Study Design Obtaining and Describing Samples Matching Matching

Study Design Obtaining and Describing Samples

Matching • Matching involves selecting controls so that the distribution of potential confounders (e. g. age or smoking) is as similar as possible to that amongst the cases. In practice this is only utilized in case-control studies, but it can be done in two ways: • Pair matching - selecting for each case one or more controls with similar characteristics (e. g. same age and smoking habits) • Frequency matching - ensuring that as a group the cases have similar characteristics to the controls

Inclusion/Exclusion Criteria • In a clinical trial, the investigators must specify Inclusion and exclusion criteria for participation in the study.

Inclusion/Exclusion Criteria • Inclusion criteria are characteristics that the prospective subjects must have if they are to be included in the study, while exclusion criteria are those characteristics that disqualify prospective subjects from inclusion in the study.

Inclusion/Exclusion Criteria • Inclusion and exclusion criteria may include factors such as age, sex, race, ethnicity, type and stage of disease, the subject’s previous treatment history, and the presence or absence (as in the case of the “healthy” or “control” subject) of other medical, psychosocial, or emotional conditions.

Selecting Controls • In case-control studies, selection bias can occur in the selection of cases if they are not representative of all cases within the population, or in the selection of controls if they are not representative of the population that produced the cases.

Selecting Controls • For example, in a hospital-based case-control study looking at the relationship between alcohol consumption and development of liver cirrhosis, in the first instance controls are selected from patients hospitalized due to trauma (Controls A). The exposure (alcohol consumption) is classified into 'heavy alcohol use' and 'light / no alcohol use'. Field Epidemiology Manual: https: //wiki. ecdc. europa. eu/fem/w/wiki/selection-bias-and-case-control-studies

Selecting Controls • But, how representative are hospitalized trauma patients of the population which gave rise to the cases? In the trauma ward, where the controls have been selected, there may be a higher proportion of patients who report heavy alcohol use compared to those who report heavy drinking in the population which produced the cases, leading to an underestimation of the odds ratio (OR). Compare this to the situation if controls are selected from hospitalized patients in a non-trauma ward (Controls B). Field Epidemiology Manual: https: //wiki. ecdc. europa. eu/fem/w/wiki/selection-bias-and-case-control-studies

Lack of Controls • Lack of controls can impact whether an experiment is technically good and can render results unreliable. • For example, an experiment might control for the application of a fertilizer, but fail to control for the environment required for the application.

Allocation Concealment • Concealed allocation is a procedure implemented in a randomized control trial where the individuals screening and separating the candidates into two (or more) arms of a study are blinded. • This is a consideration beyond blinding the practitioner delivering the care or the patients receiving the care.

Allocation Concealment • Even if there is a randomization process in place, it matters how it is actually implemented. • To be effective, the process must ensure that the investigators (or providers or subjects) CANNOT influence the group each person ends up in. The allocation into the groups must be “concealed. ” Otherwise, the researcher, provider delivering care, or the subjects themselves may consciously or unconsciously manipulate who gets allocated to which arm of the study—thusly defeating the benefits of randomization.

Allocation Concealment • Trials with unconcealed allocation, compared to trials on the same interventions where allocation was concealed, consistently overestimated the benefit of a treatment by as much as 40% (as measured by odds ratios). • • Schultz KF, Chalmers I, Hayers RJ, et al. Empirical evidence of bias JAMA 1995; 273: 408 -12 Schulz KF, Grimes DA. Allocation concealment in randomized trials: defending against deciphering. The Lancet 2002; 359(9306): 614 -18. Pidal J, Hrobjartsson A, et al. Impact of allocation concealment on conclusions drawn from meta-analyses of randomized trials. Int J Epidemiol 2007; 36: 847 -857 Moher D, Ba’Pham AJ, Cook, DJ et al. Does quality of reports of randomized trials affect estimates of intervention efficacy reported in meta-analysis? The Lancet 1998; 352(9128): 609 -13

Allocation Concealment • When allocation is not concealed, it doesn’t mean that the results are not valid, but it means we don’t know. Because it can potentially introduce a lot of bias, most researchers now would consider the study significantly flawed.

