Skip to main content
Log in

Statistical methods of estimating fracture risk

Statistische Methoden zur Berechnung des Frakturrisikos

  • Themenschwerpunkt
  • Published:
Wiener Medizinische Wochenschrift Aims and scope Submit manuscript

Zusammenfassung

Das Ziel dieser Arbeit ist die Analyse von statistischen Methoden, welche zur Bewertung des Frakturrisikos bei osteoporotischen Patienten zur Anwendung kommen. Die mathematischen Relationen von verschiedenen Methoden werden erklärt. (Risk – R, Risk ratio – RR, Risk difference – RD, odds – O, odds ratio – OR, Yule's Q, Yule's Y, logistic model). Für die Interpretation von statistischen Daten ist es wichtig zu beachten: 1) Relatives Risiko und Odds ratio beschreiben lediglich eine Assoziation und keine Kausalität. 2) Relatives Risiko und Odds ratio beziehen sich auf eine Population aber nicht auf einen individuellen Patienten. 3) Studien mit geringerer Fallzahl finden mit höherer Wahrscheinlichkeit eine zufällige Verbindung zwischen dem Risikofaktor und dem Outcome (Fraktur) als Studien mit größerer Fallzahl. 4) Wenn die Inzidenz eines "Outcome of interest" in der Studienpopulation niedrig ist (<10 %), und die OR nahe beim RR liegt, wird je häufiger der Outcome auftritt, die OR das RR überschätzen, wenn sie größer als 1 ist (bzw. unterschätzen, wenn sie kleiner als 1 ist). Anspruchsvolle statistische Pakete, welche viele der statistischen Tests berechnen können, sind verfügbar. Das Problem ist jedoch, dass der Forscher wissen muss welcher der richtige Test ist. Die falsche Wahl der statistischen Analyse, die falsche Interpretation der Risk ratio oder der Odds ratio und die Überbewertung eines Risikofaktors kann zu unbeabsichtigten Fehlbewertungen bei der ökonomischen Analyse von potentiellen Interventionen bei Osteoporose führen. Dieser Artikel könnte einen Beitrag für Forscher sein, die sich mit der Bewertung des Frakturrisikos beschäftigen.

Summary

The aim of the article is to present an analysis of statistical methods used for estimating fracture risk in patients with osteoporosis. Mathematical relations of different methods are explained (risk – R, risk ratio – RR, RD – risk difference, odds – O, odds ratio – OR, Yule's Q, Yule's Y, logistic model). What is important to keep in mind is that: 1) relative risk and odds ratio are statistics that only describe an association, not causation; 2) relative risk and odds ratio refer to a population, not to an individual patient; 3) the studies of small groups are more likely to find an association that might actually just be due to chance, the larger the groups, the less likely the association between a risk factor and an outcome (fracture); 4) when the incidence of an outcome of interest in the study population is low (<10 %), the OR is close to the RR, the more frequent the outcome becomes, the more the OR will overestimate the RR when it is more than 1 or underestimate the RR when it is less than 1. Sophisticated statistical packages are available which can calculate many of the tests of association but the problem is that the investigator must know which is the desirable one. The incorrect option of statistical analysis, the incorrect interpretation of risk ratio or odds ratio and overestimation of the importance of a risk factor may lead to unintentional errors in the economic analysis of potential programs or treatments in osteoporosis. This article could be a contribution for investigators who are concerned with assessment of fracture risk.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

References

  • Kanis JA, Johnell O, Oden A, Jonsson B, Dawson A, Dere W (2000) Risk of hip fracture derived from relative risk: An analysis applied to the population of Sweden. Osteoporosis Int 11: 120–127

    Article  CAS  Google Scholar 

  • Kanis JA, Johnell O, Oden A, Oglesby AK, De Laet C, Jonsson B (2001) Ten year probabilities of osteoporotic fracture according to BMD and diagnostic thresholds. J Bone Miner Res 16(Suppl 1): 194

    Google Scholar 

  • Mc Nutt LA, Wu CH, Xue X, Hafner JP (2003) Estimating the relative risk in cohort studies and clinical trials of common outcomes. Am J Epidemiol 157: 940–943

    Article  Google Scholar 

  • Sinclair JC, Bracken MB (1994) Clinically useful measures of treatment effect in binary analyses of randomized trials. J Clin Epidemiol 47: 881–889

    Article  CAS  PubMed  Google Scholar 

  • Zhang J, Yu KF (1998) What is relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 280: 1690–1691

    Article  CAS  PubMed  Google Scholar 

  • Greenland S (2004) Model – based estimation of relative risks and other epidemiologic measures in studies of common outcomes and in case – control studies. Am J Epidemiol 160 (4): 301–305

    Article  PubMed  Google Scholar 

  • Melton JL III, Atkinson E, O'Connor MK, O'Fallon WM, Riggs BL (1998) Bone density and fracture risk in men. J Bone Miner Res 13: 1915–1917

    Article  PubMed  Google Scholar 

  • Hilgers RD, Bauer P, Scheber V (2003) Einführung in die medizinische Statistik Springer Verlag Berlin Heidelberg, 1. Ausgabe, S217–S220

  • Kleinbaum DG, Klein M (2005) Logistic regression. Statistics for Biology and Health, Springer, 2nd edition 2002, corr. 4th printing 2005, S2–S45

  • Lee J (1994) Odds ratio or relative risk for cross – sectional data? Int J Epidemiol 23: 201–203

    CAS  PubMed  Google Scholar 

  • Bland JM, Altman DG (2000) The odss ratio. Brit Med J 320: 1468

    Article  CAS  PubMed  Google Scholar 

  • Davies HTO, Crombie IK, Tavakoli M (1998) When can odds ratios mislead? BMJ 316: 989–991

    CAS  PubMed  Google Scholar 

  • Last JM (2001) A dictionary of epidemiology, 4th ed, RA Spasoff, SS Harris and MC Thuriaux (eds). Oxford University Press, Oxford, pp 45–82

  • Jaeschke, Guyatt G, Shannon H, Walter S, Cook D, Heddle N (1995) Assessing the effects of treatment: measures of association. CMAJ 152: 351–357

    CAS  PubMed  Google Scholar 

  • Mikulecky M (2003) Association of nominal variables – contingency tables. Euro Rehab 9: 88–96

    Google Scholar 

  • Egger M, Davey Smith G, Philips AN (1997) Meta analyses principles and procedures. BMJ 315: 1533–1537

    CAS  PubMed  Google Scholar 

  • Black DM, Steinbuch L, Palermo L, Dargent–Molina P, Lindsay R, Hoseyni MS, Johnell O (2001) An assessment tool for predicting fracture risk in postmenopausal women. Osteoporosis Int 12: 519–528

    Article  CAS  Google Scholar 

  • Widrow B, Smith FW (1964) Pattern recognizing control systems. In: Tou JT, Wilcox RH, Computer and Information science. Collected papers on learning adaptation and control in information systems, 2nd ed. Sparten Books, Washington, pp 40–47

    Google Scholar 

  • Mikulecky M, Komornik J, Ondrejka P (1997) Automatizacia matematicko – statistickych odhadov a testov na zaklade binomickeho rozdelenia. PC software, Comtel

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Jaroslava Wendlová.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Wendlová, J. Statistical methods of estimating fracture risk. Wien Med Wochenschr 156, 569–573 (2006). https://doi.org/10.1007/s10354-006-0278-5

Download citation

  • Received:

  • Accepted:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10354-006-0278-5

Schlüsselwörter

Keywords

Navigation