Padmanabhan, Sandosh

Aim: AI research, development, and implementation are expanding at an exponential pace across healthcare. This paradigm shift in healthcare research has led to increased demands for clinical outcomes, all at the expense of a significant gap in AI literacy within the healthcare field. This has further translated to a lack of tools in creating a framework for literature in the AI in medicine domain. We propose HUMANE (Harmonious Understanding of Machine Learning Analytics Network), a checklist for establishing an international consensus for authors and reviewers involved in research focused on artificial intelligence (AI) or machine learning (ML) in medicine. Methods: This study was conducted using the Delphi method by devising a survey using the Google Forms platform. The survey was developed as a checklist containing 8 sections and 56 questions with a 5-point Likert scale. Results: A total of 33 survey respondents were part of the initial Delphi process with the majority (45%) in the 36–45 years age group. The respondents were located across the USA (61%), UK (24%), and Australia (9%) as the top 3 countries, with a pre-dominant healthcare background (42%) as early-career professionals (3–10 years’ experience) (42%). Feedback showed an overall agreeable consensus (mean ranges 4.1–4.8, out of 5) as cumulative scores throughout all sections. The majority of the consensus was agreeable with the Discussion (Other) section of the checklist (median 4.8 (interquartile range (IQR) 4.8-4.8)), whereas the least agreed section was the Ground Truth (Expert(s) review) section (median 4.1 (IQR 3.9–4.2)) and the Methods (Outcomes) section (median 4.1 (IQR 4.1–4.1)) of the checklist. The final checklist after consensus and revision included a total of 8 sections and 50 questions. Conclusions: The HUMANE international consensus has reflected on further research on the potential of this checklist as an established consensus in improving the reliability and quality of research in this field.

Background Machine learning (ML) is pervasive in all fields of research, from automating tasks to complex decision‐making. However, applications in different specialities are variable and generally limited. Like other conditions, the number of studies employing ML in hypertension research is growing rapidly. In this study, we aimed to survey hypertension research using ML, evaluate the reporting quality, and identify barriers to ML's potential to transform hypertension care. Methods and Results The Harmonious Understanding of Machine Learning Analytics Network survey questionnaire was applied to 63 hypertension‐related ML research articles published between January 2019 and September 2021. The most common research topics were blood pressure prediction (38%), hypertension (22%), cardiovascular outcomes (6%), blood pressure variability (5%), treatment response (5%), and real‐time blood pressure estimation (5%). The reporting quality of the articles was variable. Only 46% of articles described the study population or derivation cohort. Most articles (81%) reported at least 1 performance measure, but only 40% presented any measures of calibration. Compliance with ethics, patient privacy, and data security regulations were mentioned in 30 (48%) of the articles. Only 14% used geographically or temporally distinct validation data sets. Algorithmic bias was not addressed in any of the articles, with only 6 of them acknowledging risk of bias. Conclusions Recent ML research on hypertension is limited to exploratory research and has significant shortcomings in reporting quality, model validation, and algorithmic bias. Our analysis identifies areas for improvement that will help pave the way for the realization of the potential of ML in hypertension and facilitate its adoption.

Objective: Artificial intelligence and machine learning (AI/ML) are increasingly being applied to big clinical data to tackle research questions that cannot be answered with traditional statistical methods. The field is still in its nascent stages and there is a paucity of guidelines for conducting and reporting AI/ML research in hypertension. The objective was to apply the HUMANE checklist to survey the present landscape of AI/ML in hypertension to inform the development of hypertension-specific guidelines and recommendations. Design and method: The HUMANE checklist was developed by global clinical and AI/ML experts through the Delphi method. It assesses the quality of medical AI/ML articles based on whether they cover subjects expected in any peer-reviewed, clinical or AI/ML research publication. A cooping review was carried out to identify articles presenting original research in AI/ML and hypertension published in 2019–2021. Two independent reviewers applied the checklist to each article and in the case of discordance, the response was adjudicated by an AI/ML expert. Results were analysed to assess compliance with the survey (% of papers satisfying checklist requirements). Results: A total of 63 manuscripts was reviewed. A summary of results is shown in Figure 1. Highest compliance was seen for items relating to general article presentation, with compliance ranging from 68% to 98% (description of statistical analysis methods and background context, respectively). Lowest compliance was seen with checklist items relating to clinical research and AI/ML methods. 44% of reviewed articles described the demographics of their dataset and 48% stated their inclusion/exclusion criteria. Nonetheless, datasets were deemed appropriate for investigative aims in 93% of articles. 30% of manuscripts reported their calibration measures, while 73% stated their performance metrics. Internal validation was carried out in 75% of studies, but external validity was assessed in only 14% of cases. Algorithmic bias was addressed in 11% of papers. Conclusions: Application of AI/ML methods in hypertension research is growing, but the majority of current work has major shortfalls in reporting quality, model validation and algorithmic bias. Our study identifies areas of improvement to enable the full realisation of the potential of AI/ML in hypertension.

