AI/ML in Precision Medicine: A Look Beyond the Hype.

Artificial Intelligence (AI) and Machine Learning (ML) are making headlines in medical research, especially in drug discovery, digital imaging, disease diagnostics, genetic testing, and optimal care pathway (personalized care). However, the potential uses and benefits of AI/ML applications need to be distinguished from hype. In the 2022 American Statistical Association Biopharmaceutical Section Regulatory-Industry Statistical Workshop, we convened a panel of experts from FDA and industry to talk about the challenges of successfully applying AI/ML in precision medicine and how to overcome those challenges. This paper provides a summary and expansion on the topics discussed in the panel: the application of AI/ML, bias, and data quality.

Secukinumab's effect on structural damage progression in psoriatic arthritis: longitudinal mixture modelling of FUTURE-1 and FUTURE-5.

OBJECTIVES: Peripheral and axial manifestations of psoriatic arthritis (PsA) can lead to irreversible structural damage and chronic disability. Our objective was to explore predictors of radiographic progression and to increase our understanding of treatment effects in subgroups of patients with different rates of structural damage progression. METHODS: We analysed data from two large Phase-3 trials of secukinumab in PsA patients, FUTURE-1 (NCT01392326, n=606) and FUTURE-5 (NCT02404350, n=996), where different posologies ranging from 75 mg to 300 mg were used. We applied a longitudinal Bayesian mixture model with random effects to account for the variability in the repeated radiographic assessments. "Fast progressors" were defined post hoc as patients with a 50% model-estimated probability to progress at least 0.5 mTSS/year faster than an average patient. RESULTS: Higher baseline inflammation and higher body weight were identified as significant predictors of radiographic progression (multivariate model). Model-estimated structural damage progression in an average patient treated with secukinumab 150 mg subcutaneous (s.c.) was slower (0.04 mTSS/year; 95% CI -0.28, 0.34) compared to a patient treated with placebo (0.94 mTSS/year; 95% CI 0.45, 1.45). According to the model, the subgroup of "fast progressors" (hsCRP ≥26 mg/L, body weigth ≥94 kg, inadequate response to prior anti-TNF-alpha, structural damage ≥42 mTSS) treated with secukinumab 150 mg s.c. progressed at 0.56 mTSS/year (95% CI 0.02, 1.09) and 1.46 mTSS/year (95% CI 0.81, 2.11) when treated with placebo. CONCLUSIONS: Greater systemic inflammation and higher body weight at baseline were identified as significant predictors of progression. Even patients with fast radiographic progression could experience a beneficial effect with secukinumab that holds promise to prevent further mobility loss.

Inter-phenotypic differences in CYP2C9 and CYP2C19 metabolism: Bayesian meta-regression of human population variability in kinetics and application in chemical risk assessment.

Quantifying variability in pharmacokinetics (PK) and toxicokinetics (TK) provides a science-based approach to refine uncertainty factors (UFs) for chemical risk assessment. In this context, genetic polymorphisms in cytochromes P450 (CYPs) drive inter-phenotypic differences and may result in reduction or increase in metabolism of drugs or other xenobiotics. Here, an extensive literature search was performed to identify PK data for probe substrates of the human polymorphic isoforms CYP2C9 and CYP2C19. Relevant data from 158 publications were extracted for markers of chronic exposure (clearance and area under the plasma concentration-time curve) and analysed using a Bayesian meta-regression model. Enzyme function (EF), driven by inter-phenotypic differences across a range of allozymes present in extensive and poor metabolisers (EMs and PMs), and fraction metabolised (Fm), were identified as exhibiting the highest impact on the metabolism. The Bayesian meta-regression model provided good predictions for such inter-phenotypic differences. Integration of population distributions for inter-phenotypic differences and estimates for EF and Fm allowed the derivation of CYP2C9- and CYP2C19-related UFs which ranged from 2.7 to 12.7, and were above the default factor for human variability in TK (3.16) for PMs and major substrates (Fm >60%). These results provide population distributions and pathway-related UFs as conservative in silico options to integrate variability in CYP2C9 and CYP2C19 metabolism using in vitro kinetic evidence and in the absence of human data. The future development of quantitative extrapolation models is discussed with particular attention to integrating human in vitro and in vivo PK or TK data with pathway-related variability for chemical risk assessment.

