E&O_lecture 5_biomarkers_surrogate_endpoints_2021.pptx

biomarkers and surrogate endpoints Metrics – exposure and outcomes Lecture 5 Annelies Van Rie  What is a biomarker? Biomarkers and measurement error Biomarkers for clinical research Surrogate endpoints: what are they Topics fo...

biomarkers and surrogate endpoints Metrics – exposure and outcomes Lecture 5 Annelies Van Rie  What is a biomarker? Biomarkers and measurement error Biomarkers for clinical research Surrogate endpoints: what are they Topics for and why are they used? today’s class Issues with surrogate endpoints Correlation is not causation Where in the causal pathway/cascade? Biomarkers for environmental and occupational research  A biomarker is an objective, quantifiable, accurate and reproducible measurement of an exposure or an outcome Biomarkers are used instead of or to complement subjective variables, self-reported Biomarkers variables or difficult to measure variables. Surrogate endpoint = biomarker to measure : definition outcome A biomarker that predicts the outcome Surrogate endpoints are used instead of ‘hard’ clinically relevant outcomes What does a biomarker measure? A biomarker can be an indicator of a normal biological process – 25(OH) vit D plasma level (sunlight on positive emotion, not on causal pathway) a pathogenic process – tumor size, histological cancer staging a pharmacologic response to a therapeutic intervention - plasma level of a drug Biomarkers often reflect an interaction between a biological system and a health hazard Pathophysiological logical effect of presence of a pathogen – CD4 count in PLWHIV Concentration of the agent of interest - lead Concentration of products of biotransformation – 25(OH)vit D Biological effects that result from contact with the agent – skin rash (rash density) Structural change caused by the agent and detected by imaging – cavity (TB- lung tissuse)  Measurement error = difference between the measured biomarker and the true exposure Biomarkers True exposure = the underlying biological or and external factor that the biomarker is meant to measure without laboratory or other source of measurem error the causal factor for etiological research ent error the predictor in predictive research If the biomarker is not fixed and can fluctuate over time, the true exposure is integrated over the time of interest the average of the true exposure over the etiologically important time period for etiological studies The many sources of error in biomarker s Common 1. Variability and/or error introduced during specimen collection, sources of processing, and storage measurem 2. Variability and/or error introduced ent error during specimen laboratory analysis Within-batch (intra-assay) variability introduced Between batch (inter-assay) variability Between laboratory variability when Between technician variabiliy using 3. Within-person variability of the biomarker value over time bomarkers  Which type of sample is best suited? Different specimen types may reflect different exposure Reflection periods Folate and selenium concentrations in plasma represent recent dietary intake (only eat), while s concentrations measured in red blood cells reflect longer-term body stores (store + eat) Participation rates may differ for non-invasive compared to invasive specimens (more and less measurement error due to required When should missing data)the sample be collected? Diurnal variation (daily) to reduce Etiological window Steady state and half life (doesn’t matter what time point you measure it) measure How should the sample be collected? the type of anticoagulant used (type blood tube) alters ment measurable plasma concentrations of various hormones or cytokines How should samples be stored? error different storage methods can influence assay variability. Degradation of free prostate-specific antigen concentrations is twice as fast when stored at –20C versus –70C Freeze thaw cycles can impact the measurement The role of SOPs in preventing measurement error The SOP need to list specific details with regards to Instruction to participant (fasting or not,…) Method of specimen collection Handling of the specimen (room temperature or on ice,…) Specimen processing before storage (RNA degradation) Laboratory processing procedures Quality control  Avoid differential error Differential Biomarkers should be processed similarly for cases and controls regarding all criteria that may affect the assays and Blinding should occur at all stages of the process use the same storage containers (e.g., cryovials), labels, and labeling system for all samples vial identifiers should be similar for study and quality non- control samples samples should be randomly allocated within the entire batch differential Avoid non-differential error measurem Communicate with laboratory investigators before and during the study ent error Perform a pilot studies to assess collection feasibility and its effects on assay variability Follow standard procedures for collection, shipping, storing and processing samples Include QC samples in each batch include all samples from a particular subject (longitudinal study) or related group (family study) in the same batch. Why case control studies are especially vulnerable to measurement error in biomarkers Differential collection, processing, storage, analysis of samples between cases and controls Laboratory staff not blinded to case or control status Specimens on exposure are obtained after the diagnosis for the cases and the biological effects of the disease or its treatment affects the exposure biomarker Solution: cohort study? Reduces many issues of case control studies, but may need to exclude cases in which the biomarker was measured in the preclinical phase of the disease (at a minimum in a sensitivity analysis)  Biomarkers may not necessarily correlate with a person’s experiences or a patient’s clinical state Measurable changes or differences in biomarker values may be undetectable Concerns by individuals beyond Variations in biomarkers can be without effects on health measurem The goal of clinical interventions is to reduce morbidity and mortality, not to ent error change quantifiable features of patients that have no clinical effect Patients seek treatment for their diseases, not for the numerical measures that frequently but always or not perfectly correlate with their illness Kyle et al. Tavel. Curr Opin HIV AIDS.  