The importance of linking biomedical research and major players of community-based primary health care is well established. Effective collaborations among researchers, clinicians, and professionals in healthcare services can improve patient outcomes, advance healthcare, and address health challenges. Biomarkers could be histologic, radiographic, molecular, or other physiologic characteristics and can be classified into categories like genomic, proteomic, and metabolomic and used for diagnostic, prognostic, or predictive biomarkers. Combating chronic diseases contributing to around 70% of all adult deaths at the world level [1] is an important area of closed linkages between biomedical research and clinical practice, since most of the health services systems aiming at primary care play critical roles in chronic disease management [2,3].

However, biomarkers do not consider how an individual feels, functions, or survives. These are evaluated by outcome measures through patient-reported outcome (PRO) scales, tasks described in neuropsychological tests (NTs), and rating scales like the Muscle Impairment Rating Scale (MIRS) (5-point scale: 1 - 5) to assess muscle weakness severity among individuals with myotonic dystrophy type 1 (DM1). Tailored therapeutic approaches are designed based on pertinent outcome measures, rating scales, and relevant biomarkers [4]. While PROs generate ordinal data, disease-specific biomarkers and count data like the number of plaques and tangles, the number of Cytosine-Thymine-Guanine (CTG) repeats in the blood or Cellular nucleic acid-binding protein (CCTG) for Myotonic dystrophy (MD), the number of errors (Seashore Rhythm Test of HRB), the time taken to complete tasks (Tactual Performance Test of HRB), etc. are in ratio scales. Combining biomarkers, task scores, and PRO scores requires a sound aggregation method for comprehensive assessment of disease status, effect of treatment, and prognosis for better planning and intervention [5]. Several biomarkers are used for accurate evaluation of the relationship of heterogeneous disease outcomes with exposures.

Examples of biomarkers of toxic environmental exposures may include levels of Pb in bone marrow cells for anemia; 2, 3, 7, 8-tetrachloro-dibenzo-p-dioxin for chloracne; hemoglobin adducts for leukemia; serum cholesterol for atherosclerosis; etc. For chlamydia infections, Yang et al. [6] considered monocytes, neutrophils, and albumin as biomarkers. A combination of tumor markers like AFP, CA-125, and CA-19.9 can improve cancer detection. In the exposure–disease continuum, Carmine et al. [7] found 21 possible nominal relationships in epidemiology studies. However, evaluation of relationship is problematic when a subset of biomarkers is inadequate to represent the chosen dimensions. Finding critical dimensions and corresponding biomarkers could help. For diagnosis and assessment of the severity of Alzheimer’s disease (AD), the illustrative list of biomarkers includes counts of the total number of diffuse or neuritic plaques, dense neurofibrillary tangles (NFTs) in tissue samples from multiple brain regions, etc. Other neuropathological markers are the presence of amyloid plaques, deposition of Aβ42, etc.

Existing methods for combining biomarkers, like linear combinations, logistic regression models, nonparametric approaches, etc., are based on different sets of assumptions and differ with respect to their strengths and weaknesses. Yang et al. [6] suggested estimating the distribution of the optimal linear combination of biomarkers and finding diagnostic accuracy through receiver operating characteristic (ROC) curve—area under the ROC curve (AUC) analysis assuming multivariate normal distribution. Major problem areas of the approach are:

• Multivariate normality of the biomarkers needs to be tested. Relevant questions are (i) how best to test multivariate normality and (ii) what to do in case multivariate normality test fails.
• Sound measures of individual disease status are unknown.
• Disease statuses are often binary, i.e., an individual is either with the disease or without the disease. Thus, binary testing needs to be done on pooled samples.
• Probabilities of wrongly specifying sensitivity and (1 – specificity) give poor estimations when the proportion of the subgroup with the disease is high, the sample size is small, etc [7].
• Similar symptoms arising from different combinations of different diseases with non-specific causes may lead to situations where negative biomarkers and clinical features may fail to provide evidence of an early stage of a disease but may appear later [8].
• A large number of possible combinations of a set of biomarkers involves extensive computations.
• Low prevalence and small sample size may give rise to unreliable confidence intervals.

