

STATISTICS 

Year : 2015  Volume
: 6
 Issue : 4  Page : 222224 

Common pitfalls in statistical analysis: Odds versus risk
Priya Ranganathan^{1}, Rakesh Aggarwal^{2}, CS Pramesh^{3}
^{1} Department of Anaesthesiology, Tata Memorial Centre, Parel, Mumbai, Maharashtra, India ^{2} Department of Gastroenterology, Sanjay Gandhi Postgraduate Institute of Medical Sciences, Lucknow, India ^{3} Department of Surgical Oncology, Division of Thoracic Surgery, Tata Memorial Centre, Parel, Mumbai, Maharashtra, India
Date of Web Publication  12Oct2015 
Correspondence Address: Priya Ranganathan Department of Anaesthesiology, Tata Memorial Centre, Ernest Borges Road, Parel, Mumbai  400 012, Maharashtra India
Source of Support: None, Conflict of Interest: None  Check 
DOI: 10.4103/22293485.167092
Abstract   
In biomedical research, we are often interested in quantifying the relationship between an exposure and an outcome. “Odds” and “Risk” are the most common terms which are used as measures of association between variables. In this article, which is the fourth in the series of common pitfalls in statistical analysis, we explain the meaning of risk and odds and the difference between the two.
Keywords: Biostatistics, odds ratio, risk
How to cite this article: Ranganathan P, Aggarwal R, Pramesh C S. Common pitfalls in statistical analysis: Odds versus risk. Perspect Clin Res 2015;6:2224 
Introduction   
Researchers are often interested in evaluating the association between an exposure and an outcome. In other words, they are interested in knowing whether the presence of a risk factor or performing an intervention alters the risk of an outcome as compared to the absence of the risk factor or of the intervention (the “control” situation).
Risk and Odds: Definitions   
“Risk” refers to the probability of occurrence of an event or outcome. Statistically, risk = chance of the outcome of interest/all possible outcomes. The term “odds” is often used instead of risk. “Odds” refers to the probability of occurrence of an event/probability of the event not occurring. At first glance, though these two concepts seem similar and interchangeable, there are important differences that dictate where the use of either of these is appropriate.
Let us look at the hypothetical example of a randomized trial comparing endoscopic sclerotherapy (n = 65) versus band ligation (n = 64) for the treatment of bleeding esophageal varices [Table 1]. The overall risk of death = 47/129 ([number of deaths]/[all outcomes i.e., all deaths + survivors]) =0.36. The overall odds of death = 47/82 ([number of deaths]/[number of nondeaths, i.e., survivors]) =0.57. The risk of death in the ligation group was 18/64 (28% or 0.28), and the risk of death in the sclerotherapy group was 29/65 (44% or 0.44). By contrast, the odds of death in the two groups was 18/46 (0.39) and 29/36 (0.81), respectively.  Table 1: A randomized trial of sclerotherapy versus ligation for esophageal varices (hypothetical data)
Click here to view 
Relationship of Risk and Odds   
In the example above, for the same data set, the chances of death appear markedly different when expressed as risks and odds.
[Table 2] shows the risk and odds for different event rates. As “a” decreases with respect to “b” (probability of outcome becomes less), the odds and risk are similar. For rare events (i.e., if “a” is small and “a + b” approaches “b”), a/(a + b) ≈ a/b and risk approximates odds. Therefore, though “odds” does not represent true risk, its value is close to risk when the event rates are low (typically <10%).^{[1]}
Relative Risk and Odds Ratio   
The relative risk (also known as risk ratio [RR]) is the ratio of risk of an event in one group (e.g., exposed group) versus the risk of the event in the other group (e.g., nonexposed group). The odds ratio (OR) is the ratio of odds of an event in one group versus the odds of the event in the other group.
An RR (or OR) of 1.0 indicates that there is no difference in risk (or odds) between the groups being compared. An RR (or OR) more than 1.0 indicates an increase in risk (or odds) among the exposed compared to the unexposed, whereas a RR (or OR) <1.0 indicates a decrease in risk (or odds) in the exposed group. As for other summary statistics, confidence intervals can be calculated for RR and OR.
In the same example, the RR of death in ligation group versus sclerotherapy group = 0.28/0.44 = 0.63. This means that the risk of death after ligation is 63% of the risk of death after sclerotherapy. This implies that ligation decreases the risk of death by 37% (calculated as 100 minus 63%) as compared with sclerotherapy. The OR for death in ligation group versus sclerotherapy group = 39/81 = 0.48. This means that the odds of death after ligation is 48% of the odds of death after sclerotherapy, or that ligation decreases the odds of death by 52% as compared with sclerotherapy.
Relationship of Risk Ratio and Odds Ratio   
Though OR also indicates the nature of association between exposure and outcome, it is not identical to RR. The relationship of OR and RR is complex. [Table 3] shows RR and OR for different event rates. When there is no association between exposure and outcome, both OR and RR are identical and equal to 1.0 [Table 3]a. When there is an association between an exposure and an outcome, OR exaggerates the estimate of their relationship (is farther from 1.0 than RR). Thus, when RR <1, OR is lower than RR [Table 3]b; by contrast, when RR is more than 1.0, OR is higher than the RR [Table 3]c,[Table 3]d,[Table 3]e. When the outcome is rare (typically <10%), the value of OR is not too different from that of RR, and the two can be used interchangeably irrespective of whether the risk is lower [Table 3]b or higher [Table 3]c in the exposed group as compared to the unexposed.^{[1]} As event rates increase [Tables 3]d or [Table 3]e, the two ratios diverge and can no longer be used interchangeably.
When Should One Use Risk Ratio and Odds Ratio?   
Calculation of risk requires the use of “people at risk” as the denominator. In retrospective (casecontrol) studies, where the total number of exposed people is not available, RR cannot be calculated and OR is used as a measure of the strength of association between exposure and outcome. By contrast, in prospective studies (cohort studies), where the number at risk (number exposed) is available, either RR or OR can be calculated.
Multiple logistic regression, a frequently used multivariate technique, calculates adjusted ORs and not RRs.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
References   
1.  Sedgwick P. Relative risks versus odds ratios. BMJ 2014;348:g1407. 
[Table 1], [Table 2], [Table 3]
This article has been cited by  1 
Custom Focal Trough in ConeBeam Computed Tomography Reformatted Panoramic Versus Digital Panoramic for Mental Foramen Position to Aid Implant Planning 

