Assessing gastroesophageal reflux disease (GERD) threat 5 years following laparoscopic vertical sleeve gastrectomy (LVSG) and laparoscopic Roux-en-Y gastric bypass (LRYGB): which measure of association is appropriate, odds ratio (OR) or relative risk (RR)?

Introduction

Background

Weight loss surgery is an efficient method for improving obesity-related health outcomes (1). The International Federation for Surgery for Obesity and Metabolic Disorder (IFSO) 8^th Global Registry Report issued in 2023 (2) revealed that 480,970 bariatric procedures were performed between 2021 and 2022; with laparoscopic vertical sleeve gastrectomy (LVSG) (60.4%) and laparoscopic Roux-en-Y gastric bypass (LRYGB) (29.5%) being the two most frequently performed procedures. Though both operative approaches are credited with achieving long-term weight loss and improving comorbidities, numerous studies have shown worse gastroesophageal reflux disease (GERD) outcomes following LVSG, and discretion has been encouraged in performing LVSG in patients with pre-existing severe GERD (3). We therefore assessed GERD consequences at 5 years after LVSG and LRYGB procedures based on five randomized controlled trials (RCTs) (4-8). The aim of this paper was to compare the odds ratio (OR) and the relative risk (RR) as a measure of effect size and strength of association for only one categorial outcome, i.e., GERD. These two statistical measures in clinical research are often misused, misunderstood and misinterpreted, as demonstrated by several analysis in clinical research (9-11).

Rationale and knowledge gap

The uncertainty surrounding the usage of RR and OR may be secondary to inadvertently mixing up the mathematical description of the risk and odds (12). This is because individuals apply risk (probability) and odds (not probability) interchangeably in a casual manner. Even though both statistical methods measure the same thing, the ‘likelihood’ of a specific outcome or event is measured differently. The “risk” is measured as the ratio of events to the total number of all possible events (all subjects), whereas the “odds” is measured as the ratio of the number of events to the number of non-events. This difference between risk (probability) and odds (not probability) is crucial to the researchers and decision makers to appreciate the fundamental differences between these two statistical concepts. Although in a very special situation (e.g., rare disease assumption), the numerical value of the two ratios could be very close/similar, they are by definition dissimilar due to their fundamental conceptual differences. Recently, Khan (13) provided details on the differences between RR and OR.

Referring to previous reports (14,15) “Odds ratio (OR) originally was proposed to determine whether the probability of an event (or disorder) is the same or different across two groups, generally in a high-risk group and in a low-risk group”. This is one of several instances in the epidemiological and public health literature where odds is erroneously and exchangeably used with probability and/or risk. Stare et al. (16) wrote, “Most readers perceive OR as a RR. But since such perception is mostly correct, there is nothing (or almost nothing) wrong with that”—which is a fundamentally flowed statement. Andrade (17) erroneously wrote, “ORs are interpreted in the same way as RRs”. Interestingly, Grant (18) described ORs are necessary evil in medical research and suggests to convert them to RRs, however, the RR and OR are not statistically comparable when the outcome is common (19). Simon (9) failed to distinguish between OR and RR when both of them were described as the relative likelihood of an event occurring between two groups without any regards to the differences in the denominator of risk and odds. Lee (20) and Schmidt and Kohlmann (21) discussed RR and OR in the context of cross-sectional studies without differentiating them. OR was used by Pearce (22) in case-controlled studies. Zhang and Yu (23) discussed the limitation of estimating OR by RR in cohort studies. A generalization of the OR and RR for a r × k contingency table was proposed by Edwardes and Baltzan (24). For estimating the association between multiple risk factors, Ospina et al. (25) used the RR as an alternative to OR. For cluster randomized trials, Ukoumunne et al. (26) compared the RR, OR and risk difference. Merrill (27) noted OR as the ‘relative probabilities’ of disease in case-control studies without recognizing that ‘probabilities’ are different from odds.

Regardless of whether odds are used interchangeably with probability (RR), the OR is a very important measure of association in epidemiological and clinical studies because it (I) provides an estimate of strength of relationship between two binary variables; (II) enables us to examine the effect of predictor variables on the outcome variable via logistic regression; and (III) has a special and very convenient interpretation in case-control studies.