Selection of antihypertensive treatment according to self-defined ethnicity is recommended by some guidelines but might be better guided by individual genotype rather than ethnicity or race. We compared the extent to which variation in blood pressure response across different ethnicities may be explained by genetic factors: genetically defined ancestry and gene variants at loci known to be associated with blood pressure. We analyzed data from 5 trials in which genotyping had been performed (n=4696) and in which treatment responses to β-blockers, angiotensin-converting enzyme inhibitors, angiotensin receptor blocker, thiazide or thiazide-like diuretic and calcium channel blocker were available. Genetically defined ancestry for proportion of African ancestry was computed using the 1000 genomes population database as a reference. Differences in response to the thiazide diuretic hydrochlorothiazide, the β-blockers atenolol and metoprolol, the angiotensin-converting enzyme inhibitor lisinopril, and the angiotensin receptor blocker candesartan were more closely associated to genetically defined ancestry than self-defined ethnicity in admixed subjects. A relatively small number of gene variants related to loci associated with drug-signaling pathways ( KCNK3, SULT1C3, AMH, PDE3A, PLCE1, PRKAG2 ) with large effect size (−3.5 to +3.5 mm Hg difference in response per allele) and differing allele frequencies in black versus white individuals explained a large proportion of the difference in response to candesartan and hydrochlorothiazide between these groups. These findings suggest that a genomic precision medicine approach can be used to individualize antihypertensive treatment within and across populations without recourse to surrogates of genetic structure such as self-defined ethnicity.

Genome-wide association studies have identified >50 common variants associated with kidney function, but these variants do not fully explain the variation in eGFR. We performed a two-stage meta-analysis of associations between genotypes from the Illumina exome array and eGFR on the basis of serum creatinine (eGFRcrea) among participants of European ancestry from the CKDGen Consortium ( n Stage1 : 111,666; n Stage2 : 48,343). In single-variant analyses, we identified single nucleotide polymorphisms at seven new loci associated with eGFRcrea ( PPM1J , EDEM3, ACP1, SPEG, EYA4, CYP1A1 , and ATXN2L ; P Stage1 <3.7×10 −7 ), of which most were common and annotated as nonsynonymous variants. Gene-based analysis identified associations of functional rare variants in three genes with eGFRcrea, including a novel association with the SOS Ras/Rho guanine nucleotide exchange factor 2 gene, SOS2 ( P =5.4×10 −8 by sequence kernel association test). Experimental follow-up in zebrafish embryos revealed changes in glomerular gene expression and renal tubule morphology in the embryonic kidney of acp1- and sos2 -knockdowns. These developmental abnormalities associated with altered blood clearance rate and heightened prevalence of edema. This study expands the number of loci associated with kidney function and identifies novel genes with potential roles in kidney formation.

To identify novel genes influencing blood pressure response to thiazide diuretic therapy for hypertension, we conducted genome-wide association meta-analyses of ≈1.1 million single-nucleotide polymorphisms in a combined sample of 424 European Americans with primary hypertension treated with hydrochlorothiazide from the Pharmacogenomic Evaluation of Antihypertensive Responses study (n=228) and the Genetic Epidemiology of Responses to Antihypertensive study (n=196). Polymorphisms associated with blood pressure response at P <10 –5 were tested for replication of the associations in independent samples of hydrochlorothiazide-treated European hypertensives. The rs16960228 polymorphism in protein kinase C, α replicated for same-direction association with diastolic blood pressure response in the Nordic Diltiazem study (n=420) and the Genetics of Drug Responsiveness in Essential Hypertension study (n=206), and the combined 4-study meta-analysis P value achieved genome-wide significance ( P =3.3×10 −8 ). Systolic or diastolic blood pressure responses were consistently greater in carriers of the rs16960228 A allele than in GG homozygotes (>4/4 mm Hg) across study samples. The rs2273359 polymorphism in the GNAS-EDN3 region also replicated for same-direction association with systolic blood pressure response in the Nordic Diltiazem study, and the combined 3-study meta-analysis P value approached genome-wide significance ( P =5.5×10 −8 ). The findings document clinically important effects of genetic variation at novel loci on blood pressure response to a thiazide diuretic, which may be a basis for individualization of antihypertensive drug therapy and identification of new drug targets.