Glossary

A   |  B   |  C   |  D   |  E   |  F   |  G   |  H   |  I   |  K   |  L
M   |  N   |  O   |  P   |  Q   |  R   |  S   |  T   |  V   |  W   |  Z

Glossary terms for 'P'

P valueBased on statistical tests, the probability of finding an effect (more precisely, a value of the test statistic) as large or larger than that found in the study by chance alone if the null hypothesis is really true. For example, if the null hypothesis is that drinking coffee is not associated with the risk of myocardial infarction, and the study found that the relative risk of myocardial infarction among coffee drinkers compared with nondrinkers was 2.0 with a p value of 0.10, there was a 10% probability of finding of a relative risk of 2.0 or larger in this study if there was no association between coffee drinking and myocardial infarction in the population.
Paired measurementsMeasurements closely linked with one another in some way, such as those done on different sides of the same person, different members of a twin pair, or (most commonly) those done on the same participant at two different points in time, such as before and after an intervention. For example, in a study of the effect of an exercise program on glycohemoglobin levels in patients with type II diabetes, paired measurements of glycohemoglobin included measurements made at baseline and again after 3 months of exercise.
ParticipantSomeone who participates in a research study. The term “participant” is often preferred over “subject” because it emphasizes that the person enrolled in the study is an active participant in advancing science, not merely a subject being experimented upon. For example, in a study of a new drug for treatment of insomnia, the participants are the people who are eligible for and enroll in the study.
Peer reviewReview of a protocol, proposal or manuscript by peers of the investigator who prepared these documents. For example, proposals submitted for funding to the NIH undergo a peer review process in which scientists in the same field score the protocol using well-defined criteria. Similarly, manuscripts submitted to medical journals are peer reviewed by scientists who help the journal editors decide whether the manuscript should be published.
Per-protocol analysisIn a clinical trial, an analysis approach in which data from participants are only included if the participants adhered to the study protocol, which is typically defined as taking or using the study intervention as instructed. For example, in a randomized trial of surgery compared with physical therapy for treatment of severe osteoarthritis of the knee, a per protocol analysis would include data only from participants in the surgery group who actually underwent surgery and from participants in the physical therapy group who were adherent to the physical therapy regimen. See also intention-to-treat analysis.
Person-timeThe sum of the amounts of time each of the subjects in a study or population is at risk, used as the denominator for calculation of incidence rates. It can be calculated as the number of subjects who are at risk of an outcome multiplied by their average time at risk. For example, the total person-time of follow-up among the 1000 subjects who had an average of 2.5 years at risk was a total of 2500 person-years, although 5% of the subjects were followed for 1 month or less. See also incidence rate.
Phase I trialAn early phase, generally unblinded, uncontrolled trial of escalating doses of a new treatment in a small number of human volunteers to test its safety. For example, a phase 1 trial of a new drug for treatment of menopausal hot flashes would generally include a small number of volunteers (with or without hot flashes) who receive escalating doses of the drug to determine its effects on blood counts, liver and renal function, physical findings, symptoms, and other unexpected adverse events.
Phase II trialA small randomized (and preferably blinded) trial to test the effect of a range of doses of a new treatment on side effects as well as on surrogate or clinical outcomes. For example, a phase II trial of a new drug for treatment of hot flashes that has been shown to be safe in a phase I trial might enroll a small number of postmenopausal women with hot flashes, randomly assign them to 2 or 3 different doses of the new medication or placebo, and then follow them to determine the frequency of hot flashes, as well as side effects.
Phase III (pivotal) trialA randomized (and preferably blinded) trial that is large enough to test the efficacy and safety of a new treatment. For example, if the optimal dose of a new treatment for hot flashes has been established in a phase II trial and the new treatment was acceptably safe, the next step would be a large phase III trial in which postmenopausal women with hot flashes are randomly assigned to the new treatment or placebo and followed for the occurrence of hot flashes and adverse effects.
Phase IV trialA large study, which may or may not be a randomized trial, conducted after a drug is approved by a regulatory agency such as the US Food and Drug Administration (FDA), often to determine the drug?s safety over a longer term than is possible in a Phase III trial. For example, after a new drug for the treatment of menopausal hot flashes has been approved by the FDA, a phase IV trial might include women with less severe hot flashes than those included in the phase III trial.
Pilot studyA small study conducted to determine whether a full-scale study is feasible, as well as to optimize the logistics and maximize the efficiency of the full-scale study. For example, a pilot trial of restorative yoga for prevention of diabetes in patients with insulin resistance might aim to refine and standardize the yoga intervention and show that it is feasible to recruit and randomize the participants with insulin resistance to yoga and control groups.
Placebo controlAn inactive control that is indistinguishable from the active drug or intervention used in a randomized trial. For example, in a randomized, placebo-controlled trial of a new treatment for incontinence, the placebo should look, smell, taste and feel the same as the new medication that is being tested.
PlagiarismA type of scientific misconduct in which an investigator appropriates another person's ideas, results, or words without giving appropriate credit. For example, using another investigator's description of a new measurement method without appropriate attribution constitutes plagiarism.
Polychotomous categorical variablesCategorical variables with 3 or more categories. For example blood group, which includes A, B and O, is a polychotomous categorical variable.
PopulationA complete set of people with specified characteristics. For example, the adult population of the United States with type II diabetes could be defined as all U.S. adults who are taking a glucose-lowering medication or who have a fasting blood sugar level above 125 mg/dL.
Population-based sampleA sample of people who represent an entire population. For example, the National Health and Nutrition Examination Survey (NHANES), which provides data on a random sample of the entire population of the United States, is a population-based sample.