Objective

We are therefore going to explore the issue of OR and RR to determine the occurrence of GERD at 5 years following two different bariatric procedures, i.e., LVSG and LRYGB. The risk (probability) of GERD (outcome) is the number of patients experienced GERD relative to the total number of patients who either suffered or not suffered GERD, i.e., total number of patients. On the other hand, the odds of GERD represent the number of patients experience GERD relative to the number of patients who did not suffer GERD (no GERD) rather than total number of patients. The numerator of both the risk and odds for GERD is the same, but the denominators are different.

Methods

Search strategies and data collation

Medline, PubMed, EMBASE, Cochrane Register of Systematic Reviews, and Science Citation Index were assessed systematically not only to identify RCTs comparing LVSG and LRYGB (Figure 1) but to make sure that all published papers meet the inclusion criteria. Search strategies for each database combined a number of medical subject headings (MeSH) headings which included “laparoscopy”[MeSH Terms] OR “laparoscopy”[All Fields] OR “laparoscopic”[All Fields], “gastric sleeve”[All Fields] OR “sleeve gastrectomy” OR “vertical sleeve gastrectomy” [All Fields], “gastric roux-en-y gastric bypass”[All Fields] OR “gastric bypass”, “gastroesophageal reflux disease”[All Fields] OR “gastro-oesophageal reflux disease”[All Fields], “weight loss surgery”[All Fields], “bariatric surgery”[All Fields], “manometry”[All Fields], “lower esophageal OR oesophageal sphincter”[All Fields], “esophageal OR oesophageal function”[All Fields], “esophageal OR oesophageal motility disorder”[All Fields], “esophageal OR oesophageal motor disorder”[All Fields], “esophageal OR oesophageal dysmotility”[All Fields], “outcomes”[All Fields], “randomized OR randomized controlled trials”[All Fields] AND “comparative trials”[All Fields]. All the retrieved articles’ bibliographies were analysed for any additional references. A thorough literature search by two authors (M.A.M. and Z.H.) was conducted to select records that confirmed compliance with the inclusion criteria. The extracted data from selected studies were compared, and agreement was reached through discussion or if needs be via a contact with corresponding authors.

Figure 1 PRISMA flow chart. GERD, gastroesophageal reflux disease; LRYGB, laparoscopic Roux-en-Y gastric bypass; LVSG, laparoscopic vertical sleeve gastrectomy; OR, odds ratio; RCT, randomized controlled trial; RR, relative risk.

Inclusion criteria

Only RCTs comparing LVSG and LRYGB in adult patients (≥18 years) with morbid obesity published from January 1999 until March 2024 in English language investigating GERD outcomes were included for analysis.

Exclusion criteria

Non-human studies, duplicate studies, abstracts, conference articles, opinion pieces, editorial letters, case studies, reviews, and meta-analyses were excluded from the final review.

Statistical analysis

Five RCTs (4-8) were analyzed (LVSG n=513, LRYGB n=498) (Table 1), all of whom reported 5-year follow-up data on GERD. Table 2 provides the binary data of 5 RCTs comparing LVSG vs. LRYGB and its impact on GERD at 5-year follow-up.

The RR is defined as the ratio (proportion) of the risks for GERD between LVSG and LRYGB. Similarly, OR is defined as the ratio of odds for patients with GERD undergoing these two bariatric procedures. In our analysis, the outcome is “GERD” and the interventions are “LVSG and LRYGB”. The comprehension of the difference between RR and OR depends on the understanding of the fundamental difference in the definition of probability and odds—both represent likelihood of GERD but on different scales. Risk (probability) is a number between 0 and 1. Odds is a number between 0 and ∞ (infinity). Due to the inappropriate interchangeable use of the words risk and odds, often misleading statements are made such as “the OR overestimates the RR and thus may inappropriately affect clinical decision-making or policy development, leading to unintentional errors in the analysis of treatments” (28). The reality is that odds can never estimate risk, and vice versa, and therefore there is no question of over- or underestimation of one by another. Undoubtedly, there is no statistical reasoning of applying OR as an estimate of RR.

Results

Odds provides direct comparison

Both the odds and risk have the same numerator, which in our case is the number of GERD patients, but they have different scales. For odds, the scale (denominator) is the number of patients with no GERD. Therefore, odds is the ‘likelihood’ of GERD relative to no GERD. Whereas the scale (denominator) of risk is the total number of patients with GERD and no GERD (total number of patients). Therefore, if the aim is to compare the outcome of an intervention (LVSG and LRYGB), it makes more sense to use odds, i.e., patients with no GERD as the denominator.