The paper transforms scores of each chosen biomarker, item of different PROs, and rating scale to follow normal distribution ensures multivariate normality of the study variables, facilitating meaningful aggregation to get a normally distributed indexreflecting overall disease status in a comprehensive fashion satisfying desired properties, including quantification of changes by one or a sample of patients over time and statistical inferences.

For accurate prediction of brain β-amyloid (Aβ) for AD, Pan et al. [10] developed an integrated model based on plasma biomarkers like Aβ42, Aβ40, NfL, T-tau, and P-tau181, with APOE genotypes categorized into different groups with respect to Aβ risks, along with scores of tests like the Montreal Cognitive Assessment (MOCA) [11], Activities of Daily Living (ADL) [12], Mini-Mental State Examination (MMSE) [13], and different cognitive domains observed across major demographic variables, which are often used for clinical diagnoses by experienced neurologists. Chen et al. [14] found that the combination of blood-based biomarkers (BBMs) and MemTrax (an online memory test used as a digital biomarker of cognitive impairment) is optimal to predict amyloid load in early stages of AD. Digital biomarkers like the Boston Remote Assessment for Neurocognitive Health (BRANCH) and the Face-Name Associative Memory Exam (FNAME) have shown positive association with Aβ burden and thus increase the feasibility of combining BBMs with digital biomarkers to predict Aβ burden [15,16]. After reviewing 20 collaboration models, Paquette et al. [17] identified implementation barriers like lack of knowledge, financial support, robust management structure, and facilitators for collaboration. Only seven of such models showed measurable impact.

Popular biomarkers for mental disorders include, among others, cortisol levels, neurotransmitter activity (like serotonin and dopamine), neuroimaging showing brain activity patterns, etc., which are assessed through various methods like blood tests, saliva samples, and neuroimaging techniques. Lack of biomarker investigations, especially for better understanding of the pathophysiology of mental disorders and treatment responses to psychiatric conditions, was observed [18]. The need for evaluation of disease status, treatment effect, and efficiency of the clinical team in patient-centered care (PCC), including measurement and aggregation of measures, was suggested [19,20]. Associations between PCC and healthcare outcomes for inpatients, physical and mental health status through PROs, physician-induced demand behaviors, etc. were explored [21]. PCC helps to improve healthcare outcomes for several chronic diseases, and the integration of multidisciplinary team playing plays a significant role in chronic disease management [22,23].

A cohort study by Sivakumaran et al. [24] indicates that a combination of transcriptional biomarkers (like AIRE, B2M, CD19, CXCL10, CXCL13, NCAM1, NLRC4, NLRP1, NLRP2, NOD2, TLR6, TLR8, etc.) and characteristics of patients (like Body Mass Index (BMI), age, gender, tobacco and/or alcohol users previous TB diagnosis, severe thinness, etc.) can better identify failure, especially at the early stage of tuberculosis therapy, and predict TB treatment outcomes through the ROC-AUC approach.

For septic patients, scores of Sequential Organ Failure Assessment (SOFA) are used to evaluate organ dysfunction, enabling stratification of risks for suspected diseases. However, SOFA suffers from poor sensitivity in predicting mortality outside the ICU [25]. Comparing scoring systems of disseminated intravascular coagulation (DIC), Zaiema et al. [26] found that NLR, JAAM-2 DIC, and PAI-1 scores are more reliable in early screening of ICU sepsis patients and improving survival rates. PAI-1 levels ≥ 32.79 ng/ml also provide improved predictions of early mortality within the first 7 critical days.

For the full article, please go through this link: https://www.pubtexto.com/pdf/?aggregation-of-biomarkers-and-patient-reported-outcomes