 Khaled Beshtawi,Emad Qirresh,Mohamed Parker,Shoayeb Shaik   Journal of Clinical Imaging Science. 2020; 10: 34   [Pubmed]  [DOI]   2 
Intensive care utilization in patients with endstage liver disease: A populationbased comparative study of cohorts with and without comorbid hepatocellular carcinoma in taiwan 

 JenKuei Peng,HaoHsiang Chang,Irene J Higginson,Wei Gao   EClinicalMedicine. 2020; 22: 100357   [Pubmed]  [DOI]   3 
Crossing boundaries online: A hybrid method study of patients’ convergence of mass and interpersonal communication on forums. (Preprint) 

 Remco Sanders,Annemiek J. Linn,Theo Araujo,Rens Vliegenthart,Mies van Eenbergen,Julia C.M. van Weert   Journal of Medical Internet Research. 2020;   [Pubmed]  [DOI]   4 
Place of death and factors associated with hospital death in patients who have died from liver disease in England: a national populationbased study 

 JenKuei Peng,Irene J Higginson,Wei Gao   The Lancet Gastroenterology & Hepatology. 2019; 4(1): 52   [Pubmed]  [DOI]   5 
Identifying sequential episodes of pharmacotherapy as a method for assessing treatment failure in comparative effectiveness research 

 Greg Carney,Malcolm Maclure,Ken Bassett,Suzanne Taylor,Colin R. Dormuth   Pharmacoepidemiology and Drug Safety. 2019;   [Pubmed]  [DOI]   6 
Effects of racism on the socioemotional wellbeing of Aboriginal Australian children 

 D. M. Macedo,L. G. Smithers,R. M. Roberts,Y. Paradies,L. M. Jamieson   International Journal for Equity in Health. 2019; 18(1)   [Pubmed]  [DOI]   7 
Making sense of medieval mouths: Investigating sex differences of dental pathological lesions in a late medieval Italian community 

 Trent M. Trombley,Sabrina C. Agarwal,Patrick D. Beauchesne,Caroline Goodson,Francesca Candilio,Alfredo Coppa,Mauro Rubini   American Journal of Physical Anthropology. 2019;   [Pubmed]  [DOI]   8 
Does ethnicracial identity modify the effects of racism on the social and emotional wellbeing of Aboriginal Australian children? 

 Davi M. Macedo,Lisa G. Smithers,Rachel M. Roberts,Dandara G. Haag,Yin Paradies,Lisa M. Jamieson,Valsamma Eapen   PLOS ONE. 2019; 14(8): e0220744   [Pubmed]  [DOI]   9 
Cardiologist and Diabetologist crosstalk in the era of cardiovascular outcome trials of novel glucoselowering drugs 

 Francesco Maranta,Lorenzo Cianfanelli,Maria Regoni,Domenico Cianflone   IJC Heart & Vasculature. 2018; 21: 80   [Pubmed]  [DOI]   10 
Risk factor for infection in spine surgery: Are the results correct? 

 Jonny Alejandro Garcia,Alejandra Del Castillo,Juan Camilo Hernández Vargas   American Journal of Infection Control. 2018;   [Pubmed]  [DOI]  