The models of RR and OR are illustrated using the count data in a 2×2 contingency table. Y represents an outcome variable with two levels, i.e., GERD and no GERD, whereas X signifies an intervention variable with two levels, i.e., LVSG and LRYGB, as in Table 2. The values in each of the four cells represent the counts (frequencies). Therefore, in Table 2, “a” and “c” represent patients who developed GERD following LVSG and LRYGB after 5 years, respectively, and “b” and “d” represent patients with no GERD following LVSG and LRYGB, respectively. Thus, the risk (or probability) of GERD is a/(a + b) or c/(c + d), i.e., GERD/(GERD + no GERD), but the odds is a/b or c/d, i.e., GERD/no GERD. The risk and not the odds is comparable to probability.

Probability

The probability of patients having GERD is the proportion of patients with GERD divided by the total number of patients (those with GERD + those with no GERD). In Table 2,

• For the LVSG group, $P\left(GERD\right)=\frac{a}{\left(a+b\right)}=\frac{138}{\left(138+375\right)}=\frac{138}{513}=0.269$;
• For the LVSG group, $P\left(noGERD\right)=\frac{b}{\left(a+b\right)}=\frac{375}{\left(138+375\right)}=\frac{375}{513}=0.731$.

The occurrence of GERD is complementary to the non-occurrence of GERD; therefore, P(GERD)=1−P(noGERD)=1−0.731=0.269.

Risk

The risk of GERD in LVSG and LRYGB groups is the probability of patients suffering GERD. This is a proportion of the number of patients with GERD relative to the total number of patients with GERD and no GERD in LVSG and LRYGB patients.

In Table 2, the risk of GERD in the LVSG and LRYGB groups is calculated as:

• For the LVSG group, $R_{LVSG}=\frac{a}{\left(a+b\right)}=\frac{138}{\left(138+375\right)}=\frac{138}{513}\times 100\%=26.90\%$;
• For the LRYGB group, $R_{LRYGB}=\frac{c}{\left(c+d\right)}=\frac{27}{\left(27+471\right)}=\frac{27}{498}\times 100\%=5.42\%$.

Odds

The odds of GERD for two procedures are as follows:

• For the LVSG group, $Odd{s}_{LVSG}=\frac{a}{b}=\frac{138}{375}=0.368$;
• For the LRYGB group, $Odd{s}_{LRYGB}=\frac{c}{d}=\frac{27}{471}=0.057$.

Note that the odds of no GERD=1Odds(GERD):

• For the LVSG group, $Odd{s}_{LVSG}=\frac{1}{0.368}=2.717$;
• For the LRYGB group, $Odd{s}_{LRYGB}=\frac{1}{0.055}=17.544$.

The odds of patients suffering GERD are directly related to the odds of patients not suffering GERD, and vice versa.

Remark

Sum of risk of ‘GERD’ and ‘no GERD’ is one. Product of odds of ‘GERD’ and ‘no GERD’ is one.

Odds range from 0 to ∞ (infinity).

• Odds(GERD) =1 implies that GERD and no GERD are likewise possible;
• Odds(GERD) >1 implies that GERD is more plausible compared to no GERD;
• Odds(GERD) <1 implies that GERD is less plausible compared to no GERD.

It is important to appreciate that the OR is a proportion of two odds, whereas the RR is a proportion of two probabilities (i.e., risks). In the context of clinical research, McHugh (29) discussed the importance of OR.

RR

The probability (or risk) of GERD in the LVSG group is RLVSG=aa+b and in the LRYGB group is RLRYGB=cc+d.

When comparing two treatment groups, if the RR <1, then one treatment group is better than the other. However, if the RR =1, the efficacy of both treatment groups is equal. Conversely, if one treatment group has the RR >1, then that treatment group has a higher risk of GERD.

The risk of GERD in LVSG patients compared to LRYGB patients measures the effect of both treatments. So, the RR is the ratio of the probability (risk) of the GERD in the LVSG group when compared with LRYGB group, i.e., RRGERD=RLVSGRLRYGB=a(a+b)c(c+d)=13851327498=13.717118.444=0.2690.054=4.96.