Artificial Intelligence and Machine Learning Applied at the Point of Care.

INTRODUCTION: The increasing availability of healthcare data and rapid development of big data analytic methods has opened new avenues for use of Artificial Intelligence (AI)- and Machine Learning (ML)-based technology in medical practice. However, applications at the point of care are still scarce. OBJECTIVE: Review and discuss case studies to understand current capabilities for applying AI/ML in the healthcare setting, and regulatory requirements in the US, Europe and China. METHODS: A targeted narrative literature review of AI/ML based digital tools was performed. Scientific publications (identified in PubMed) and grey literature (identified on the websites of regulatory agencies) were reviewed and analyzed. RESULTS: From the regulatory perspective, AI/ML-based solutions can be considered medical devices (i.e., Software as Medical Device, SaMD). A case series of SaMD is presented. First, tools for monitoring and remote management of chronic diseases are presented. Second, imaging applications for diagnostic support are discussed. Finally, clinical decision support tools to facilitate the choice of treatment and precision dosing are reviewed. While tested and validated algorithms for precision dosing exist, their implementation at the point of care is limited, and their regulatory and commercialization pathway is not clear. Regulatory requirements depend on the level of risk associated with the use of the device in medical practice, and can be classified into administrative (manufacturing and quality control), software-related (design, specification, hazard analysis, architecture, traceability, software risk analysis, cybersecurity, etc.), clinical evidence (including patient perspectives in some cases), non-clinical evidence (dosing validation and biocompatibility/toxicology) and other, such as e.g. benefit-to-risk determination, risk assessment and mitigation. There generally is an alignment between the US and Europe. China additionally requires that the clinical evidence is applicable to the Chinese population and recommends that a third-party central laboratory evaluates the clinical trial results. CONCLUSIONS: The number of promising AI/ML-based technologies is increasing, but few have been implemented widely at the point of care. The need for external validation, implementation logistics, and data exchange and privacy remain the main obstacles.

Perinatal Antiretroviral Intensification to Prevent Intrapartum HIV Transmission When Antenatal Antiretroviral Therapy Is Initiated Less Than 8 Weeks Before Delivery.

INTRODUCTION: Infants born to women living with HIV initiating combination antiretroviral therapy (cART) late in pregnancy are at high risk of intrapartum infection. Mother/infant perinatal antiretroviral intensification may substantially reduce this risk. METHODS: In this single-arm Bayesian trial, pregnant women with HIV receiving standard of care antiretroviral prophylaxis in Thailand (maternal antenatal lopinavir-based cART; nonbreastfed infants 4 weeks' postnatal zidovudine) were offered "antiretroviral intensification" (labor single-dose nevirapine plus infant zidovudine-lamivudine-nevirapine for 2 weeks followed by zidovudine-lamivudine for 2 weeks) if their antenatal cART was initiated ≤8 weeks before delivery. A negative birth HIV-DNA polymerase chain reaction (PCR) followed by a confirmed positive PCR defined intrapartum transmission. Before study initiation, we modeled intrapartum transmission probabilities using data from 3738 mother/infant pairs enrolled in our previous trials in Thailand using a logistic model, with perinatal maternal/infant antiretroviral regimen and predicted viral load at delivery as main covariates. Using the characteristics of the women enrolled who received intensification, prior intrapartum transmission probabilities (credibility intervals) with/without intensification were estimated. After including the transmission data observed in the current study, the corresponding Bayesian posterior transmission probability was derived. RESULTS: No intrapartum transmission of HIV was observed among the 88 mother/infant pairs receiving intensification. The estimated intrapartum transmission probability was 2·2% (95% credibility interval 0·5-6·1) without intensification versus 0·3% (0·0-1·6) with intensification. The probability of superiority of intensification over standard of care was 94·4%. Antiretroviral intensification appeared safe. CONCLUSION: Mother/infant antiretroviral intensification was effective in preventing intrapartum transmission of HIV in pregnant women receiving ≤8 weeks antepartum cART.