Randomization • In randomized controlled trials, the research participants are assigned by chance, rather than by choice, to either the experimental group or the control group. • Randomization reduces bias as much as possible. Randomization is designed to "control" (reduce or eliminate if possible) bias by all means.

Randomization • The completely randomized design is probably the simplest experimental design, in terms of data analysis and convenience. With this design, participants are randomly assigned to treatments. • A completely randomized design relies on randomization to control for the effects of extraneous variables. The experimenter assumes that, on average, extraneous factors will affect treatment conditions equally; so any significant differences between conditions can fairly be attributed to the independent variable. Stat Trek, Experimental Design: http: //stattrek. com/experiments/experimental-design. aspx? Tutorial=AP

Randomization • With a randomized block design, the experimenter divides participants into subgroups called blocks, such that the variability within blocks is less than the variability between blocks. Then, participants within each block are randomly assigned to treatment conditions. Because this design reduces variability and potential confounding, it produces a better estimate of treatment effects. • The randomized block design is an improvement over the completely randomized design. Both designs use randomization to implicitly guard against confounding. But, in this example, only the randomized block design explicitly controls for gender. Stat Trek, Experimental Design: http: //stattrek. com/experiments/experimental-design. aspx? Tutorial=AP

Randomization • A matched pairs design is a special case of the randomized block design. It is used when the experiment has only two treatment conditions; and participants can be grouped into pairs, based on some blocking variable. Then, within each pair, participants are randomly assigned to different treatments. • In this example, 1000 participants are grouped into 500 matched pairs. Each pair is matched on gender and age. For example, Pair 1 might be two women, both age 21. Pair 2 might be two women, both age 22, and so on. The matched pairs design is an improvement over the completely randomized design and the randomized block design. Like the other designs, the matched pairs design uses randomization to control for confounding. However, unlike the others, in this example, this design explicitly controls for two potential lurking variables - age and gender. Stat Trek, Experimental Design: http: //stattrek. com/experiments/experimental-design. aspx? Tutorial=AP

Simple Random Sampling • In simple random sampling, every unit has an equal chance of being selected Other Effective Sampling Methods: https: //faculty. elgin. edu/dkernler/statistics/ch 01/1 -4. html

Stratified Sampling • In stratified sampling, the population is separated using some characteristic, and then a proportional random sample is taken from each. • Stratum/groups are created and then units are picked randomly. Other Effective Sampling Methods: https: //faculty. elgin. edu/dkernler/statistics/ch 01/1 -4. html

Systematic Sampling • A systematic sample is obtained by selecting every nth individual from the population. The first individual selected corresponds to a random number between 1 and n.

Cluster Sampling • • • In stratified sampling, the population is split up into groups (strata) based on some characteristic. A cluster sample is obtained by selecting all individuals within a randomly selected collection or group of individuals. Cluster sampling is used when the population is already broken up into groups (clusters), and each cluster represents the population. A certain number of clusters is selected.

Question • A study evaluated patients with degenerative joint disease of the knee who were randomized to receive physical therapy or surgery. The patients and the researcher were not told which treatment the patients would get until after they were enrolled into the study. Which of the following does this illustrate: • Allocation Concealment • Matching • Stratification

Answer • Allocation Concealment • A study can be “unblind” in the traditional sense but yet the allocation can be concealed. • If the patients had known what they were going to get, either surgery or physical therapy, the patients may have elected not to be enrolled, and, therefore, the generalizability of the study might have been open to question. Once they were in the study, though, both the patients and the treating physician or physical therapist were certainly aware of the treatment the patient was receiving.

Question • A researcher wants to select a sample of 10 people from a population of 100. She has a list of all 100 people. She assigns each person a number from 1 to 100. The researcher then picks a number, in this example 6, as the starting number. She then selects every tenth person for the sample. This is an example of: • Cluster Sampling • Simple Random Sampling • Stratified Sampling • Systematic Sampling

Answer • Systematic Sampling • She selects every tenth person for the sample (because the sampling interval = 100/10 = 10). The final sample would contain those individuals who were assigned the following numbers: 6, 16, 26, 36, 46, 56, 66, 76, 86, 96