Interpretations

As RR conveys how many times more probable is the occurrence of GERD in LVSG vs. LRYGB patients, our analysis has revealed that the former group of patients has a 4.96 times higher risk of experiencing GERD compared to the latter group (and vice versa) at 5 years follow-up.

OR

The odds of GERD in the LVSG group is ODLVSG=ab.

Similarly, the odds of GERD in the LRYGB group is ODLRYGB=cd.

The OR of the LVSG cohort compared to the LRYGB cohort is the ratio of the two odds: OR=ORLVSGORLRYGB=abcd=a×dc×b=13837527471=138×47127×375=64,99810,125=6.42.

Interpretations

For the above example, OR =6.42 (not 6.42%). Thus, the odds of GERD for the LVSG group is 6.42 times compared to those who underwent LRYGB. Since this OR is much greater than 1, confirming 6.42 times higher odds of GERD following LVSG. Regardless of the statistics used, both OR and RR are showing stronger a association between GERD and LVSG.

Avoid division by zero/continuity correction

In many cases, a slightly amended estimator of OR is used by adding 0.5 to each cell count to avoid division by 0 as suggested by Agresti (30) and Shoukri and Pause (31). Hence, for the above example, the two amended OR (OR*) become: ODLVSG∗=(a+0.5)(b+0.5)=(138.5)(375.5)=0.369 and ODLRYGB∗=(c+0.5)(d+0.5)=(27.5)(471.5)=0.058.

Then, the OR* of the LRYGB vs. LVSG becomes OR∗=ODLVSG∗ODLRYGB∗=0.3690.058=6.36.

All statistical packages have this choice to adjust for continuity correction.

Definition of odds in terms of probabilities

According to the Boston Public School webpage (32), the odds are defined as “the probability the event will occur divided by the probability that the event will not occur”. That is, odds=p1−p, where p is the probability that the GERD occurs and (1−p) is the probability that the GERD does not occur. This definition of odds is used in the specification of the logistic regression model.

The above definition of odds as the ratio of two probabilities should not confuse anyone that odds is the same as probability. This is because the ratio of the probability of an event (GERD) and that of a non-event (no GERD) is not a probability. Note the following explanations:

Using the notations in Table 2 we have p=aa+b and 1−p=1−aa+b=a+b−aa+b=ba+b.

Then the odds of GERD against no GERD is defined as the ratio of their probabilities as p1−p=aa+bba+b=ab.

Thus, although the odds is defined as the ratio of probabilities of GERD and no GERD, it is essentially the ratio of the numbers of GERD and no GERD.

Theorem 1

The ratio of RR of GERD and no GERD is the same as the OR of GERD and no GERD in the LVSG group vs. the LRYGB group.

In notation, ORGERD=RRLVSGRRLRYGB.

Proof

From the cell counts in Table 2, RR of GERD in the LVSG group relative to the LRYGB group is: RRGERD(LVSGLRYGB)=[riskGERD(LVSG)][riskGERD(LRYGB)]=RRLVSGRRLRYGB=a/(a+b)c/(c+d)=ac×c+da+b.

Similarly, RR of no GERD in the LVSG group relative to the LRYGB group is:RRNoGERD(LVSGLRYGB)=[riskNoGERD(LVSG)][riskNoGERD(LRYGB)]=b/(a+b)d/(c+d)=db×c+da+b.

Now, the ratio of the above two RRs, for GERD and no GERD, is ac×c+da+bdb×c+da+b=a/cb/d=a×db×c.

The odds of GERD in the LVSG group is a/b and that in the LRYGB group is c/d. Hence, the OR of GERD in the LVSG group relative to the LRYGB is ORLVSG=a/bc/d=a×db×c.

Inference on RR and OR

To find the confidence interval and perform a statistical test of significance on RR and OR, their sampling distributions are required. Since the distributions of RR and OR are not readily available, log transformation of these statistics is used for inferential purposes. This is because both ln(RR) and ln(OR) follow an approximate normal distribution.

For inference on RR, use the log transformation ln(RR) with approximate standard error (SE), SE[ln(RR)]=1a−1a+b+1c−1c+d.

For inference on OR, use the log transformation ln(OR) with approximate SE, SE[ln(OR)]=1a+1b+1c+1d.

To test the null hypothesis, H0:ln(RR)=0 against the alternative hypothesis Ha:ln(RR)≠0 use the test statistic, Z=ln(RR)SE[ln(RR)].

