• RelativE risK: Compares how many times as high the disease risk is for the exposed group compared to the non-exposed group. For instance, in patients who had uncemented prosthesis, the risk of aseptic loosening increases three times with exposure to NSAID compared to patients who were not exposed to NSAID.

• THE Odds ratio: Estimate of the relative risk if the disease is not frequent.

• IncidenCE: Concerns the frequency of new cases. Can be expressed as number of new disease cases accumulated over time (cumulated incidence proportion), or as the >velocity< at which new cases arise (incidence rate).

• PrEvalenCE: Concerns the population of diseased at a given time. The prevalence proportion is the proportion or part of the study population which has the disease in question at a given time.

• Mortality rate: Is estimated as the number of deaths in relation to the time of risk.

• Confounding: An alternative cause of disease which is unevenly distributed between exposed and non-exposed persons. Three conditions must be met for a factor to be a confounder. The factor must be an independent risk factor for the development of the disease (1), there must be a statistical association between the incidence of the exposure and the confounder (2), and finally, the confounder cannot be a part of the chain of causes between exposure and effect (3). Confounding can be avoided / reduced by restriction, matching or stratification.

• Confidence interval (CI): An expression of the statistical precision of the observed measure of association, e.g. the Rate or Hazard ratio; 95 % Confidence Interval means that if we repeat the data collection and analyses many times, the 95 % Confidence interval will include the correct value of measurements in 95 % of the cases. Confidence Intervals indicate to which extent random variation can explain the registered survival and is closely connected with the number of operations being part of the analysis. A wide confidence interval indicates that there is a considerable uncertainty about the real prosthesis survival, while, on the contrary, to a lesser extent, a narrow interval indicates that the prosthesis survival can be interpreted as a result of random variation. If the value corresponding with no effect (as the Relative risk or Hazard ratio 1, or treatment difference at 0) fall outside the 95 % Confidence Interval, the result will be statistically significant at 0.05 level. If the Confidence Interval includes 1 or 0 the result is not statistically significant.

• P - value: Is calculated in relation to a specific hypothesis, usually the Nil Hypothesis which states that there is no connection between exposure and disease; e.g. Relative risk=1. The P-value is to regard as a measure of the relative accordance between the Nil Hypothesis and the collected data.

• Survival analyses:

A. Outcome variable is the time until the event occurs.

B. We are interested in one event; more events cause competing risk problems.

C. The purpose of the analysis is to estimate and interpret survival and/or mortality; to compare survival and/or mortality; to assess the relation between explanatory variables and the survival time.

• survival curves: Curves or plots which present the proportion of patients who have not experienced the defined event (e.g. death, revision of prosthesis) in relation to the time. The Kaplan-Meier method is the most used to present survival curves.

• Log Rank Test: A statistical test which compares the survival of two or more groups during the entire follow-up period (as opposed to comparison of the survival within a determined time period, e.g. five years survival).

• Cox proportional hazard model: A statistical model which is used to analyze survival data. The model compares two or more different categories (e.g. three types of prosthesis) calculating the Hazard Ratios (can be interpreted as measure of the relative risk) with 95 % CI.

• Hazard ratio: Expresses the effect of each variable included in the Cox model in relation to the reference group, adjusted for other variables in the model.

If we compare the survival for patients in three prosthesis groups with revision as outcome;

Thus, Hazard Ratios are a comparison of the incidence of revision in two different categories of patients. If the Hazard Ratio is 1.00 there is no difference in the incidence of revision when the two patient categories are compared. On the other hand, a Hazard Ratio <1 will indicate that the incidence of revision in a given patient category is lower than the incidence in the reference category.

In case the stated 95 % CI for Hazard Ratio do not include 1.00 it can be concluded that the given category of patients have an incidence of revision which differs from the reference category and that this difference probably cannot be explained by random variation. In other words, there exist stastistically significant differences. On the other hand, if the 95 % CI include 1.00 it is not possible to determine whether the incidence is different in the two categories.

Example: In an analysis of all patients with a primary total hip replacement with first revision as end point, the Hazard Ratio was 0.49 (95 % CI:0,35-0,69) when we compared patients older than 74 years with patients younger than 50 years. Thus, the incidence of the first revision was relatively 51 % lower among patients older than 74 years compared to patients younger than 50 years. The relatively narrow CI and the fact that 1.00 is not included indicate that this difference between the two patient categories is established very precisely and probably cannot be ascribed to random variation.

