely defined. Estimates vary because of referral patterns, diagnostic era, outcome definitions, and surveillance around diagnosis. We evaluated cancer-category-specific associations in a matched cohort of clinically recognized CD. Methods: We used longitudinal electronic health record (EHR) data from Clalit Health Services for a propensity-matched cohort. Adults with EHR-coded CD were matched to controls on demographic, socioeconomic, comorbidity, and inflammatory variables. Pre-index invasive malignancies and non-invasive neoplasms were excluded. Dated EHR-coded invasive oncology outcomes were analyzed using Cox models. A restricted dated-event cohort, lag analyses, competing-risk modeling, hemoglobin adjustment, and age-at-index strata assessed robustness. Results: The primary matched cohort included 8143 individuals: 1006 with CD and 7137 controls, contributing 49,330.5 person-years. CD was associated with increased hazard of an EHR-coded invasive oncology outcome (hazard ratio [HR] 1.61, 95% confidence interval [CI] 1.47-1.77; p<0.001). Strongest signals were hematological malignancy codes (HR 1.99), lymphoma codes (HR 1.90), and gastrointestinal (GI) cancer codes (HR 2.71). Associations persisted after one-year and two-year lags. In the dated-event sensitivity cohort (161 CD; 1610 controls), CD remained associated with invasive cancer (HR 1.68, 95% CI 1.31-2.14), with the strongest signals for lymphoma (HR 2.81) and GI cancer (HR 2.25). The association was essentially unchanged under competing-risk modeling (Fine-Gray subdistribution HR 1.69) and after hemoglobin adjustment (HR 1.61), and was present in both age strata. Neither breast nor lung cancer was associated. Lymphoma codes included peripheral T-cell lymphomas recorded at intra-abdominal and extranodal sites, the pattern most consistent with enteropathy-associated T-cell lymphoma (EATL). Conclusions: In clinically recognized CD, cancer hazard was elevated and category-specific, concentrated in hematological, lymphoid, and GI codes with a gut-oriented T-cell lymphoma signal. The findings support targeted clinical vigilance, not expanded screening, and describe relative associations that require registry-linked confirmation.

Acute lymphoblastic leukemia (ALL) is the most common pediatric malignancy and is associated with profound metabolic reprogramming. This exploratory study aimed to characterize the metabolomic profiles of plasma and cerebrospinal fluid (CSF) in children with newly diagnosed pre-B-cell acute lymphoblastic leukemia (pre-B ALL) prior to therapy. Metabolomic analyses were performed using mass spectrometry-based platforms combined with multivariate statistical approaches (PCA, OPLS-DA, SVM-RFE, EBAM). In plasma, we identified 41 significantly altered metabolites (FDR < 0.011), revealing a distinct signature that differentiated pre-B ALL patients from healthy controls. Specifically, patients exhibited elevated levels of hypoxanthine, xanthine, and phosphatidylcholine derivatives, alongside reduced concentrations of L-cysteine and prasterone sulfate, indicating systemic dysregulation of purine, lipid, and amino acid metabolism. In CSF, we observed a distinct metabolic profile characterized by coordinated disturbances in purine degradation, phospholipid metabolism, and sphingolipid pathways. Notably, correlation analysis between the two matrices suggested that systemic metabolic shifts, particularly in purine metabolism (e.g., hypoxanthine and xanthine levels), are mirrored within the central nervous system microenvironment. These findings indicate that children with pre-B ALL exhibit specific metabolic alterations in both compartments before treatment. This work serves as a proof-of-concept for applying metabolomics in pediatric oncology, highlighting the necessity for further validation in larger, prospective cohorts to assess the clinical utility of these profiles.