A 95% confidence interval for ln(RR) is given by ln(RR)±1.96×SE[ln(RR)].

For the data in Table 2, ln(RR) = ln(4.96) = 1.60, SE[ln(RR)]=1a−1a+b+1c−1c+d=1138−1138+375+127−127+471=1138−1513+127−1498=0.20081, z=1.600.20081=7.97, and P value <0.001.

The 95% confidence interval for ln(RR) is ln(RR)±1.96×SE[ln(RR)], or 1.60±1.96×0.20081, that is (1.2064, 1.994) in the ln(RR) scale.

The exponential of the limits of this confidence interval gives the 95% confidence interval in the original RR scale, which is (3.341, 7.345). As this interval does not contain ‘1’ inside it, the estimated value of RR is significant at 5% significance level, which reaffirms the significance of RR by the P value (<0.001) of the test.

Similarly, for inference on OR, we get ln(OR) = ln(4.96) = 1.60, SE[ln(OR)]=1a+1b+1c+1d=1138+1375+127+1471=0.00725+0.00267+0.03704+0.00212=0.04908=0.22152, z=1.600.22152=7.22, with P value <0.001.

Therefore, the test is highly significant. That is, the ln(OR) is significantly different from zero.

The 95% confidence interval for ln(OR) is ln(OR)±1.96× SE[ln(OR)], or 1.60±1.96×0.22152, that is, (1.166, 2.034) in the ln(OR) scale.

The 95% confidence interval in the ln(OR) scale is (1.166, 2.034) and in the original OR scale is (3.209, 7.645) (Table 2). As the 95% confidence interval of OR does not contain ‘1’ inside it and the P value is less than 0.05, based on both 95% confidence interval and P value, the OR is statistically significant. Thus, there is a very strong association between LVSG and GERD as evidenced by both RR and OR.

RR and OR from incidence and prevalence

In epidemiology and public health, the two frequently used terms are incidence and prevalence. They may be related but very distinct. They are related to RR and OR. Prevalence (also called prevalence rate) is defined as the probability that a proportion of individuals in a population has a given disease/disorder at a specific point in time or over a defined period. So, it measures the actual number of cases alive with a disease/disorder either at a given period in time (period prevalence) or at a particular date in time (point prevalence). Incidence (also called incidence rate) is referred to a measure of the occurrence of new cases of disease/disorder during a particular period of time within a specified population. Normally, prevalence is used for cross-sectional data and incidence for longitudinal data. Thus, in our analysis, the incidence of GERD is calculated over a 5-year period in two distinct populations, i.e., following LVSG or LRYGB over a 5-year period.

Incidence rate and RR

Incidence rate is a measure that reflects how frequently new cases of a disease/disorder appear within a population over a specific time period. It’s a dynamic measure, indicating the risk of developing the condition during that time. Thus, incidence rate is another way to represent the risk at a point in time. In our analysis, we have calculated the incidence of GERD at 5 years following LVSG and LRYGB. The ratio of incidences in the LVSG group (I_LVSG) divided by that in the LRYGB group (I_LRYGB) gives the RR.

That is, RR=ILVSGILRYGB.

From the data in Table 2, ILVSG=aa+band ILRYGB=cc+d.

As the incidence rate (risk, i.e., GERD) in any of these bariatric surgical groups increases, the value of the RR grows larger.

Conversion of OR to RR

So, there is a contentious issue of whether or not the OR should be converted to RR. Most of the epidemiologists feel that if OR is properly interpreted and explained, there is no need for such a conversion. However, some investigators still attempt to convert OR to RR in order to make it easier to understand by non-specialized readers. Zhang and Yu (23) and Cummings (33) separately proposed conversion formulas of OR to RR as RR=OR1−ILRYGB+ILRYGB×OR.

It is interesting to note that (1−ILRYGB)=1−c(c+d)=dc+d=ILRYGB0, which is the no GERD rate in the LRYGB group. Clearly, the sum of ILRYGB and ILRYGB0 is one. Similarly, for the LVSG group, (1−ILVSG)=1−a(a+b)=ba+b=ILVSG0, and hence, ILVSG and ILVSG0 add to one.

Based on our data in Table 2, ILVSG=aa+b=RLVSG (risk of GERD in the LVSG group) and ILRYGB=cc+d=RLVSG (risk of GERD in the LRYGB group).

