Retrospective study

from Wikipedia, the free encyclopedia
Categories of patient-oriented studies

Retrospective study ( Latin retrospectare , 'to look back' ) is a term used in clinical research . It is a study from the main group of observational studies and the local subgroup of longitudinal studies . A study is referred to as retrospective if the data was collected before the start of the study. In contrast, in a prospective study, the data are collected again after the start of the study especially for this study.

The common goal of these two types of longitudinal studies is to describe a possible statistical relationship between certain influences - such as genetic deviations, pollutants, lifestyle habits, medication or medical treatments - on the one hand and certain changes in health that occur afterwards, such as illness, improvement , or recovery on the other hand.

In the course of the now widespread emergence of very extensive digitized medical databases , especially of ( anonymized ) patient data, the efficiency and therefore the importance of retrospective studies has increased significantly.

history

Retrospective studies have almost the same history as prospective ones. An early example is a study from 1933 using data from 132 families in a small town in Tennessee , USA, in which a relationship between pulmonary tuberculosis and family contacts was described as a function of age groups.

Main types of retrospective studies

In addition to the three main types of retrospective studies described below, there are also combinations of these, which in a specific case can usually be developed based on knowledge of the main types.

Case-control study

Case- control studies are carried out as follows. Test subjects are selected for whom the event to be examined, for example a certain illness or health disorder, is present (case group). In addition, a comparison group of test persons who did not have this event is selected (control group). It should be noted that these should correspond to those of the case group for important properties, such as age and gender. This is known as matching . The subjects are then examined and / or asked whether and to what extent they were exposed to possible causal factors. Statistical methods are then used to analyze whether these factors occurred more frequently or less often in the case group than in the control group. If there are statistically significant differences, it may be possible to develop important hypotheses about the causes of the disease or health disorder - for example with regard to the effects of certain environmental toxins or certain diet and lifestyle habits.

Retrospective cohort study

In contrast to the case-control study, the retrospective cohort study assumes one possible factor - such as a specific medical treatment. From the patient files of the examined group ( cohort ) it is then filtered out whether certain desired - or also undesired - changes (events) are recorded after one or more time periods. Such changes over time, provided they are statistically significant within the cohort, and also non-changes, are suitable for either confirming or invalidating basic hypotheses about the effects and side effects of treatments. They can also stimulate the development of new hypotheses.

Nested case-control study

In nested case-control studies, both the case group and the control group are subsets of what is usually a very large cohort. The common origin in a cohort already guarantees a rough similarity (comparability) of case and control groups. As a rule, this rough matching is further refined by exact correspondences according to age, gender, etc. It is also customary to use several groups for each person in the case group, e.g. B. to include four - similar (matched) people in the control group. The results of this third main type of retrospective study are considered to be particularly reliable and have grown in importance since the explosion of very large digitized medical databases.

Importance for genetic studies

Retrospective studies of the case-control study type are the basis of most studies on possible genetic influences in the development of diseases and health disorders, as well as on the interaction between genetic makeup and the environment (gene-environment interaction). The most important process here are genome-wide association studies and epidemiological description of copy number variations (copy number variations, CNV) in the human genome.

Advantages of retrospective studies

  • Suitable for rare occurrences: Very large databases are required to record and clarify rare, but dangerous, side effects of drugs or treatments, and retrospective studies are usually the best or even the only way to get results.
  • Suitable for harmful influences: Possible risks from environmental toxins or previously overlooked pollutants in food and luxury goods can only be recorded through retrospective studies, since targeted administration of such substances is not justifiable.
  • Uninfluenced data collection: As a rule, the data were collected and recorded without any subsequent additional use from a retrospective study being known. A direct or indirect influence on the data collection by goals of the later use can therefore be excluded in these cases.
  • Immediate and quick feasibility: By the time the study begins, all data to be used have already been collected and recorded.
  • Low costs: collection and recording of the data are usually billed at the beginning of the study as part of other (previous) projects or treatments. Subsequent cost sharing due to renewed use of the data is not common.
  • Involvement of ethics committees is omitted: The usual applications for authorization of the study eliminated by an ethics committee, as their responsibilities are not affected. The data to be used has already been recorded and it is only necessary to ensure that it is used anonymously .

Disadvantages of retrospective studies

  • Like all observational studies, retrospective studies can suggest possible causal relationships - in some cases imperatively - but cannot definitively prove them.
  • Compared to randomized controlled trials , observational studies usually correctly determine the direction of cause and effect, but often overestimate the magnitude of the treatment effects.
  • Unwanted side effects of treatments tend to be underestimated compared to randomized controlled trials.
  • Possible additional, disruptive factors ( confounders ) are often inadequately recorded in the evaluated data material or are completely absent.
  • Since one relies on the patient's memory and old records, such studies are prone to errors. Has the patient perhaps simply forgotten a causal event or mixed up the sequence? Patients tend to forget something that they do not have a causal connection with their illness or do not remember it very well ( recall bias ).
  • Compared to randomized controlled studies, the group of treated persons is not put together at random, which means that the results may be directed in a non-representative direction ( sample bias ).