Development and validation of a Bayesian survival model for inclusion body myositis.

BACKGROUND: Associations between disease characteristics and payer-relevant outcomes can be difficult to establish for rare and progressive chronic diseases with sparse available data. We developed an exploratory bridging model to predict premature mortality from disease characteristics, and using inclusion body myositis (IBM) as a representative case study. METHODS: Candidate variables that may be potentially associated with premature mortality were identified by disease experts and from the IBM literature. Interdependency between candidate variables in IBM patients were assessed using existing patient-level data. A Bayesian survival model for the IBM population was developed with identified variables as predictors for premature mortality in the model. For model selection and external validation, model predictions were compared to published mortality data in IBM patient cohorts. After validation, the final model was used to simulate the increased risk of premature death in IBM patients. Baseline survival was based on age- and gender-specific survival curves for the general population in Western countries as reported by the World Health Organisation. RESULTS: Presence of dysphagia, aspiration pneumonia, falls, being wheelchair-bound and 6-min walking distance (6MWD in meters) were identified as candidate variables to be used as predictors for premature mortality based on inputs received from disease experts and literature. There was limited correlation between these functional performance measures, which were therefore treated as independent variables in the model. Based on the Bayesian survival model, among all candidate variables, presence of dysphagia and decrease in 6MWD [m] were associated with poorer survival with contributing hazard ratios (HR) 1.61 (95% credible interval [CrI]: 0.84-3.50) and 2.48 (95% CrI: 1.27-5.00) respectively. Excess mortality simulated in an IBM cohort vs. an age- and gender matched general-population cohort was 4.03 (95% prediction interval 1.37-10.61). CONCLUSIONS: For IBM patients, results suggest an increased risk of premature death compared with the general population of the same age and gender. In the absence of hard data, bridging modelling generated survival predictions by combining relevant information. The methodological principle would be applicable to the analysis of associations between disease characteristics and payer-relevant outcomes in progressive chronic and rare diseases. Studies with lifetime follow-up would be needed to confirm the modelling results.

Investigating combined toxicity of binary mixtures in bees: Meta-analysis of laboratory tests, modelling, mechanistic basis and implications for risk assessment.

Bees are exposed to a wide range of multiple chemicals "chemical mixtures" from anthropogenic (e.g. plant protection products or veterinary products) or natural origin (e.g. mycotoxins, plant toxins). Quantifying the relative impact of multiple chemicals on bee health compared with other environmental stressors (e.g. varroa, viruses, and nutrition) has been identified as a priority to support the development of holistic risk assessment methods. Here, extensive literature searches and data collection of available laboratory studies on combined toxicity data for binary mixtures of pesticides and non-chemical stressors has been performed for honey bees (Apis mellifera), wild bees (Bombus spp.) and solitary bee species (Osmia spp.). From 957 screened publications, 14 publications provided 218 binary mixture toxicity data mostly for acute mortality (lethal dose: LD50) after contact exposure (61%), with fewer studies reporting chronic oral toxicity (20%) and acute oral LC50 values (19%). From the data collection, available dose response data for 92 binary mixtures were modelled using a Toxic Unit (TU) approach and the MIXTOX modelling tool to test assumptions of combined toxicity i.e. concentration addition (CA), and interactions (i.e. synergism, antagonism). The magnitude of interactions was quantified as the Model Deviation Ratio (MDR). The CA model applied to 17% of cases while synergism and antagonism were observed for 72% (MDR > 1.25) and 11% (MDR < 0.83) respectively. Most synergistic effects (55%) were observed as interactions between sterol-biosynthesis-inhibiting (SBI) fungicides and insecticide/acaricide. The mechanisms behind such synergistic effects of binary mixtures in bees are known to involve direct cytochrome P450 (CYP) inhibition, resulting in an increase in internal dose and toxicity of the binary mixture. Moreover, bees are known to have the lowest number of CYP copies and other detoxification enzymes in the insect kingdom. In the light of these findings, occurrence of these binary mixtures in relevant crops (frequency and concentrations) would need to be investigated. Addressing this exposure dimension remains critical to characterise the likelihood and plausibility of such interactions to occur under field realistic conditions. Finally, data gaps and further work for the development of risk assessment methods to assess multiple stressors in bees including chemicals and non-chemical stressors in bees are discussed.

