| Term | Explanation |
| Absolute risk reduction (ARR) | The difference between the rates in the control and the treatment group. |
| Baye’s revision | A formula for revising belief in the light of new information |
| Blinded | Single blinded – the patient doesn’t know if on trial therapy or placebo; double blinded – the doctor also doesn’t know. |
| Cohort study | A study design where patients are followed over time without an intervention and control group – observational. |
| Confidence interval (CI) | Two figures delineating the range within which the true value is likely to fall 95% of the time if you repeated the trial many times |
| Control event rate (CER) | The rate of events in the group given placebo. |
| Diagnostic probability | Your degree of certainty that a diagnosis is correct. |
| Diagnostic threshold | The level of certainty below which a diagnosis is so unlikely that investigating further is unwarranted. |
| Experimental event rate (EER) | The rate of events in the group given the active treatment |
| Hazard ratio | The ratio of two incidence rates. Not intuitively easy to grasp, so look for more clinically meaningful comparators. |
| Likelihood ratio negative (LR-) | A small number (e.g. < 0.3) indicates a useful test if the result is negative – i.e. will be helpful at excluding a diagnosis. |
| Likelihood ratio positive (LR+) | A large number (e.g. > 3) indicates a useful test if a result is positive (i.e. will make the diagnosis much more likely). |
| Meta-analysis | A mathematical process for pooling results from a systematic review. |
| Negative predictive value (NPV) | Probability that a disease is absent if the test is negative. |
| Number needed to harm (NNH) | Number you will have treated before one patient derives harm from the treatment. |
| Number needed to investigate (NNI) | The number of patients you need to test to detect one case |
| Number needed to treat (NNT) | The number of patients one needs to treat in order to secure benefit for one individual. |
| Observational study | A study design where you watch without intervening. |
| Odds | A ratio expressed in a form more easily used in Bayesian revision. (Events occurring/events not occurring.) |
| Odds ratio (OR) | A ratio of two odds ratios. Commonly used in meta-analysis outputs. |
| Positive predictive value (PPV) | Probability that a disease is present if the test is positive. |
| Posterior probability | Your degree of certainty after revising knowledge in the light of new information. |
| Prior probability | Your degree of certainty before revising knowledge with new information. |
| Randomised controlled trial (RCT) | An experimental design incorporating a group given the trial therapy compared with a group given either placebo or standard care. |
| Rate | A proportion (e.g. events occurring /all possible events) |
| Relative risk reduction (RRR) | Efficacy measured as a proportional rather than an absolute difference. |
| Sensitivity (Sens) | Probability that test is positive if the disease is present. |
| Significance level | A measure of the degree of certainty about trial result. |
| Specificity (Spec) | Probability that the test is negative if the disease is absent. |
| Systematic review | A formal overview of literature on a specific topic, done according to strict rules. |
| Treatment threshold | The level of diagnostic certainty above which treatment is appropriate without further testing. |