Environmental carcinogenesis is traditionally explained by the accumulation of genetic alterations induced by exogenous carcinogens. However, most exposed cells do not undergo malignant transformation because intrinsic tumor-suppressive mechanisms limit the expansion of damaged clones. Among these mechanisms, cellular senescence represents a major barrier that restricts proliferation following DNA damage, oncogenic stress, and other carcinogen-induced insults. In this review, we examine environmental carcinogenesis within a probabilistic evolutionary framework in which tumor initiation depends not only on mutation acquisition but also on the ability of rare cells to evade senescence-mediated growth arrest. Environmental carcinogens contribute to cancer development by increasing genomic instability, altering tissue microenvironments, and modifying selective pressures, whereas senescence acts as a critical constraint on clonal evolution. We further discuss how aging, immune surveillance, DNA-repair capacity, and tissue-specific factors influence the likelihood of senescence escape and malignant progression. This integrative perspective highlights carcinogenesis as a multistep stochastic process shaped by the interaction between mutational events, cellular fitness barriers, and microenvironmental selection. Understanding how these factors collectively regulate transformation may improve mechanistic models of cancer risk and identify new opportunities for prevention and early intervention.

Immunity plays critical roles in cutaneous melanoma. We investigated the association between the pro-inflammatory interleukin-1β gene (IL-1B) rs16944 (-511 C > T) promoter single nucleotide polymorphism (SNP) and cutaneous melanoma susceptibility and clinical features. This observational case-control study included 133 patients with cutaneous melanoma and 900 healthy controls from Northeastern Italy. The rs16944 C > T polymorphism was determined by genomic DNA restriction fragment analysis. The IL-1B rs16944 C allele (OR = 1.44, p = 0.014) and CC genotype (OR = 1.48, p = 0.037) were associated with modestly higher odds of melanoma. Among melanoma patients, the CC genotype was associated with Stage I disease (OR = 2.45, p = 0.016) and Breslow thickness ≤ 0.75 mm (OR = 2.27, p = 0.049), whereas it was less frequent in Stage IV melanoma (OR = 0.37, p = 0.029). The CT genotype was associated with Stage IV melanoma (OR = 3.03, p = 0.014) and with lower-limb (OR = 2.45, p = 0.038) and lower-extremity (OR = 3.09, p = 0.005) melanoma. These divergent associations are exploratory and do not establish disease trajectory or a functional effect of rs16944; mechanistic interpretation remains uncertain because IL-1β expression or activity was not measured in this cohort. To our knowledge, this is the first report of an association between a genetic polymorphism and lower-limb/lower-extremity melanoma. These exploratory findings require confirmation in independent cohorts.

Extranodal natural killer/T-cell lymphoma, nasal type (ENKTL-NT), is a rare and aggressive lymphoma for which real-world data from resource-limited settings remain scarce. This study evaluated the efficacy and safety of frontline multimodal treatment strategies for localized ENKTL-NT.

We retrospectively analyzed 62 patients with localized ENKTL-NT treated at the Vietnam National Cancer Hospital between May 2019 and June 2025. Patients received concurrent chemoradiotherapy followed by VIPD or VIDL, or sequential chemoradiotherapy with GELOX/PGEMOX. Treatment responses, survival outcomes, and toxicities were assessed.

The median age was 44 years, and 67.7% of patients had stage I disease. Baseline EBV-DNA positivity was detected in 46.8%, while 54.8% had PINKE scores of 1-2. The overall response rate was 93.5%, including an 85.5% complete response rate. Grade 3-4 neutropenia was the most common adverse event (18.7%). The estimated 3-year progression-free survival (PFS) and overall survival (OS) rates were 76.3% and 79.5%, respectively. Multivariable Cox analysis identified PINKE risk stratification as an independent predictor of both PFS and OS.

Frontline multimodal treatment strategies achieved high response rates, favorable survival outcomes, and manageable toxicities in localized ENKTL-NT. The PINKE score remained an important prognostic factor, supporting risk-adapted treatment selection in routine clinical practice.