Second-line cabozantinib versus nivolumab in advanced renal cell carcinoma: Systematic review and indirect treatment comparison.

BACKGROUND: Nivolumab and cabozantinib, two new treatment options for previously-treated advanced/metastatic renal cell carcinoma (aRCC), have recently been approved. METHODS: Two independent reviewers performed study selection, data extraction, and risk of bias assessment. Indirect treatment comparisons were carried out by directly assessing HR differences and statistical modeling of Kaplan-Meier curves from these two trials. RESULTS: Publications identified showed that no head-to-head comparisons had been carried out. Two indirect treatment comparisons used agreed that there was no significant difference in OS between cabozantinib and nivolumab and that cabozantinib significantly improved PFS compared to nivolumab. CONCLUSIONS: The field of aRCC treatments is evolving rapidly, creating opportunities for individualized treatments and challenges for clinicians to keep up with the evidence. In lieu of availability of direct comparisons of all options, advanced modeling results presented herein can help to inform and improve personalized treatments.

The Yin-Yang of CYP3A4: a Bayesian meta-analysis to quantify inhibition and induction of CYP3A4 metabolism in humans and refine uncertainty factors for mixture risk assessment.

Quantifying differences in pharmacokinetics (PK) and toxicokinetics (TK) provides a science-based approach to refine uncertainty factors (UFs) for chemical risk assessment. Cytochrome P450 (CYP) 3A4-the major hepatic and intestinal human CYP-and the P-glycoprotein (Pgp) transporter share a vast range of common substrates for which PK may be modulated through inhibition or induction in the presence of grapefruit juice (GFJ) or St. John's wort (SJW), respectively. Here, an extensive literature search was performed on PK interactions for CYP3A4 and Pgp substrates after oral co-exposure to GFJ and SJW. Relevant data from 109 publications, extracted for both markers of acute (Cmax) and chronic [clearance and area under the plasma concentration-time curve (AUC)] exposure, were computed into a Bayesian hierarchical meta-analysis model. Bioavailability (F) and substrate fraction metabolised by CYP3A4 (Fm) were identified as the variables exhibiting the highest impact on the magnitude of interaction. The Bayesian meta-regression model developed provided good predictions for magnitudes of inhibition (maximum 5.3-fold with GFJ) and induction (maximum 2.3-fold with SJW). Integration of CYP3A4 variability, F, Fm and magnitude of interaction provided the basis to derive a range of CYP3A4 and Pgp-related UFs. Such CYP3A4 and Pgp-related UFs can be derived in the absence of human data using in vitro TK evidence for CYP3A4/Pgp inhibition or induction as conservative in silico options. The future development of quantitative in vitro-in vivo extrapolation models for mixture risk assessment is discussed with particular attention to integrating human in vitro and in vivo P/TK data on interactions with pathway-related variability.

Challenges for Optimizing Real-World Evidence in Alzheimer's Disease: The ROADMAP Project.

ROADMAP is a public-private advisory partnership to evaluate the usability of multiple data sources, including real-world evidence, in the decision-making process for new treatments in Alzheimer's disease, and to advance key concepts in disease and pharmacoeconomic modeling. ROADMAP identified key disease and patient outcomes for stakeholders to make informed funding and treatment decisions, provided advice on data integration methods and standards, and developed conceptual cost-effectiveness and disease models designed in part to assess whether early treatment provides long-term benefit.

Clinical and economic burden of severe asthma: A French cohort study.

