Epidemiological study designs, 3 E2040: Introduction to Epidemiology and Environmental Health Andrea Dalecká, PhD Kasia Kordas, PhD > Today’s learning objectives •Describe the features of intervention studies •Describe the reason for and process of randomization in intervention studies •Discuss advantages and disadvantages of intervention studies •Explain what problems with epi studies randomization addresses But first, a bit of review Epidemiological study Epidemiology allows us to quantify the association between exposure/factor and outcome Data for this comes from epidemiological studies Different epi designs are appropriate for different purposes and types of questions We are here today Study designs can be grouped by type Descriptive Analytic Observational Intervention And a few words about bias What is bias in epi? Three categories of bias 4: Confounding bias. Smoking is an example of distorted association... | Download Scientific Diagram Finally, to intervention studies… Goal of public health What is an intervention study? •Intervention studies are research designs where researchers actively assign participants to receive specific interventions to assess their effects on health outcomes. • •These studies are distinct from observational studies, where researchers simply observe and record outcomes without manipulating any variables. • •Investigator intervenes; controls what the exposure is; chooses which participants are exposed and not exposed • •Intervention studies evaluate both, the benefits and the risks of new forms of intervention (medications, exercise, vaccination… etc.) • •Also known as randomized controlled trials (RCT). • There are different types of RCT… > Clinical vs Laboratory Research A Comparative Analysis for 2024 - CCRPS > Feedback and Analysis on the Safety of Fluoride in Drinking Water – KnowYourH2O > Fortify Health on LinkedIn: #fortifyhealth #ffi #ffinetwork #foodfortification Clinical trials •Test new medical, surgical, or diagnostic interventions in human patients. •They are primarily designed to determine the safety and efficacy of these interventions. •Successful clinical trials can lead to the development of new therapies for various diseases There are different types of RCT… Non-clinical trials Oriented on public health and behavioral changes. Examples: •Water fluoridation, •grain fortification with folate, •workplace exercise program, diet, •wearing a step counter, •school-based health clinic... 1.Define the population to study 2.Sample people from that population 3.Randomly assign the exposure/intervention (X) or the control •Inactive treatment (placebo) •Active treatment (standard of care) •It is “randomized” because exposure is assigned randomly 4.Follow up for the outcome (Y) •How many in each group have developed disease? How many have improved? 5.Compare outcome between exposed and unexposed. We can calculate: •Risk (rate) ratio of outcome in exposed compared to unexposed •Risk difference • • • What happens in a randomized trial? We’re measuring incident (new) outcomes here, so we can measure risk! How are subjects selected? •Based on eligibility criteria (also called inclusion and exclusion criteria) •Characteristics that define the individuals you want to participate in the study •Inclusion criteria: •Participants with the potential to benefit from the intervention and a high probability of developing the outcomes of interest •Exclusion criteria: •Higher risk of unwanted events (allergic reactions, pregnant women, children) •Risk of not complying with the study protocol • •Must be explicitly specified in a study protocol, before the study begins. Why? -Allows clear selection of the sample -Allows homogeneity of the sample -Allows others to reproduce the study, assess if the sample was appropriate for the research question -Identifies the individuals to whom findings apply Example Why randomize? •Different people participate in randomized trials •Older vs. younger •Men vs. women •Sicker vs. healthier •Better vs. poorer: SES, physical activity, mental health, sleep, diet •Different people respond differently to experimental condition/treatment •Possibly due to genetic traits •Possibly due to above characteristics • Purpose of randomization Condition B Condition A Who will tend to do better? Purpose of randomization Condition A Condition B •If there are enough participants, we hope that randomization will increase the likelihood that the groups will be comparable on characteristics about which we may be concerned (such as sex, age, race, and severity of disease). •Randomization does not guarantee comparability – by chance, groups could differ. If groups that are being randomized are large enough and the randomization procedure is free of bias, they will tend to be similar. •The critical element of randomization is the unpredictability of the next assignment A diagram of a computer with people around it Description automatically generated Badge Question Mark