Biomarkers for clinical research  Human or environmental Endobiotic or xenobiotic Endobiotic: normally present in the human body and needed for healthy bodily function Types of Example: cholesterol biomarkers Xenobiotic: agent that is foreign to the human body Example: lead Diagnostic or prognostic Susceptibility, exposure or outcome  Fang-Shu Ou, 2020 Clinical use of a biomarker Biomarkers: spectrum for use and specifications Infectious or Cance Non-communicable r diseases Terminology in cancer biomarkers Yanqiu Liu et al. Front. Biomark ers across the disease pathway Biomarkers in clinical trials EC: endothelial cell CEC: circulating endothelial cells CECP: circulating endothelial cell progenitors US, ultrasound CT, computed tomography; DCE-MRI: dynamic contrast- enhanced MRI PET: positron emission tomography FACS: fluorescence activated cell sorting IHC: immunohistochemistry RT-PCR, reverse transcription- PCR www.nature.com/clinicalpractice TTP: time to progression doi:10.1038/ncponc1150 Surrogate endpoints What is a surrogate endpoint? Outcome measures can be Direct, clinically meaningful endpoints that are direct measures of how patients feel, function and survive. Indirect or surrogate endpoints that replace endpoints and act as “surrogates” for clinically meaningful endpoints A surrogate endpoint is a laboratory measure (or sign or symptom) used as a substitute for a clinically relevant outcome Examples of surrogate endpoints Characteristics of a surrogate endpoint? Any surrogate endpoint should provide reliable data about whether the treatment or intervention provides clinically meaningful benefit there must be an association between the response by surrogate measures and the response by true clinical outcomes (only with temporal differences) A valid surrogate end point must both correlate with and accurately predict the outcome of interest A ‘hard’ outcome or surrogate endpoint? Golden rule: focus on clinically relevant outcomes What impact a patient’s life? morbidity (quality of life) or mortality (survival time) It is often not easy task to perform a controlled, randomized, clinical study with a hard, patient relevant outcome (morbidity and mortality is often a rare outcome) It is logical (and ethical) that researchers attempt to measure something less severe and then correlate this with the ‘hard’ outcomes by statistical methods with the use of endpoints validated in the literature. Surrogate endpoints are therefore increasingly used in clinical studies  Primary clinical endpoints, such as death, can be rare Why is Clear clinical endpoints may only develop many years there an after the intervention. (long study) Surrogate endpoints can provide interim evidence about explosion of the safety and efficacy of treatments while more definitive clinical data is collected. interest in The US FDA allows accelerated approval of an overall drug development plan by allowing submission of an surrogate application and marketing of a drug based on the evidence obtained using a surrogate endpoint, with endpoints further studies demonstrating that direct patient benefit is underway in clinical Surrogate endpoints can allow researchers to design smaller, more efficient studies, thus reducing the research? number of subjects exposed to an experimental treatment and shortening the time to approval of new effective treatments Biomarkers and surrogate endpoints: effect on sample size in observational research Power of a study depends on cohort size, variability of data, length of follow-up, and frequency of the disease of interest Biomarkers and surrogate endpoints can increase the power of observational studies by improving exposure assessment, hence decreasing exposure misclassification improving accuracy of health-risk estimates identify a more homogeneous population reduce biases due to health outcome misclassification. Pernot et al. Cancer Epidemiology, Biomarkers & Prevention. 2014 Potential issues of the use of surrogate markers: reflecting about causal pathways Key question: do the effects of a treatment/intervention on a surrogate endpoint reliably predict the effects of a treatment/intervention on a clinically meaningful endpoint? Multidimensional causal (causal pie) pathways can result in false positive or false negative results when assessing effectiveness using surrogate endpoints Surrogate endpoints that are strongly associated with clinical efficacy measures in natural history observations but are not in the causal pathway of the disease process can provide misleading information about clinical efficacy It often is unclear what the magnitude and duration of effect on a causal pathway is required to meaningfully affect the clinical efficacy measure Flemming and Powers Stat Med. Examp le Interferon Biomarker: CD4 count HIV disease in Mother to child HIV infected pregnant transmission of infant women HIV Causal pathway: HIV virus (viral load) Antiretroviral drug Nevirapine Risk of MTCT of HIV increases with decreases in CD4 count Treatment with interferon increases CD4 count but does not reduce MTCT Only interventions on the causal pathway (HIV VL) will have a clinically meaningful effect Flemming Stat Med.  The primary efficacy end point of standard phase III trials in TB is an unfavorable outcome Biomarke (treatment failure, recurrence, or death or study dropout during treatment) measured 24 months after the end of treatment rs do not Month 2 culture conversion is (was) considered a valid surrogate endpoint for treatment efficacy in always tuberculosis Use of this surrogate endpoint could accelerate the approval of new TB drugs by several years work Phase II trials showed significantly higher rates of month 2 culture conversion of a 4-month gatifloxacin regimen compared to the standard 6-month regimen Noninferiority of the 4-month regimen to the TB drugs standard regimen with respect to the primary efficacy end point was not shown, even in the presence of higher month 2 culture conversion rates  When used as outcomes in clinical trials, biomarkers can act as surrogates or substitutes for clinically meaningful endpoints To be a valid surrogate endpoint, there must be Surrogate solid scientific evidence that a biomarker consistently and accurately predicts a clinical endpoints: outcome (benefit or harm) caution is Even biomarkers that have been statistically validated as surrogates for a clinical endpoint needed may not be part of the pathophysiological pathway that results in that endpoint, resulting in mistaking correlation for causation Surrogate endpoints never prove conclusively that a treatment result in a particular clinical endpoint. Rather, the research process progressively reduces uncertainty about the relationship between an intervention, a surrogate endpoint, and a clinical endpoint Kyle et al. Tavel. Curr Opin HIV AIDS.  