Then, (1−ILVSG)=b(a+b)=RNoGERDLVSG (risk of no GERD in the LVSG group) and (1−ILRYGB)=dc+d=RNoGERDLRYGB (risk of no GERD in the LRYGB group).

Therefore, the RR of GERD in the LVSG group relative to the LRYGB group is RRGERD(LVSG|LRYGB)=RLVSGRLRYGB=ILVSGILRYGB.

Similarly, the RR of no GERD in the LVSG group relative to the LRYGB group is RRNoGERD(LVSG|LRYGB)=RNoGERD(LVSG)RNoGERD(LRYGB)=(1−ILVSG)(1−ILRYGB).

The RR of GERD in the LVSG group relative to the LRYGB group, [RRGERD(LVSG|LRYGB)] is the product of that of the no GERD, [RRNoGERD(LVSG|LRYGB)] and OR of GERD in the LVSG group relative to the LRYGB group, [ORGERD(LVSG|LRYGB)]. The OR of GERD in the LVSG group relative to the LRYGB group is ORGERD(LVSG|LRYGB)=abcd=adbc and RRNoGERD(LVSG|LRYGB)=RNoGERD(LVSG)RNoGERD(LRYGB)=ba+bdc+d=b(c+d)d(a+b).

The product of the two becomes, ORGERD(LVSG|LRYGB)×RRNoGERD(LVSG|LRYGB)=adbc×b(c+d)d(a+b)=aa+bcc+d=RRGERD(LVSG|LRYGB).

Hence, RRGERD(LVSG|LRYGB)=RRNoGERD(LVSG|LRYGB)×ORGERD(LVSG|LRYGB).

Remark

From the numerical data in Table 2, we have ORGERD(LVSG|LRYGB)=0.3680.057=6.45 and RRNoGERD(LVSG|LRYGB)=375513471498=0.7300.945=0.77.

Hence, RRGERD(LVSG|LRYGB)=6.45×0.77=4.96.

Hu et al. (34) proposed a method to convert adjusted group-level disease risk reported on a probability scale to the logOR for the study of livestock.

Discussion

The RR and the OR both compared the likelihood of an event (GERD) occurring in two bariatric surgical groups (LVSG and LRYGB), but the RR compares probabilities (or risk) while the OR compares odds. In other words, the RR does not exhibit effect magnitude alone but instead is a ratio of two conditional probabilities that vary with outcome prevalence. The OR, on the other hand, is a true effect magnitude measure, which means it measures the actual, underlying size of an effect in a population as opposed to the estimated effect size from a sample and therefore represents the fold increase in odds of outcome (GERD) from an unexposed state (non-surgical) to an exposed state or two different exposed states, i.e., LVSG and LRYGB. The issue of OR often overestimating RR is a controversial one. Recently, Doi et al. (35) addressed this issue, stating “The RR varies for reasons other than the magnitude of the effect because it is a ratio of two posterior probabilities, both of which are dependent on baseline prevalence of an outcome. On the other hand, the OR, the other commonly used ratio, measures solely the effect magnitude and has no relationship to the prevalence of an outcome in a study nor does it overestimate the RR as is commonly thought”. These authors feel the primary use of the RR in clinical trials and meta-analysis as its direct interpretation is not meaningful and RR should be replaced with OR.

Comparison of RR and OR

• OR determines the association between two variables, whereas RR determines the relationship in risk status based on some variables.
• OR determines the association between an intervention (LVSG or LRYGB) and risk (GERD). RR, on the other hand, tells us how an intervention changes risk.
• When the event/disorder is rare or very unusual in two different treatment groups, the RR and OR are comparable in magnitude. The OR, however, is larger compared to the RR when the event is more frequent. In our analysis, the event, i.e., GERD, following LVSG compared to LRYGB was large and therefore OR applied for calculating effect size.
• The OR can overestimate and magnify the risk, especially when the event is more common. This was not seen in our analysis.
• One benefit of OR is that it is not reliant on if one focuses on the event/disorder (GERD) occurring or not, i.e., OR is symmetric to which outcome level is of interest; however, RR is not symmetric. That is, Odds(GERD) = 1/Odds(no GERD).
• If the OR for an event (e.g., GERD) departs from 1 considerably, the OR of its non-event (no GERD) will also deviate from 1 markedly, but in the opposite direction.