OBJECTIVE: To describe the clinical and economic burden of severe asthma in France over 12 months. METHODS: Data were retrieved from the observational, prospective "Cohorte Obstruction Bronchique et Asthme" (COBRA) cohort, which has enrolled nearly 1000 asthma patients since 2007 from throughout France. Patients undergoing treatment with GINA step-4 or 5 medications uninterruptedly for 12 months (thus defining "severe asthma") were identified and their clinical data used to describe the clinical burden of asthma (exacerbations, symptoms outside exacerbations, and level of asthma control). Patients' utilization of healthcare resources was described and used to estimate the direct medical costs incurred to treat severe asthma. RESULTS: 155 patients were included in the present study. Over the 12-month period of interest, 128 (83%) patients experienced at least one asthma exacerbation, 22 (14%) patients were hospitalized for asthma, 133 (86%) patients experienced continuous symptoms outside exacerbations, and 77 (50%) patients experienced important limitations in daily life activities. The median number of asthma-related drugs used was 4. The mean estimated annual asthma-related cost was 8,222 euros (standard deviation, SD = 11,886), including 7,229 euros (SD = 11,703) for controller medications. CONCLUSION: Symptoms outside exacerbation periods are highly prevalent in severe asthma patients, for whom the main driver of medical costs is controller medication.

A retrospective observational analysis of post-pandemic influenza-related outcomes in the United Kingdom, 2010-2014.

This study set out to evaluate influenza- and respiratory-related illnesses recorded during primary care physician consultations in England following the H1N1 pandemic in 2009 and to enable the development of a dynamic disease model. Data were obtained from the Clinical Practice Research Datalink of primary care records over four influenza seasons (2010-2014). The primary outcome of the study was incidence of influenza- and respiratory-related diagnoses, calculated per practice and by season and age group. Upper respiratory tract infection diagnoses were most frequently recorded (mean seasonal practice level incidence; 3,762 consultations per 100,000 [SD = 1,989]), and influenza-related diagnoses were least frequently recorded across all seasons, except one. Incidence rates for the under 18 population were higher than those for the general population, in particular for upper respiratory tract infection (range of 8,024-9,950 versus 3,228-4,120, respectively) and otitis media diagnoses (2,668-3,652 versus 782-1,057, respectively). For influenza-related diagnoses, the 65+ age group, the 0 to <2 and 2 to <4 groups had a higher risk (risk ratio = 1.33, 1.12 and 1.16, respectively) than other age groups. This study provides valuable insight into the incidence of influenza- and respiratory-related diagnoses in the primary care setting in England, and suggests a higher burden of disease in young children and the elderly. The study also indicates that some influenza illness is likely to be reported under respiratory-related diagnoses, given the low incidence of influenza-related diagnoses in the study.

Impact of increased influenza vaccination in 2-3-year-old children on disease burden within the general population: A Bayesian model-based approach.

INTRODUCTION: During the 2013-2014 influenza season, Public Health England extended routine influenza vaccination to all 2- and 3-year-old children in England. To estimate the impact of this change in policy on influenza-related morbidity and mortality, we developed a disease transmission and surveillance model informed by real-world data. METHODS: We combined real-world and literature data sources to construct a model of influenza transmission and surveillance in England. Data were obtained for four influenza seasons, starting with the 2010-2011 season. Bayesian inference was used to estimate model parameters on a season-by-season basis to assess the impact of targeting 2- and 3-year-old children for influenza vaccination. This provided the basis for the construction of counterfactual scenarios comparing vaccination rates of ~2% and ~35% in the 2- and 3- year-old population to estimate reductions in general practitioner (GP) influenza-like-illness (ILI) consultations, respiratory hospitalizations and deaths in the overall population. RESULTS: Our model was able to replicate the main patterns of influenza across the four seasons as observed through laboratory surveillance data. Targeting 2- and 3-year-old children for influenza vaccination resulted in reductions in the general population of between 6.2-9.9% in influenza-attributable GP ILI consultations, 6.1-10.7% in influenza-attributable respiratory hospitalizations, and 5.7-9.4% in influenza-attributable deaths. The decrease in influenza-attributable ILI consultations represents a reduction of between 4.5% and 7.3% across all ILI consultations. The reduction in influenza-attributable respiratory hospitalizations represents a reduction of between 1.2% and 2.3% across all respiratory hospitalizations. Reductions in influenza-attributable respiratory deaths represent a reduction of between 0.9% and 2.4% in overall respiratory deaths. CONCLUSION: This study has provided evidence that extending routine influenza vaccination to all healthy children aged 2 and 3 years old leads to benefits in terms of reduced utilization of healthcare resources and fewer respiratory health outcomes and deaths.

