Futurists have anticipated that novel autonomous technologies, embedded with machine learning (ML), will substantially influence healthcare. ML is focused on making predictions as accurate as possible, while traditional statistical models are aimed at inferring relationships between variables. The benefits of ML comprise flexibility and scalability compared with conventional statistical approaches, which makes it deployable for several tasks, such as diagnosis and classification, and survival predictions. However, much of ML-based analysis remains scattered, lacking a cohesive structure. There is a need to evaluate and compare the performance of well-developed conventional statistical methods and ML on patient outcomes, such as survival, response to treatment, and patient-reported outcomes (PROs). In this article, we compare the usefulness and limitations of traditional statistical methods and ML, when applied to the medical field. Traditional statistical methods seem to be more useful when the number of cases largely exceeds the number of variables under study and a priori knowledge on the topic under study is substantial such as in public health. ML could be more suited in highly innovative fields with a huge bulk of data, such as omics, radiodiagnostics, drug development, and personalized treatment. Integration of the two approaches should be preferred over a unidirectional choice of either approach.

The global public health crisis caused by COVID-19 has lasted longer than many of us would have hoped and expected. With its high uncertainty and limited control, the COVID-19 pandemic has undoubtedly asked a lot from all of us. One important central question is: how resilient have we proved in face of the unprecedented and prolonged coronavirus pandemic? There is a vast and rapidly growing literature that has examined the impact of the pandemic on mental health both on the shorter (2020) and longer (2021) term. This not only concerns pandemic-related effects on resilience in the general population, but also how the pandemic has challenged stress resilience and mental health outcomes across more specific vulnerable population groups: patients with a psychiatric disorder, COVID-19 diagnosed patients, health care workers, children and adolescents, pregnant women, and elderly people. It is challenging to keep up to date with, and interpret, this rapidly increasing scientific literature. In this review, we provide a critical overview on how the COVID-19 pandemic has impacted mental health and how human stress resilience has been shaped by the pandemic on the shorter and longer term. The vast literature is dominated by a wealth of data which are, however, not always of the highest quality and heavily depend on online and self-report surveys. Nevertheless, it appears that we have proven surprisingly resilient over time, with fast recovery from COVID-19 measures. Still, vulnerable groups such as adolescents and health care personnel that have been severely impacted by the COVID-19 pandemic do exist. Large interindividual differences exist, and for future pandemics there is a clear need to comprehensively and integratively assess resilience from the start to provide personalized help and interventions tailored to the specific needs for vulnerable groups.

Lithium is a first-line treatment in bipolar disorder, but individual response is variable. Previous studies have suggested that lithium response is a heritable trait. However, no genetic markers of treatment response have been reproducibly identified. Here, we report the results of a genome-wide association study of lithium response in 2563 patients collected by 22 participating sites from the International Consortium on Lithium Genetics (ConLiGen). Data from common single nucleotide polymorphisms (SNPs) were tested for association with categorical and continuous ratings of lithium response. Lithium response was measured using a well established scale (Alda scale). Genotyped SNPs were used to generate data at more than 6 million sites, using standard genomic imputation methods. Traits were regressed against genotype dosage. Results were combined across two batches by meta-analysis. A single locus of four linked SNPs on chromosome 21 met genome-wide significance criteria for association with lithium response (rs79663003, p=1·37 × 10(-8); rs78015114, p=1·31 × 10(-8); rs74795342, p=3·31 × 10(-9); and rs75222709, p=3·50 × 10(-9)). In an independent, prospective study of 73 patients treated with lithium monotherapy for a period of up to 2 years, carriers of the response-associated alleles had a significantly lower rate of relapse than carriers of the alternate alleles (p=0·03268, hazard ratio 3·8, 95% CI 1·1-13·0). The response-associated region contains two genes for long, non-coding RNAs (lncRNAs), AL157359.3 and AL157359.4. LncRNAs are increasingly appreciated as important regulators of gene expression, particularly in the CNS. Confirmed biomarkers of lithium response would constitute an important step forward in the clinical management of bipolar disorder. Further studies are needed to establish the biological context and potential clinical utility of these findings. Deutsche Forschungsgemeinschaft, National Institute of Mental Health Intramural Research Program.

Bipolar disorder (BD) is a genetically complex mental illness characterized by severe oscillations of mood and behaviour. Genome-wide association studies (GWAS) have identified several risk loci that together account for a small portion of the heritability. To identify additional risk loci, we performed a two-stage meta-analysis of >9 million genetic variants in 9,784 bipolar disorder patients and 30,471 controls, the largest GWAS of BD to date. In this study, to increase power we used ∼2,000 lithium-treated cases with a long-term diagnosis of BD from the Consortium on Lithium Genetics, excess controls, and analytic methods optimized for markers on the X-chromosome. In addition to four known loci, results revealed genome-wide significant associations at two novel loci: an intergenic region on 9p21.3 (rs12553324, P =  5.87 × 10 ^{− 9}; odds ratio (OR) = 1.12) and markers within ERBB2 (rs2517959, P =  4.53 × 10 ^{− 9}; OR = 1.13). No significant X-chromosome associations were detected and X-linked markers explained very little BD heritability. The results add to a growing list of common autosomal variants involved in BD and illustrate the power of comparing well-characterized cases to an excess of controls in GWAS.

Phenotypic misclassification (between cases) has been shown to reduce the power to detect association in genetic studies. However, it is conceivable that complex traits are heterogeneous with respect to individual genetic susceptibility and disease pathophysiology, and that the effect of heterogeneity has a larger magnitude than the effect of phenotyping errors. Although an intuitively clear concept, the effect of heterogeneity on genetic studies of common diseases has received little attention. Here we investigate the impact of phenotypic and genetic heterogeneity on the statistical power of genome wide association studies (GWAS). We first performed a study of simulated genotypic and phenotypic data. Next, we analyzed the Wellcome Trust Case-Control Consortium (WTCCC) data for diabetes mellitus (DM) type 1 (T1D) and type 2 (T2D), using varying proportions of each type of diabetes in order to examine the impact of heterogeneity on the strength and statistical significance of association previously found in the WTCCC data. In both simulated and real data, heterogeneity (presence of “non-cases”) reduced the statistical power to detect genetic association and greatly decreased the estimates of risk attributed to genetic variation. This finding was also supported by the analysis of loci validated in subsequent large-scale meta-analyses. For example, heterogeneity of 50% increases the required sample size by approximately three times. These results suggest that accurate phenotype delineation may be more important for detecting true genetic associations than increase in sample size.