outline https://mrctcenter.org/clinical-research-glossary/glossary-words/randomization/ There are different types/levels of randomization •Individual – each person is randomized to an experimental/treatment group and the individual is the unit of analysis • • •Cluster – a group of people is randomized and the cluster is the unit of analysis • Condition A Condition B Condition A Condition B Example clusters for randomization Individual vs cluster RCT Individual vs cluster RCT Why would you wish to cluster-randomize? •Unethical to assign individuals (ex., breastfeeding promotion in baby-friendly hospital) •Difficult to implement experiment on individuals (ex., children in daycare) •Difficult to mask assignments. Example cluster-randomized trial Mentor mothers in the video intervention arm provided standard-of-care counseling plus the MOVIE intervention; mentor mothers in the control arm provided standard of care only. The video intervention mentor mothers were tasked with delivering the video intervention during their regular perinatal home visits, which typically include counseling on infant feeding methods. Example cluster-randomized trial Stratified randomization •Concern about comparability of the groups in terms of one or a few important characteristics that may influence response to intervention/experiment (Sex, classroom, city, etc.) •Randomization does not 100% ensure comparability •Randomization within strata can help increase the likelihood of comparability of the study groups. •We first stratify our study population by each variable that we consider important and then randomize participants to treatment groups within each stratum. • Older ages Younger ages Older ages Younger ages Condition A Condition B Condition A Condition A Condition B Condition A Condition B What is randomisation? Let’s review randomized trials Blinding •Blinding refers to the concealment (masking) of group allocation from one or more individuals involved in a RCT. •Blinding is particularly important when: •The outcome is a subjective measure, such as pain •The intervention may affect the participant‘s behaviour (measured as exposure) • •How can subjects be masked? •Using a placebo (an inert substance that looks, tastes, and smells like the active agent) • •Use of a placebo does not automatically guarantee that the patients are masked •Some participants may try to determine whether they are taking the placebo or active drug • https://nautil.us/sugar-pill-nation-294292/ A close-up of a bottle of pills Description automatically generated Types of blinding •Participants are not aware whether receiving treatment or placebo. Neither the subject nor the researcher is aware of the participant's current treatment. The intervention being studied is concealed from the participants, the individuals administering the treatment, and the evaluators of the results Why do we double- or triple- blind the RCTs? •Observers who think the intervention is effective may probe the subjects who are taking the intervention differently (Observer bias) •Clinicians who know the subject is taking the placebo may give extra care to compensate not receiving an active treatment (Co-intervention bias) • https://www.covid19recovery.net/blog/understanding-clinical-trials •Happens in behavioral, educational, or clinical care interventions • •Cross-over trials •Participants serve as their own controls • •Wait-list control •Everyone gets the intervention, but some people randomized to wait before they get it What if blinding is not possible? No intervention Control Intervention Wait Control Intervention Factorial design •You can test two interventions at once! How convenient! •Ex., Harvard Women’s Health study. Randomized women to both vitamin E and low-dose aspirin (LDA): • • • • • • •Neither drug had strong effects on cardiovascular disease or cancer. There were some weak effects on specific outcomes like stroke. •Factorial designs also allow testing of effect modification: ex., is the effect of vitamin E and aspirin taken together stronger than the sum of their individual effects? • Vitamin E: yes Vitamin E: no Low dose aspirin: yes Vit E and LDA LDA only Low dose aspirin: no Vit E only Neither (placebo) Possible issues during the RCTs • Data collection: Information bias •We want comparable measurement between intervention and control groups: this reduces information bias •To prevent information bias, you need to measure outcome equivalently for intervention and control groups • •Example, exercise frequency and intensity: •In a study of online video intervention to increase exercise, the intervention group may report it more accurately because they’re more attentive to it •One way to solve this: not rely on self-reports. Maybe look at activity time or steps on a smartwatch. There are privacy and feasibility concerns, though. •Information bias can affect randomized trials. Randomization doesn’t solve information bias! •Randomization doesn't improve