The ALLEGRO phase 2a and 2b/3 studies demonstrated that ritlecitinib, an oral JAK3/TEC family kinase inhibitor, is efficacious at doses of ≥ 30 mg in patients aged ≥ 12 years with alopecia areata (AA). The objective of this study was to evaluate the safety of ritlecitinib in an integrated analysis of four studies in AA. Two cohorts were analyzed: a placebo-controlled and an all-exposure cohort. Proportions and study size–adjusted incidence rates (IRs) of adverse events (AEs) of interest and laboratory abnormalities are reported. In the placebo-controlled cohort (n = 881; median exposure: 169 days), the proportion of ritlecitinib-treated patients with AEs was 70.2–75.4
Importance:Early-onset colorectal cancer (EOCRC) (diagnosed age <50 years) incidence is increasing globally, in parallel with increased consumption of ultraprocessed foods (UPFs). The role of UPFs in early-onset colorectal neoplasia remains underexplored. Objective:To evaluate the association between UPF consumption and risk of EOCRC precursors. Design, Setting, and Participants:This prospective cohort study included participants of the Nurses' Health Study II, an ongoing US prospective cohort of female registered nurses established in 1989. Participants were followed up from June 1, 1991, through June 1, 2015. Data were analyzed from October 2024 to July 2025. UPF intake, derived from food-frequency questionnaires administered every 4 years and classified using the Nova system, was modeled as quintiles of energy-adjusted servings per day. Of the nurses enrolled, those who had completed the baseline 1991 food-frequency questionnaire, undergone at least 1 lower endoscopy before age 50 years after baseline, had no history of cancer (except for nonmelanoma skin cancer) before endoscopy, and no colorectal polyp or inflammatory bowel disease were included. Main Outcomes and Measures:Incidence of EOCRC precursors, including conventional adenomas and serrated lesions, confirmed via medical records and pathology reports. Multivariable logistic regression models with generalized estimating equations for clustered data were used to estimate adjusted odds ratios (AORs) and 95% CIs, accounting for known and putative risk factors. Results:Among 29 105 female participants (mean [SD] age, 45.2 [4.5] years) over 24 years of follow-up, 1189 cases were documented of early-onset conventional adenomas and 1598 serrated lesions. UPFs provided 34.8% of total daily calories (median, 5.7 [IQR, 4.5-7.4] servings per day). Participants with higher UPF intake had an increased risk of early-onset conventional adenomas (highest vs lowest intake: AOR, 1.45; 95% CI, 1.19-1.77; overall P < .001) but not serrated lesions (AOR, 1.04; 95% CI, 0.89-1.22; P = .48 for trend). Findings were consistent after further adjustment for body mass index, type 2 diabetes, dietary factors (fiber, folate, calcium, and vitamin D), and Alternative Healthy Eating Index-2010 score. Conclusions and Relevance:In this study, higher UPF intake was associated with increased risk of early-onset colorectal conventional adenomas. These data highlight the important role of UPFs in early-onset colorectal tumorigenesis and support improving dietary quality as a strategy to mitigate the increasing burden of EOCRC.
Restricted methodological naturalism (RMN) is the stipulation that, within the epistemic domain of scientific inquiry, if x explains y, then x is natural (that is, not supernatural). In this article, I argue that RMN cannot do its main job as a stipulation, namely, the job of guarding against present and future "dead-end" explanatory paradigms and encouraging consideration of promising novel explanations. RMN cannot do this job because of two well-known problems: what I call the intelligibility and truth-seeking problems. My argument proceeds in three basic stages: (1) a brief rundown of RMN, in the interest of clarifying the sort of position I have in mind; (2) a recasting of the intelligibility and truth-seeking problems for RMN in case studies from historical and present science; and finally (3) concluding remarks about the staying power of these problems in light of recent work on the intellectual history of naturalism.
We give an asymptotic lower bound on the number of field extensions generated by algebraic points on superelliptic curves over $\mathbb{Q}$ with fixed degree $n$, discriminant bounded by $X$, and Galois closure $S_n$. For $C$ a fixed curve given by an affine equation $y^m=f(x)$ where $m\geq 2$ and $\text{deg }f(x) = d \geq m$, we find that for all degrees $n$ divisible by $\gcd(m,d)$ and sufficiently large, the number of such fields is asymptotically bounded below by $X^{c_n}$, where $c_n \to 1/m^2$ as $n \to \infty$. This bound is determined explicitly by parameterizing $x$ and $y$ by rational functions, counting specializations, and accounting for multiplicity. We then give geometric heuristics suggesting that for $n$ not divisible by $\gcd (m, d)$, degree $n$ points may be less abundant than those for which $n$ is divisible by $\gcd (m, d )$. Namely, we discuss the obvious geometric sources from which we expect to find points on $C$ and discuss the relationship between these sources and our parametrization. When one a priori has a point on $C$ of degree not divisible by $\gcd(m,d)$, we argue that a similar counting argument applies. As a proof of concept we show in the case that $C$ has a rational point that our methods can be extended to bound the number of fields generated by a degree $n$ point of $C$, regardless of divisibility of $n$ by $\gcd (m ,d)$.
The unprecedented success of deep learning (DL) makes it unchallenged when it comes to classification problems. However, it is well established that the current DL methodology produces universally unstable neural networks (NNs). The instability problem has caused an enormous research effort -- with a vast literature on so-called adversarial attacks -- yet there has been no solution to the problem. Our paper addresses why there has been no solution to the problem, as we prove the following mathematical paradox: any training procedure based on training neural networks for classification problems with a fixed architecture will yield neural networks that are either inaccurate or unstable (if accurate) -- despite the provable existence of both accurate and stable neural networks for the same classification problems. The key is that the stable and accurate neural networks must have variable dimensions depending on the input, in particular, variable dimensions is a necessary condition for stability. Our result points towards the paradox that accurate and stable neural networks exist, however, modern algorithms do not compute them. This yields the question: if the existence of neural networks with desirable properties can be proven, can one also find algorithms that compute them? There are cases in mathematics where provable existence implies computability, but will this be the case for neural networks? The contrary is true, as we demonstrate how neural networks can provably exist as approximate minimisers to standard optimisation problems with standard cost functions, however, no randomised algorithm can compute them with probability better than 1/2.