Cabozantinib versus everolimus, nivolumab, axitinib, sorafenib and best supportive care: A network meta-analysis of progression-free survival and overall survival in second line treatment of advanced renal cell carcinoma.

BACKGROUND: Relative effect of therapies indicated for the treatment of advanced renal cell carcinoma (aRCC) after failure of first line treatment is currently not known. The objective of the present study is to evaluate progression-free survival (PFS) and overall survival (OS) of cabozantinib compared to everolimus, nivolumab, axitinib, sorafenib, and best supportive care (BSC) in aRCC patients who progressed after previous VEGFR tyrosine-kinase inhibitor (TKI) treatment. METHODOLOGY & FINDINGS: Systematic literature search identified 5 studies for inclusion in this analysis. The assessment of the proportional hazard (PH) assumption between the survival curves for different treatment arms in the identified studies showed that survival curves in two of the studies did not fulfil the PH assumption, making comparisons of constant hazard ratios (HRs) inappropriate. Consequently, a parametric survival network meta-analysis model was implemented with five families of functions being jointly fitted in a Bayesian framework to PFS, then OS, data on all treatments. The comparison relied on data digitized from the Kaplan-Meier curves of published studies, except for cabozantinib and its comparator everolimus where patient level data were available. This analysis applied a Bayesian fixed-effects network meta-analysis model to compare PFS and OS of cabozantinib versus its comparators. The log-normal fixed-effects model displayed the best fit of data for both PFS and OS, and showed that patients on cabozantinib had a higher probability of longer PFS and OS than patients exposed to comparators. The survival advantage of cabozantinib increased over time for OS. For PFS the survival advantage reached its maximum at the end of the first year's treatment and then decreased over time to zero. CONCLUSION: With all five families of distributions, cabozantinib was superior to all its comparators with a higher probability of longer PFS and OS during the analyzed 3 years, except with the Gompertz model, where nivolumab was preferred after 24 months.

Padeliporfin vascular-targeted photodynamic therapy versus active surveillance in men with low-risk prostate cancer (CLIN1001 PCM301): an open-label, phase 3, randomised controlled trial.

BACKGROUND: Vascular-targeted photodynamic therapy, a novel tissue-preserving treatment for low-risk prostate cancer, has shown favourable safety and efficacy results in single-arm phase 1 and 2 studies. We compared this treatment with the standard of care, active surveillance, in men with low-risk prostate cancer in a phase 3 trial. METHODS: This randomised controlled trial was done in 47 European university centres and community hospitals. Men with low-risk, localised prostate cancer (Gleason pattern 3) who had received no previous treatment were randomly assigned (1:1) to vascular-targeted photodynamic therapy (4 mg/kg padeliporfin intravenously over 10 min and optical fibres inserted into the prostate to cover the desired treatment zone and subsequent activation by laser light 753 nm with a fixed power of 150 mW/cm for 22 min 15 s) or active surveillance. Randomisation was done by a web-based allocation system stratified by centre with balanced blocks of two or four patients. Best practice for active surveillance at the time of study design was followed (ie, biopsy at 12-month intervals and prostate-specific antigen measurement and digital rectal examination at 3-month intervals). The co-primary endpoints were treatment failure (histological progression of cancer from low to moderate or high risk or death during 24 months' follow-up) and absence of definite cancer (absence of any histology result definitely positive for cancer at month 24). Analysis was by intention to treat. Treatment was open-label, but investigators assessing primary efficacy outcomes were masked to treatment allocation. This trial is registered with ClinicalTrials.gov, number NCT01310894. FINDINGS: Between March 8, 2011, and April 30, 2013, we randomly assigned 206 patients to vascular-targeted photodynamic therapy and 207 patients to active surveillance. Median follow-up was 24 months (IQR 24-25). The proportion of participants who had disease progression at month 24 was 58 (28%) of 206 in the vascular-targeted photodynamic therapy group compared with 120 (58%) of 207 in the active surveillance group (adjusted hazard ratio 0·34, 95% CI 0·24-0·46; p<0·0001). 101 (49%) men in the vascular-targeted photodynamic therapy group had a negative prostate biopsy result at 24 months post treatment compared with 28 (14%) men in the active surveillance group (adjusted risk ratio 3·67, 95% CI 2·53-5·33; p<0·0001). Vascular-targeted photodynamic therapy was well tolerated. The most common grade 3-4 adverse events were prostatitis (three [2%] in the vascular-targeted photodynamic therapy group vs one [<1%] in the active surveillance group), acute urinary retention (three [2%] vs one [<1%]) and erectile dysfunction (two [1%] vs three [1%]). The most common serious adverse event in the vascular-targeted photodynamic therapy group was retention of urine (15 patients; severe in three); this event resolved within 2 months in all patients. The most common serious adverse event in the active surveillance group was myocardial infarction (three patients). INTERPRETATION: Padeliporfin vascular-targeted photodynamic therapy is a safe, effective treatment for low-risk, localised prostate cancer. This treatment might allow more men to consider a tissue-preserving approach and defer or avoid radical therapy. FUNDING: Steba Biotech.