A 2014 large-scale Cochrane survey compared the reliability of observational studies with that of randomized controlled trials. The differences were so insignificant that the authors recommended that when evaluating studies, the special circumstances of each individual study should be carefully considered instead of being based on generalized judgments about study types.

See also

literature

  • Richard Doll : Cohort studies: history of the method. II. Retrospective cohort studies. In: Social and preventive medicine. Volume 46, number 3, 2001, pp. 152-160, doi : 10.1007 / BF01299724 .
  • AM Euser, C. Zoccali, KJ Jager, FW Dekker: Cohort studies: prospective versus retrospective. In: Nephron. Clinical practice. Volume 113, number 3, 2009, pp. C214-c217, doi : 10.1159 / 000235241 (review).
  • JM Gamble: An introduction to the fundamentals of cohort and case-control studies. In: The Canadian journal of hospital pharmacy. Volume 67, number 5, 2014, pp. 366–372, doi : 10.4212 / cjhp.v67i5.1391 , PMC 4214579 (free full text).
  • Wilhelm Gaus; Rainer Muche: Medical Statistics. Applied biometrics for doctors and health professions , 2nd, revised edition, Schattauer Verlag, Stuttgart 2017, ISBN 978-3-7945-3241-4 .
  • DA Grimes, KF Schulz: Cohort studies: marching towards outcomes. In: The Lancet . Volume 359, number 9303, 2002, pp. 341-345, doi : 10.1016 / S0140-6736 (02) 07500-1 , PDF .
  • U. Held: What types of study designs are there and how are they used correctly? In: SwissMedical Forum , Volume 10, Number 41, pp. 712-714, doi : 10.4414 / smf.2010.07304 .
  • JA Rosenfeld, A. Patel: Chromosomal Microarrays: Understanding Genetics of Neurodevelopmental Disorders and Congenital Anomalies. In: Journal of pediatric genetics. Volume 6, number 1, March 2017, pp. 42–50, doi : 10.1055 / s-0036-1584306 , PMC 5288005 (free full text) (Review of the importance of large retrospective case-control studies in investigating the spread and effects of Copy number variations (CNV) in the human genome)

Web links

Individual evidence

  1. ↑ CM Seiler: Patient-Oriented Research in Surgery . In: Manfred Georg Krukemeyer, Hans-Ullrich Spiegel (Hrsg.): Surgical research . Georg Thieme Verlag, Stuttgart 2005, ISBN 978-3-13-133661-3 , p. 205–212 ( limited preview in Google Book search).
  2. ^ WH Frost: Risk of Persons in Familial Contact with Pulmonary Tuberculosis. In: American Journal of Public Health and the nation's health. Volume 23, number 5, 1933, pp. 426-432, doi : 10.2105 / AJPH.23.5.426 , PMC 1558187 (free full text).
  3. ^ Richard Doll : Cohort studies: history of the method. II. Retrospective cohort studies. In: Social and preventive medicine. Volume 46, number 3, 2001, pp. 152-160, doi : 10.1007 / BF01299724 .
  4. ↑ a b c d e D. I. Sessler, PB Imrey: Clinical Research Methodology 1: Study Designs and Methodologic Sources of Error. In: Anesthesia and analgesia. Volume 121, number 4, October 2015, pp. 1034-1042, doi : 10.1213 / ANE.0000000000000815 (review), PDF .
  5. ^ A b c d e D. I. Sessler, PB Imrey: Clinical Research Methodology 2: Observational Clinical Research. In: Anesthesia and analgesia. Volume 121, number 4, 2015, pp. 1043-1051, doi : 10.1213 / ANE.0000000000000861 (review), PDF .
  6. ^ YY Teo: Common statistical issues in genome-wide association studies: a review on power, data quality control, genotype calling and population structure. In: Current opinion in lipidology. Volume 19, Number 2, 2008, pp. 133-143, doi : 10.1097 / MOL.0b013e3282f5dd77 (Review), PDF .
  7. ↑ ER Riggs, DM Church, K. Hanson, VL Horner, EB Kaminsky, RM Kuhn, KE Wain, ES Williams, S. Aradhya, HM Kearney, DH Ledbetter, ST South, EC Thorland, CL Martin: Towards an evidence-based process for the clinical interpretation of copy number variation. In: Clinical genetics. Volume 81, number 5, 2012, pp. 403-412, doi : 10.1111 / j.1399-0004.2011.01818.x , PMC 5008023 (free full text) (review).
  8. ^ DA Grimes, KF Schulz: Cohort studies: marching towards outcomes. In: The Lancet . Volume 359, number 9303, 2002, pp. 341-345, doi : 10.1016 / S0140-6736 (02) 07500-1 , PDF .
  9. ^ Gaus, Wilhelm; Muche, Rainer: Medical Statistics. Applied biometrics for doctors and health professions , 2nd, revised edition, Schattauer Verlag, Stuttgart 2017, ISBN 978-3-7945-3241-4 , p. 39.
  10. ^ A. Anglemyer, HT Horvath, L. Bero: Healthcare outcomes assessed with observational study designs compared with those assessed in randomized trials. In: The Cochrane database of systematic reviews. Number 4, 2014, MR000034, doi : 10.1002 / 14651858.MR000034.pub2 (Review).