CONCLUSION: “A difference to be a difference must make a difference” Surrogat e endpoin ts Where to measure in the cascade of effects? Should diagnostics save lives? The effectiveness of an innovation being evaluated (eg, a diagnostic) depends on events further downstream in the care cascade, which in turn depend on health system context. How can we design innovative studies to show if and how surrogate endpoints alter subsequent causal events or influence patient outcomes?  What level of scientific evidence is needed for a valid biomarker? Ideally, what one would need to have is a comprehensive understanding of: Principal causal pathways through which the disease process affects how a patient feels, functions or survives Extent to which degree the surrogate endpoint captures the meaningful “on-target” effects of the intervention on the causal pathways of the disease process Existence of any “off-target” effects of the intervention that would meaningfully affect the clinical efficacy measures and yet would not be captured by the surrogate endpoint Unfortunately, this type of knowledge is seldom Flemming and Powers Stat Med. available  Measurement validity: can the surrogate endpoint be measured objectively and reproducibly? Evaluation Internal or study validity : within the study, can the surrogate endpoint be measured not of just with precision and reproducibility, but also with accuracy and does the endpoint correlate surrogate strongly with the clinical endpoint for which it is serving as a surrogate? endpoints External validity: can this surrogate be used in closely related to the studies, i.e. other populations or other related treatment studies? External validity: can biomarkers be relied upon to serve as surrogates for other related clinical endpoints or other classes of treatments, beyond the contexts in which they were developed? Surrogate endpoints in a complex and messy world: what is the way forward? We need a shift from a fixation on tools to patient-centered solutions, from trials of standalone innovations to evaluations of complex, multisectoral health interventions Start by mapping out the exact point in the care pathway in which an innovation is inserted and theorize how it may make a difference and what barriers may impede its effects on health outcomes. Use theory-driven health systems and implementation research to confirm or refute assumptions of how an innovation might work along the mapped- out care pathway and examine the impact of innovations on the surrogate endpoints along the care cascade. Do not create unreasonable expectations and take into account the difference between surrogate endpoints and patient outcomes in dissemination of findings Pai et al. BMJ Global Health 2018 Biomarkers for environmental and occupational research  Environmental agents include: Contaminants of the general environment: soil, water, air Agents used in the personal environment: food, cosmetics, utensils Toxic agents in specific environments: home, workplace, recreational Exposure to environmental agents is often unknown or un-sensed by individuals, complicating accurate documentation through questionnaires Individual vs population measurement of present exposures Population level measurement Ecological approach Assumptions about individual behaviors that influence the exposure Individual measurement: Personal sensors Takes into account personal behavior (especially avoidance of exposure) Active and passive samplers Not always acceptable for prolonged periods of time Biomarkers are moving into people’s & patient’s daily life Biomarkers: role of dose and route of exposure Potential dose is the amount of the chemical ingested, inhaled, or applied to the skin. Applied dose is the amount that is absorbed or deposited in the body Internal dose is the amount that is available for interaction with biologically significant molecular targets. Biologically effective dose is the amount that has interacted with a target site and alters a physiologic function.  Sampling for exposure measurement Random selection of subjects and grouping of those with apparently common levels of exposure May result in very few measurements of exposures at both ends of the spectrum – which are often the ones of greatest interest in epidemiological studies Pre-defined strata of exposure and random sampling within strata Exposure groups are defined based on knowledge of tasks of personnel, personal activities, sources of contaminants, devices used for removal of contaminants, … Each zone should meet 4 criteria: Similarity in work or other activity – similar tasks/activities Similarity with regards to hazardous agents – similar potential for exposure Environmental similarity – similar personal protection, ventilation,.. Identifiability and non-overlap Value of measuring past exposure? Assumption that current and past exposure are similar Infer best estimates by combining current and past exposure measures How was sampling performed? Same as exposure risk group of interest? Similar methods used? Same mixture measured? … CONCLUSION Biomarkers are frequently used in any type of epidemiological research Great care in study design and study execution is needed to avoid measurement error introduced through the use of biomarkers Especially case control studies are vulnerable to differential error Surrogate markers are increasingly popular Need to ensure that there is proven (or at least high likelihood) that the surrogate endpoint is associated with the ‘hard’ clinically meaningful endpoint In diagnostic research: what are clinically meaningful and realistic ‘hard’ endpoints ?