Pragmatic Multicriteria Decision Analysis (MCDA) Combined With Advanced Pharmacoepidemiology for Benefit-Risk Assessments of Medicines Adapted to the Real-Life Constraints of Regulators: Development and Case Study.

BACKGROUND: Multicriteria decision analysis (MCDA) represents a promising method for benefit-risk assessment. Our goal was to develop features of pragmatic MCDA (EVIDEM [Evidence and Value: Impact on DEcisionMaking]) addressing real-life regulatory decision-making needs, incorporate advanced pharmacoepidemiology, and test the resulting benefit-risk framework using a case study. METHODS: The Intervention Outcomes domain of EVIDEM was transformed into a generic benefit-risk framework including clinical efficacy, patient-reported outcomes, and adverse event (AE) criteria. The concept of relative benefit-risk balance (RBRB) was developed for comparability across products and therapeutic areas and over time. Evidence matrix was designed to include most relevant data from trials, observational studies, and models, including Bayesian and longitudinal modeling. The framework was tested with a panel of stakeholders using efalizumab for psoriasis as retrospective case study. Uncertainty was explored. RESULTS: The MCDA benefit-risk tree was adapted with psoriasis-specific subcriteria. Panelists assigned similar weights to benefits (0.48; SD, 0.20-0.70) and risks (0.52; SD, 0.10-0.60), with large variations reflecting diverse perspectives. Panelist scores reflected higher efficacy versus placebo, lower efficacy versus active comparators, and serious and fatal AEs identified postlicensing. Efalizumab's RBRB was positive at licensing in 2004 (0.29, scale -1 to +1) and ranged from -0.41 (vs active comparators) to 0.01 (vs placebo) in 2009, when its market authorization was withdrawn. Retesting indicated good reproducibility. Panelists acknowledged good face validity and the importance of criteria beyond benefit-risk in real-life assessments. CONCLUSIONS: The approach allows quantification and visualization of benefit-risk over time and across comparators. Combination of pragmatic MCDA designed to integrate criteria beyond benefit-risk and advanced statistics supports application of MCDA to further accountable benefit-risk assessments for real-life decision making.

An Indirect Comparison of Everolimus Versus Axitinib in US Patients With Advanced Renal Cell Carcinoma in Whom Prior Sunitinib Therapy Failed.