Post hoc hypothesesHypotheses that are formulated after data have been analyzed. For example, in a study of the association between insomnia and the risk of stroke, the hypothesis that insomnia increases the risk of diverticulitis is a post hoc hypothesis.
PowerThe probability of correctly rejecting the null hypothesis in a sample if the actual effect in the population is equal to or greater than a specified effect size. For example, suppose that exercise leads to an average reduction of 20 mg/dL in fasting glucose among diabetic women in the entire population. If an investigator set power at 90% and drew a sample from the population on numerous occasions, each time carrying out the same study with the same measurements, then in 9 of every 10 studies the investigator would correctly reject the null hypothesis and conclude that exercise reduces fasting glucose levels. See also beta.
Practice-based research networksNetworks in which physicians from community settings work together to study research questions of interest. For example, a study from a practice-based research network of treatments for carpal tunnel syndrome in primary care practice showed that most patients improved with conservative therapy. This contrasted with previous literature from academic medical centers that indicated that the majority of patients with carpal tunnel syndrome required surgery.
PrecisionThe degree to which measurement of a variable is reproducible, with nearly the same value each time it is measured. For example, a beam scale can measure body weight with great precision, whereas an interview to measure severity of depression is more likely to produce values that vary from one observer to the next.
Preclinical trialStudies that occur before an intervention is tested in humans. Such trials might include cells, tissues, or animals. For example, the U.S. Food and Drug Administration requires preclinical trials in two different animal species to document safety before new drugs can be tested in humans.
Predictive validityA term that describes how well a measurement represents the underlying phenomenon it is intended to measure, based upon its ability to predict related outcomes. For example, the predictive validity of a measurement of depression would be strengthened if it was associated with the subsequent risk of suicide.
Predictor variableIn considering the association between two variables, the one that occurs first or is more likely on biologic grounds to cause the other. For example, in a study to determine if obesity is associated with an increased risk of sleep apnea, obesity would be the predictor variable. In a randomized trial analyzed by intention to treat, the predictor variable is group assignment.
PretestAn evaluation of specific questionnaires, measures, or procedures that can be carried out by study staff before a study starts to assess functionality, appropriateness or feasibility. For example, pretesting the data entry and database management system could be done by having study staff complete forms with missing, out of range, and illogical data to ensure that the data editing system identifies these errors.
PrevalenceThe proportion of persons who have a disease or condition at one point in time. Prevalence is affected by both the incidence of a disease and duration of the disease. For example, the prevalence of systemic lupus erythematosis is the proportion of people who have this condition at a specific point in time; it might increase if the disease becomes more common or if treatment improves such that persons with the disease live longer.
Principal investigatorThe person who has ultimate responsibility for the design and conduct of a study, and the analysis and presentation the study findings. For example, the institutional review board asked to speak with the study?s principal investigator because some members had questions about the protocol.
Probability sampleA random process, usually using a table of random numbers or a computer algorithm, to guarantee that each member of a population has a specified chance of being included in the sample, thereby providing a rigorous basis for making inferences from the sample to the population. For example, a observations from a probability sample of 5% of persons with chronic obstructive pulmonary disease (COPD) based on hospital discharge diagnoses from all hospitals in California should provide reliable findings about risk factors for rehospitalization and death.
Propensity scoreThe estimated probability that a study participant will have a specified value of a predictor variable, most often the probability of receiving a particular treatment. Controlling for the propensity score (e.g., by matching, stratification, or multivariable analysis) is one method for dealing with confounding by indication: Instead of adjusting for all factors that might be associated with the outcome, the investigator creates a multivariate model to predict receipt of the treatment. Each subject is then assigned a predicted probability of treatment (the propensity score), which can then be used as the only confounder when estimating the association between the treatment and the outcome. For example, the investigators used a propensity score to adjust for the factors associated with the use of aspirin to determine the association between aspirin use and colon cancer.
ProposalA document that includes a study protocol, budget and other administrative and supporting information that is written for the purpose of obtaining funding from a granting agency. For example, the National Institutes of Health (NIH) requires proposals for funding of multiple types of research.
Prospective cohort studyA study design in which a defined group (the cohort) of study participants has baseline values of predictor variables measured and then is followed over time for specific outcomes. For example, the Nurses Health Study is a prospective cohort study of risk factors for common diseases in women. The cohort is a sample of registered nurses in the United States and the outcomes have included cardiovascular diseases, cancer, and mortality.
Protected health informationIndividually identifiable health information. Federal health privacy regulations (called HIPAA regulation after the Health Insurance Portability and Accountability Act), require researchers to maintain the confidentiality of protected health information in research. For example, protected health information should not be stored on flash drives or sent via regular e-mail.
ProtocolThe detailed written plan for a study. For example, the study protocol specified that only subjects who could understand English at the 8th grade level were eligible for participation.
Publication biasA distortion of the published literature that occurs when published studies are not representative of all studies that have been done, usually because positive results are submitted and published more often than negative results. For example, publication bias was suspected by authors of a meta-analysis that found that 6 small positive studies, but only 1 large negative study, had been published.

Glossary material from Hulley SB et al. Designing Clinical Research, 4th ed. Philadelphia, Lippincott Williams & Wilkins, 2013.