Background: Inequitable access to clinical trials persists in blood cancers despite ongoing equity, diversity, and inclusion (EDI) efforts. Despite the critical role of clinical trials in improving survival and outcomes, recruitment remains suboptimal, limiting patient access to potentially life-saving therapies. Practical and scalable approaches are therefore needed to address the non-medical barriers that hinder patient readiness upstream of enrolment. Methods: Myeloma Canada led a national, multi-phase initiative using human-centred design (HCD) to operationalize EDI in clinical trials. Following an initial systems level workshop, the two-day Phase 0 workshop used a HCD approach that convened a purposively selected multidisciplinary group of stakeholders to co-design operational solutions for non-medical barriers affecting trial participation for patients. Given the use of purposive sampling, the results should be interpreted as reflecting a balanced range of diverse, informed perspectives across the Canadian clinical trial ecosystem. Results: Participants identified persistent cultural, logistical, financial, and linguistic barriers, along with fragmented awareness of available supports. Across diverse personas and care settings, all groups independently converged on a human-centred, equity-focused pre-trial navigation model supported by simple digital tools, including AI-enabled infrastructure drawing on curated resources from validated sources with appropriate governance, privacy, and oversight. Digital tools were proposed to support, rather than replace, human support and to align with existing health system realities. Conclusions: This hypothesis-generating work proposes a feasible, sustainable, and scalable equity-embedded, protocol-agnostic navigation framework. Its external hub-and-spoke structure can reduce non-medical barriers, strengthen trial access and accrual, and enhance representativeness. Pilot implementation that assesses feasibility, uptake, workflow impact, equity effects, and implementation burden is warranted.

Background/Objectives: Multiple myeloma (MM) disproportionately affects older adults and racial/ethnic minorities, many of whom face socioeconomic barriers to care. Immunomodulators (IMiDs) are central to MM therapy, but their benefit relies on timely initiation and sustained adherence. In safety-net settings, barriers such as insurance approvals, Patient Medication Assistance Program (PMAP) enrollment, and Risk Evaluation and Mitigation Strategy (REMS) requirements may influence adherence. This study evaluated IMiD adherence during the first two years of therapy among socioeconomically vulnerable MM patients in a large safety-net hospital, with attention to PMAP enrollment and treatment delays. Methods: We conducted a retrospective cohort study of adults (≥18 years) with newly diagnosed MM between October 2010 and January 2022 who initiated lenalidomide or pomalidomide at a safety-net hospital. Adherence was measured using the proportion of days covered (PDC). Delays in initiation were defined as ≥1 month from diagnosis to first IMiD prescription fill. This threshold was selected a priori based on the REMS-mandated 28-day prescription supply limit for lenalidomide and pomalidomide, which establishes a natural monthly treatment cycle; a delay exceeding one full 28-day cycle before therapy initiation is operationally meaningful within the REMS framework. Results: Eighty patients met criteria (mean age 55.1 years; 56.3% Hispanic/Latino; 31.3% African American). Most were unemployed (63.8%), 42.5% uninsured, and 57.5% enrolled in PMAP. Only 12.5% achieved PDC ≥ 80%. Median PDC was higher among PMAP enrollees (0.40 vs. 0.31; p = 0.0293) and transplant-ineligible patients (0.46 vs. 0.27; p = 0.0218). Delays ≥ 1 month occurred in 65% of patients. Conclusions: Adherence to IMiDs was suboptimal in this younger safety-net cohort. Although PMAP improved adherence, program delays may reduce early treatment benefits, underscoring the need for streamlined access and targeted interventions to address both patient- and system-level barriers.

In the companion article in this issue of Current Oncology, 'Management of AML in Older Patients: An Updated Canadian Consensus', the authors have presented the third iteration of Canadian consensus guidelines on AML treatment in the elderly. While many aspects of AML treatment in the elderly have become better defined, some old questions remain and new questions and controversies have arisen. Here, we address three topics on which there is, at this time, no universal consensus. The first is the development and features of new risk-stratification systems specifically aimed at older, less-intensively treated patients. Several different systems now exist in parallel, potentially causing confusion amongst clinicians. The second topic is good-prognosis AML subtypes (IDH1- and NPM1-mutated AML). In the Canadian context, IDH1-mutated AML is of particular interest due to the new availability of ivosidenib in Canada. The third topic is adverse-risk AML subtypes (FLT3- and TP53-mutated AML). FLT3-mutated AML remains problematic, although it is anticipated that new drug approvals and measurable residual disease-based treatment approaches may alleviate this, at least in part. And in particular, TP53-mutated AML remains a major problem. Ongoing clinical trial enrollment is essential.