PURPOSE: The purpose of this study was to perform a weight-adjusted indirect comparison to approximate the relative efficacy of everolimus versus axitinib among patients with second-line metastatic renal cell carcinoma in whom sunitinib therapy previously failed. METHODS: Individual patient data from the RECORD-1 (Renal Cell Cancer Treatment With Oral RAD001 Given Daily) Phase III clinical trial provided information for patients taking everolimus. Summary baseline clinical and demographic characteristics and progression-free survival (PFS) outcomes were available for patients taking axitinib who were included in the AXIS (axitinib versus sorafenib) Phase III clinical trial. A Bayesian latent class mixture model differentiating responders and nonresponders and with imbedded Weibull regression on PFS was used to identify sex, Memorial Sloan-Kettering Cancer Center risk score, and time receiving prior sunitinib therapy as prognostic factors for PFS based on posterior probability >95%. Patients taking everolimus were weighted up or down based on their combination of prognostic variables. Weights were calculated by dividing the proportion of patients observed in AXIS for a given characteristic by the proportion observed in RECORD-1 and taking the product of the values derived for all three weighting variables considered. Weighted PFS distributions were derived with bootstrapped 95% CIs and compared with those reported for the AXIS trial. FINDINGS: After weighting, distributions of the 3 key baseline characteristics were more closely aligned between the 2 studies; however, some differences remained. A slightly lower rate of poor-risk patients was evident in RECORD-1 (30%) versus AXIS (36%), and a 9% lower proportion of males was observed in the everolimus group compared with the axitinib group. Distributions of time receiving prior sunitinib therapy were almost equivalent between the treatment arms. A median PFS of 4.7 months (95% CI, 3.5-10.6 months) was observed for patients in the weighted everolimus group compared with 4.8 months (95% CI, 4.5-6.4 months) in the AXIS trial. IMPLICATIONS: Similar median PFS point estimates and overlapping CIs suggest that everolimus and axitinib have similar efficacy. Although these results do not negate the need for direct comparison, this study may be used to inform clinical and reimbursement decisions until such evidence is available.

Modeling of In-Utero and Intra-Partum Transmissions to Evaluate the Efficacy of Interventions for the Prevention of Perinatal HIV.

BACKGROUND: Antiretroviral treatments decrease HIV mother-to-child transmission through pre/post exposure prophylaxis and reduction of maternal viral load. We modeled in-utero and intra-partum HIV transmissions to investigate the preventive role of various antiretroviral treatments interventions. METHODS: We analysed data from 3,759 women-infant pairs enrolled in 3 randomized clinical trials evaluating (1) zidovudine monotherapy, (2) zidovudine plus perinatal single-dose nevirapine or (3) zidovudine plus lopinavir/ritonavir for the prevention of mother-to-child transmission of HIV in Thailand. All infants were formula-fed. Non-linear mixed effect modeling was used to express the viral load evolution under antiretroviral treatments and the probability of transmission. RESULTS: Median viral load was 4 log10 copies/mL (Interquartile range: 3.36-4.56) before antiretroviral treatments initiation. An Emax model described the viral load time-course during pregnancy. Half of the maximum effect of zidovudine (28% decrease) and lopinavir/ritonavir (72% decrease) were achieved after 98 and 12 days, respectively. Adjusted on viral load at baseline (Odds ratio = 1.50 [95% confidence interval: 1.34, 1.68] per log10 copies/mL increment), antiretroviral treatments duration (OR = 0.80 [0.75, 0.84] per week increment) but not the nature of antiretroviral treatments were associated with in-utero transmission. Adjusted on gestational age at delivery (<37 weeks, OR = 2.37 [1.37, 4.10]), baseline CD4 (Odds ratio = 0.79 [0.72, 0.88] per 100 cells/mm3 increment) and predicted viral load at delivery (OR = 1.47 [1.25, 1.64] per log10 copies/mL increment), single-dose nevirapine considerably reduced intra-partum transmission (OR = 0.32 [0.2, 0.51]). CONCLUSION: These models determined the respective contributions of various antiretroviral strategies on prevention of mother-to-child transmission. This can help predict the efficacy of new antiretroviral treatments and/or prevention of mother-to-child transmission strategies particularly for women with no or late antenatal care who are at high risk of transmitting HIV to their offspring. TRIAL REGISTRATION: This analysis is based on secondary data obtained from three clinical trials. ClinicalTrials.gov. NCT00386230, NCT00398684, NCT00409591.