• Validity: Or the accuracy. Validity refers to the degree of what the data measure compared to what they are supposed to measure.

• Evaluation of registry data: There are two questions to be answered:

1. Completeness of patient registration – Defined as a part of all patients who had the operation in question and who are in fact registered in the Registry.

2. The validity of registered data – Defined as a percentage of persons in the Registry with a given characteristic (e.g. age, sex, type of operation, diagnosis) who really have these characteristics. The validation of registered data can be expressed as the predictive value of positive registration.

• Sensitivity: An expression of the test’s ability to classify the diseased correctly; for instance sensitivity of postoperative questionnaire B, which is used for the registration of postoperative complications, is 0.9. That is, 90 % of the patients who had postoperative complications are registered with the complication in question in the Registry and, conversely, 10 % of the patients who had postoperative complications are not registered with the complication in the Registry.

• Specificity: An expression of the test’s ability to classify the non-diseased correctly; for instance specificity of postoperative questionnaire B, which is used for the registration of postoperative complications, is 0.3. That is, 30 % of the patients who did not have postoperative complications are registered as patients without complications in the Registry and, conversely, 70 % of the patients who did not have postoperative complications are registered with complications in the Registry.

• positive Predictive value: The probability that the patient really is diseased if the test is positive. For instance the positive predictive value of the diagnosis primary arthrosis in the Registry is 0.85. That is, 85 % of the patients registered with primary arthrosis did really have this diagnosis and 15 % of the patients with primary arthrosis should be registered with another diagnosis.

• Co-morbidity: Co-existing diseases can have an influence on survival. For instance survival of hip prosthesis in relation to first revision is 51 % lower for a patient who is older than 74 years than for a patient of the same sex who is younger than 50 years. After adjusting for co-morbidity in the analysis of survival, the survival of hip prosthesis is the same for the two patient groups. Hereby, we can conclude that higher survival before adjusting for co-morbidity is mostly due to higher morbidity for patients older than 74 years.

• The number needed to be treated (NNT): Indicates how many patients have to be in treatment in order to avoid an unfavourable incident. For instance, 100 patients must be in treatment with lipid lowering medicine for five years in order to avoid a case of AMI (made up).

• Case-Control studY: Analytical, epidemiological study based on a group of patients suffering from a defined disease, as it is studied whether there exist conditions in the diseaseds’ past which differentiate them from non-diseased. Patients with the disease (*cases*) are compared to persons without the disease (*controls / reference*). Measure of association – the Odds ratio (OR) or Relative risk (RR).

• Follow-up studY: Descriptive or analytical, epidemiological study which is based on a selected group of persons who are more or less exposed to a presumably pathogenic or prognostically significant factor. The selected persons are followed over time until they contract disease or until the study period ends in order to study whether the exposed have higher morbidity or mortality than the non-exposed. Measure of association – Relative (Incidence ratio, Relative risk, the Odds ratio) and Absolute (Risk difference, Incidence rate difference).

• RandomiZED Clinical study: An analytical study in which a group of patients receive one type of treatment and another group of patients receive a different type of treatment or no treatment. The selection between two types of treatment takes place by lot (randomization). The randomization procedure should ensure that the studied groups have the same expected prognosis and disease manifestation at group level.

• Selection bias:  Systematic errors due to errors in the procedures used in the selection of patients for the study or due to factors which can have an influence on selected patients. Bias arises when the connection between exposure and disease is different for persons who participate in the study and persons who do not participate in the study.

• Information bias: Systematic errors which occur because the information gathered about or from the participants in the study is faulty. This is called misclassified information.

• Recall bias: The most frequent type of information bias which occurs in case-control studies because the patients are interviewed about exposure information after the disease setting in. Persons who contracted a disease will remember better than persons who did not contract the disease.

• Publication bias: Is reflected in that journals have a tendency to accept manuscripts which report positive results. This shows in meta-analyses based on published data alone. Bias in meta-analyses corresponds with the selection bias in individual studies.

Reference:

Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40(5):373-383.

Fletcher RH, Fletcher SW, Wagner EH. Clinical epidemiology, the essentials. Baltimore, USA: Lippincott Williams&Wilkins, 1996.

Last JM. A dictionary of epidemiology. New York: Oxford University Press, 1995.

Olsen J, Overvad K, Juul S. Analytisk epidemiologi, 2. udgave. København, Munksgaard, 1994.