Study Design and Publication Type

The following is a brief description of the type of study design or publication types that are included in the Plant-Based Research database. Each one has strengths and limitations; a theory is best supported by multiple types of evidence. Click here to download a PDF of this chart.

Study Design Description Strengths Limitations
Randomized Controlled Trial (Parallel) Intervention study in with two groups, one of which receives the intervention and the other of which receives no intervention, or something very minimal for comparison RCTs are typically considered the "gold standard" of research and provide the strongest evidence of causality. Parallel designs are a good choice when there is plenty of money to recruit enough subjects to get the desired statistical power. Addresses the question of temporality (causes precedes effect), a requirement for causality Can be expensive, difficult to randomize diets over the long term and results may not apply to long-term outcomes
Randomized Controlled Trial (Crossover) Intervention study in which two groups, or each individual, receives both the intervention and the control experience, preferably in random order RCTs are typically considered the "gold standard" of research and provide the strongest evidence of causality. Crossover designs allow investigators to use a smaller number of subjects and test whether the results hold up after exposure to a different experience. Addresses the question of temporality (causes precedes effect), a requirement for causality Can be expensive, results may not apply to long-term outcomes
Intervention Trial (Single Arm) Intervention study with only one group - all subjects receive the intervention May be more affordable without the control group Lacks control or comparison group, so evidence is weaker
Acute Feeding Study Responses are measured following a test meal or food consumed, usually performed on fasting subjects Enables study of specific mechanisms and immediate responses to food, precise time measurements May not apply to long-term outcomes
Cohort Compares disease outcomes in more than one group of humans with different exposures - often cohort studies are observational. Risk ratios can be calculated from cohort studies. Comparing groups of humans is ideal for generalizing results to humans; observational cohort studies allow the study of long-term outcomes. Large sample sizes provide greater statistical power. More affordable way to study large groups of people over the long-term than RCTs Can be expensive to recruit; loss to follow-up may attenuate results
Case-Control Compares exposure outcomes between cases (those who have the disease) and controls (non-disease whoe were selected to be matched to cases). Odds ratios can be calculated from case-control studies. Much less expensive than cohort studies, no loss to follow-up, when the disease is rare the odds ratio approximates the risk ratio When the disease is not rare the odds ratio will overestimate the risk ratio
Nested Case-Control A case-control study in which both cases and controls are selected from a previously designated cohort. Selecting cases and controls from a defined cohort minimizes selection bias and hopefully the exposure risk in the study approximates the true exposure risk in the source population When the disease is not rare the odds ratio will overestimate the risk ratio
Observational Observes disease outcomes over time without intervening in any way. Observational studies are often cohort studies. Allows long-term study of disease outcomes over the lifetime - it is usually not possible to randomize people to overall dietary patterns for more than a few months because of adherence issues and drop outs. Reproducible results from observational, cohort studies also provide strong evidence for causality. Confounding factors that are not accounted for in the analysis may result in biased results.
Follow-Up Study No further intervention has been administered, but subjects' outcomes are measured at a defined period after the intervention has ended (often six months, a year, or several years) Useful for determing long-term outcomes following an intervention held over a discrete time period Other intervening factors may be responsible for outcomes since the end of the intervention. Loss to follow-up causes loss of statistical power in analysis
Ecologic Study Assesses correlations between conditions and outcomes in a geographic area - for instance, the correlations between death rates for heart disease by country and average consumption of total dietary fat by country Useful for exploratory research to generate hypothesis about possible casual factors for disease, may lead to observational cohort studies and RCTs Ecologic fallacy is attributing the the disease outcomes (death from CHD) to the individuals who have the highest rates of fat consumption, when these may in fact not be the same individuals
Meta-Analysis Researchers gather data from multiple studies that meet predefined inclusion criteria and analyze data as one dataset Allows for examination of overall trends and reproducibility of results Many statistical issues because often studies are not similar enough to each other to be equivalent
Review Summary of current literature on a particular research question or topic with thesis/commentary by author, using predefined inclusion criteria to include or exclude studies. A narrative review is a review advancing a particular viewpoint on the part of the author (this website is loosely a  form of narrative review). GIves a summary of current views in the literature Not an original experiment or research study
Systematic Review Summary of current literature on a particular research question or topic, using predefined inclusion criteria but also following a specific protocol accepted for systematic reviews that involves multiple reviews and a strict inclusion process. Generally seen as more rigorous than reviews Not an original experiment or research study
Prospective Refers to the timeframe over which the study was conducted: hypothesis is generated and exposure is measured before data are collected Prospective data collection addresses the question of temporality (causes precedes effect), a requirement for causality Prospective studies, especially if conducted over years, can have significant loss to follow-up
Retrospective Refers to the timeframe over which the study was conducted: hypothesis is generated after data are collected, and Looking back into datasets that have already been collected allow for investigation of important questions that were not necessarily part of the original design but are still worth asking now, also allows more thorough use of resources as data has already been gathered Often when data is collected to answer different research questions than the one being asked there are usually flaws in how well the data addresses the hypothesis. Retrospective studies cannot address the question of temporality (cause precedes effect), a requirement for causality
Cross-Sectional A survey that measures a snapshot in time - could be prevalence of disease, death rates, or exposure proportions Less expensive than RCTs or cohort studies, excellent for generating hypotheses and getting confirming data Cannot address the question of temporality (cause precedes effect), a requirement for causality
In Vitro Any research activity that does not involve humans or animals Allows study of specific mechanisms, less expensive than studying humans or animals, no ethical issues Results may not generalize to animals or people
Animal Model Experiments using animals as subjects, often rats or mice Many biological processes are common to all or many organisms. Allows measurment and experimentation over the entire life-cycle, as well as on questions that will not be asked in a human population Ethical issues, results may not generalize to humans
Life Cycle Assessment (LCAs) or Environmental Analysis LCAs are complex calculations that model the energy requirements of various foods or food systeml - because of the limited number of environmental papers, all related papers have been combined with this category Allows modeling of many potentially important variables that influence energy use and environment impact Modeling may not always reflect real-life outcomes
Comments on Published Studies Other researchers may write to journals with responses or comments on other papers that have been previously published May point out important flaws in thinking or alternative conclusions Tend to be more opinion-driven