how we collect information • Drop-ins •People in one group may unintentionally take the agent assigned to the other group. •For example: •In a trial of aspirin vs placebo, participants in the placebo group may take aspirin as an over-the-counter medication •Two steps were taken to address this problem: •(1) controls were provided with lists of aspirin-containing preparations that they should avoid, •(2) urine tests for salicylates were carried out both in the aspirin group and in the controls. • Non-compliance •Noncompliance arises when participants do not receive the treatment or intervention to which they were randomly allocated. •For example, some participants invited to go through a screening program may not attend. • •Noncompliance may be overt or covert: people may overtly articulate their refusal to comply or may stop participating in the study (non-compliers are also called dropouts from the study). •People may just stop taking the agent assigned without admitting this to the investigator or the study staff. •Whenever possible, checks on potential noncompliance are built into the study (for example, urine tests for the agent being tested or for one of its metabolites) •Consequences: •The treatment group will include some who did not receive the therapy, and the no-treatment group may include some who received the treatment. •The experimental groups are less different. •Noncompliance may be differential (related to specific characteristic)à bias. • Recruitment and retention of participants •Recruitment of sufficient numbers of eligible and willing volunteers is major challenge in randomized trials •Especially if treatment is long or difficult (has side effects, many tests or procedures). •Failure to recruit enough volunteers can leave a well-designed trial without enough participants to produce statistically valid results. •Within the limits of a randomized trial, participants must be fully informed of the risks and what arrangements have been made for their compensation if untoward effects occur. •A related problem is retaining volunteers for the full duration of the study •Participants may lose interest in the study over time, or find participation too inconvenient, long. •Investigators understand why participants often drop out of studies and develop appropriate measures to prevent losses to follow-up. • • • External vs. internal validity A diagram of different types of population Description automatically generated •Main objective is to generalize the results of a trial beyond the study population itself. •Our ability to apply the results to a broader population is called the generalizability, or external validity, of the study. • •External validity should be distinguished from internal validity. •A randomized trial is internally valid if the randomization has been properly done and the study is free of other biases and is without any of the major methodologic problems that have been discussed. •It should be ideally concluded that the observed differences in the outcomes in the two groups being compared are attributed to the hypothesized exposure under study, aside from sampling errors. Let’s wrap this up •Does randomization fix these problems? •Confounding: yes! It ensures that nothing but random chance is related to X •Selection bias: •yes if it’s due to selection into the study! It ensures that nothing but random chance is related to X •no if it’s due to loss to follow-up! Because reasons people drop out can still be related to X and Y •Generalizability: no! People who join the trial may still be different than people who don’t join •Information bias: no! There can still be measurement inaccuracies in X or Y X = Exposure Y = Outcome How to express results from a randomized trial? A trial for a new drug shows that 30% of patients in the treatment group and 50% in the control group experienced an adverse event. In clinical trials (like in any epi study), we also need to consider whether the result is clinically significant. What is that? Clinical trial registration & reporting •To avoid selective reporting, the International Committee of Medical Journal Editors adopted a policy, which became effective in 2005, that all clinical trials must be registered in a public registry before any participants are enrolled in the study. •www.Clinicaltrials.gov •Clinical trial reporting must follow established guidelines A checklist of information Description automatically generated Is using a placebo ethical? •Placebo-controlled trials •Use of an active treatment comparator in a clinical trial of a new therapy is generally the appropriate trial design when an established effective therapy exists •Placebo can be ok in some circumstances: •No established therapy •Existing evidence raises significant doubt by medical experts regarding benefit of existing therapy •Patients are resistant to existing therapy due to previous history • • Žlutý otazník Questions? Andrea Dalecká andrea.dalecka@recetox.muni.cz