Conclusions

There is a misconception that if the outcome, such as GERD, is more common in the patient’s population following a particular bariatric intervention, the OR would overstate or exaggerate the effect of the intervention or produce more “extreme” values compared to RR. This was not observed in our analysis, although the strength of association with OR was stronger. Furthermore, if the relevant prevalence or incidence data are not available, the OR provides a valid effect measure, as seen in our analysis. The OR for a given exposure is normally acquired using logistic regression model while controlling for confounders. The accessibility of this methodology in standard statistical software essentially illustrates the popularity of OR over RR. However, it does not have as insightful rationalization as the RR for non-specialist readers. In our analysis, the OR of GERD for the LVSG group compared to LRYGB is 6.42 times higher. Since this OR is much greater than 1, confirming 6.42 times higher odds of GERD following LVSG relative to LRYGB. On the other hand, the RR is 4.96, which may suggest that the OR may have slightly overstated the association or magnified the effect of LVSG for GERD. Often, the choice of RR or OR is determined by the study objective/design and choice or preference of the researcher. However, if one wants to know how any exposure, e.g., LVSG or LRYGB, affects the outcome, i.e., GERD, we feel OR is the best effect measure of association. When OR is converted to RR, which we have undertaken in our analysis using an appropriate methodology and calculations, valid interpretations of risk can be achieved. It is therefore imperative that the researchers should choose the right statistical measure of association for binary outcomes, i.e., OR or RR, in a clinical setting and use them consistently, without interchanging them, depending on what their analytical goals are.

Acknowledgments

None.

Footnote

Peer Review File: Available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-35/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-35/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. American Society for Metabolic and Bariatric Surgery (ASMBS). Estimate of Bariatric Surgery Numbers, 2011-2023. Accessed 13th May, 2024. Available online: https://asmbs.org/resources/estimate-of-bariatric-surgery-numbers/
2. International Federation for Surgery for Obesity and Metabolic Disorders. 8th global registry report. Accessed 13th May, 2024. Available online: https://www.ifso.com/pdf/8th-ifso-registry-report-2023.pdf
3. Di Lorenzo N, Antoniou SA, Batterham RL, et al. Clinical practice guidelines of the European Association for Endoscopic Surgery (EAES) on bariatric surgery: update 2020 endorsed by IFSO-EC, EASO and ESPCOP. Surg Endosc 2020;34:2332-58. [Crossref] [PubMed]
4. Zhang Y, Zhao H, Cao Z, et al. A randomized clinical trial of laparoscopic Roux-en-Y gastric bypass and sleeve gastrectomy for the treatment of morbid obesity in China: a 5-year outcome. Obes Surg 2014;24:1617-24. [Crossref] [PubMed]
5. Ignat M, Vix M, Imad I, et al. Randomized trial of Roux-en-Y gastric bypass versus sleeve gastrectomy in achieving excess weight loss. Br J Surg 2017;104:248-56. [Crossref] [PubMed]
6. Salminen P, Helmiö M, Ovaska J, et al. Effect of Laparoscopic Sleeve Gastrectomy vs Laparoscopic Roux-en-Y Gastric Bypass on Weight Loss at 5 Years Among Patients With Morbid Obesity: The SLEEVEPASS Randomized Clinical Trial. JAMA 2018;319:241-54. [Crossref] [PubMed]
7. Peterli R, Wölnerhanssen BK, Peters T, et al. Effect of Laparoscopic Sleeve Gastrectomy vs Laparoscopic Roux-en-Y Gastric Bypass on Weight Loss in Patients With Morbid Obesity: The SM-BOSS Randomized Clinical Trial. JAMA 2018;319:255-65. [Crossref] [PubMed]
8. Biter LU, 't Hart JW, Noordman BJ, et al. Long-term effect of sleeve gastrectomy vs Roux-en-Y gastric bypass in people living with severe obesity: a phase III multicentre randomised controlled trial (SleeveBypass). Lancet Reg Health Eur 2024;38:100836. [Crossref] [PubMed]
9. Simon SD. Understanding the odds ratio and the relative risk. J Androl 2001;22:533-6.
10. Holcomb WL Jr, Chaiworapongsa T, Luke DA, et al. An odd measure of risk: use and misuse of the odds ratio. Obstet Gynecol 2001;98:685-8. [Crossref] [PubMed]
11. Tajeu GS, Sen B, Allison DB, et al. Misuse of odds ratios in obesity literature: an empirical analysis of published studies. Obesity (Silver Spring) 2012;20:1726-31. [Crossref] [PubMed]
12. Khan S, Memon MA. Measuring association between two categorical variables: Revisiting risk ratio and odds ratio. Int J Stat Sciences 2020;20:153-70.
13. Khan S. Meta-Analysis Methods for Health and Experimental Studies. 1st ed. Singapore: Springer Nature; 2020.
14. Bland JM, Altman DG. Statistics notes. The odds ratio. BMJ 2000;320:1468.
15. Chen H, Cohen P, Chen S. How big is a big odds ratio? Interpreting the magnitudes of odds ratios in epidemiological studies. Commun Stat Simul Comput 2010;39:860-4.
16. Stare J, Maucort-Boulch D. Odds ratio, hazard ratio and relative risk. Advances in Methodology and Statistics 2016;13:59-67.
17. Andrade C. Understanding Misunderstandings About the Relative Risk and Odds Ratio. J Clin Psychiatry 2023;84:23f14943.
18. Grant RL. Converting an odds ratio to a range of plausible relative risks for better communication of research findings. BMJ 2014;348:f7450. Erratum in: BMJ 2014;348:g2124.
19. Persoskie A, Ferrer RA. A Most Odd Ratio:: Interpreting and Describing Odds Ratios. Am J Prev Med 2017;52:224-8. [Crossref] [PubMed]
20. Lee J. Odds ratio or relative risk for cross-sectional data? Int J Epidemiol 1994;23:201-3. [Crossref] [PubMed]
21. Schmidt CO, Kohlmann T. When to use the odds ratio or the relative risk? Int J Public Health 2008;53:165-7. [Crossref] [PubMed]
22. Pearce N. What does the odds ratio estimate in a case-control study? Int J Epidemiol 1993;22:1189-92. [Crossref] [PubMed]
23. Zhang J, Yu KF. What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998;280:1690-1.
24. Edwardes MD, Baltzan M. The generalization of the odds ratio, risk ratio and risk difference to r x k tables. Stat Med 2000;19:1901-14. [Crossref] [PubMed]
25. Ospina PA, Nydam DV, DiCiccio TJ. Technical note: The risk ratio, an alternative to the odds ratio for estimating the association between multiple risk factors and a dichotomous outcome. J Dairy Sci 2012;95:2576-84. [Crossref] [PubMed]
26. Ukoumunne OC, Forbes AB, Carlin JB, et al. Comparison of the risk difference, risk ratio and odds ratio scales for quantifying the unadjusted intervention effect in cluster randomized trials. Stat Med 2008;27:5143-55. [Crossref] [PubMed]
27. Merrill RM. Introduction to epidemiology. 9th ed. Burlington: Jones & Bartlett Learning; 2021.
28. Balasubramanian H, Ananthan A, Rao S, et al. Odds ratio vs risk ratio in randomized controlled trials. Postgrad Med 2015;127:359-67. [Crossref] [PubMed]
29. McHugh ML. The odds ratio: calculation, usage, and interpretation. Biochem Med (Zagreb) 2009;19:120-6.
30. Agresti A. An introduction to categorical data analysis. 3rd ed. Hoboken: John Wiley & Sons; 2018.
31. Shoukri MM, Pause CA. Statistical Methods for Health Sciences. 2nd ed. Boca Raton: CRC Press; 1999.
32. Boston University School of Public Health. The Difference Between "Probability" and "Odds". 2020. Accessed 18 May 2024.
33. Cummings P. The relative merits of risk ratios and odds ratios. Arch Pediatr Adolesc Med 2009;163:438-45. [Crossref] [PubMed]
34. Hu D, Wang C, O'Connor AM. A method of back-calculating the log odds ratio and standard error of the log odds ratio from the reported group-level risk of disease. PLoS One 2020;15:e0222690. [Crossref] [PubMed]
35. Doi SA, Furuya-Kanamori L, Xu C, et al. Controversy and Debate: Questionable utility of the relative risk in clinical research: Paper 1: A call for change to practice. J Clin Epidemiol 2022;142:271-9. [Crossref] [PubMed]