Odds Ratio Plot

Common Clinical Trial Graphs 200 Scatter Plot 200 Line Plot 201 Bar Chart 202 Box Plot 203 Odds Ratio Plot 203... 

As a part of logistic regression analysis, odds ratio plots are an excellent way to see how much more likely a condition is to exist based on the presence of another condition. Just by glancing at an odds ratio plot, you can see whether an independent variable is significant to the dependent variable. For instance, if the odds ratio confidence interval does not cross the value of 1, then the independent variable odds ratio is significant. Examine the following graph. 

The following is an example of an odds ratio plot. It shows the odds ratios for clinical therapy, race, gender, and baseline pain score with regard to the overall clinical success of a patient. 

Here is the SAS program that creates the preceding graph. It is a bit complex, because SAS/GRAPH does not provide horizontal box plots and this is typically what is desired for odds ratio plots. So, this sample program relies extensively on the Annotate facility to produce the plot. Notes follow the program. 

Program 6.7 Creating an Odds Ratio Plot Using PROC GPLOT... 

CREATE THE ODDS RATIO PLOT. THIS IS DONE PRIMARILY THROUGH THE INFORMATION IN THE ANNOTATION DATA SET. PUT A HORIZONTAL REFERENCE LINE AT 1 WHICH IS THE LINE OF SIGNIFICANCE. proc gplot... 

Figure 6.9 Odds ratio plot with 95% confidence intervals comparing 30-day outcomes in acute ST elevation MI patients with TIMI 3 flow at 90 minutes who underwent PCI (adjunctive or delayed) versus those who did not undergo PCI (medical therapy) in the TEVII lOB and 14 trials. (From Ref. 30.)...

P risk (odds ratio) plot of treatment emergent adverse events at PT level... 

Note that hazard ratios can be plotted in the same way that odds ratios are plotted. 

Note that the confidence intervals in Figure 15.1 are not symmetric around the estimated hazard ratio. This is because confidence intervals for hazard ratios and odds ratio and indeed ratios in general are symmetric only on the log scale (see Section 4.5.5 for further details with regard to the odds ratio). Sometimes we see plots where the x-axis is on the log scale, although it will be calibrated in terms of the ratio itself, and in this case the confidence intervals appear symmetric. 

In the upper part of Figure 15.2 we see a funnel plot of trials evaluating the effect of intravenous magnesium in the treatment of myocardial infarction. Note the absence of small trials with odds ratios greater than one (which would indicate a lack of benefit for intravenous magnesium) this... 

A description of the power (which could be via power plots and/or confidence interval plots) for the clinical trial database to detect differences in event rates (based on, e.g., risk difference or odds ratio) for various control event rates (Cooper et al., 2008). 

UI data interval plot of odds ratios between drugs and placebo for discontinuation due to AE under four models. The safest and least safe drugs compared to placebo are indicated with closed symbols and thick lines. RE Freq, frequentist random effects model RE Bayesl, Bayesian random effects model with noninformative prior RE Bayes2, Bayesian random effects model with shrinkage prior RE AB, Bayesian arm-based random effects model. 

Figure 13.3 gives side-by-side forest plots of the results of the two metaanalyses for the placebo-controlled trials for the eight outcomes. For rosiglitazone, the odds ratio estimates were consistently above 1. The 95% confidence interval for myocardial infarction was > 1. For pioglitazone, with... 

Plots of median width of 95% Cl against the number of studies under the fixed effects model, (a) Risk difference, (b) relative risk, and (c) odds ratio. 

Figure 16.1 shows median width of a 95% Cl for risk difference (Figure 16.1a), risk ratio (Figure 16.1b), and odds ratio (Figure 16.1c). We can see that the width of Cl dramatically decreases as the number of studies increases regardless of the measure. Also, the plots of p = 0.4 are generally above those corresponding to = 0.2 and 0.3, particularly, for relative risk and odds ratio. This is expected since the variance of a binomial distribution is larger as the probability is closer to 0.5.

Figure 4.5 The results from a Fisher exact test analysis to identify targets in a TNF i screen. For example, p-value (/-axis) plotted against the odds ratio (X-axis).

The only unsatisfactory feature of the fit is that the first-order plot of the observed results for pentane dips farther down than that for n-hexanol, the product whose initial rate should be boosted most strongly by the initial abundance of 1-pentene, its only parent (see Fig. 5.9). This behavior is at odds with any fit based on the network 5.43. A better fit would be obtained with a physically impossible negative value of k2S or inclusion of a mechanistically inconceivable direct pathway from 2-pentene to n-hexanol. The most likely explanation of the effect is that mass transfer from the gas phase could not quite keep pace with the extremely high initial rate of consumption of CO and H2 in the liquid, so that the concentrations of these reactants in the liquid decreased temporarily. The effect would be stronger for CO as the larger molecule, and thereby cause a temporary increase in the H2-to-CO ratio in the liquid. A higher ratio favors paraffin production at the expense of the alcohols (see Example 7.5 in Section 7.3.2) and so can explain the observed behavior. (For more detail on mass-transfer effects in this reaction, see Section 12.3). 

Figure 10. Plot of the natural logarithm of the ratio of the ion signal (S) to the zero delay ion signal (S0) vs. ionizing pulse delay time for the 21783 cm 1, J =7, odd level of dysprosium. Ionizing wavelength, 2 = 3820 A.

Information concealed in data can often be most rapidly comprehended from graphical displays. Owing to the alternation in abundance between adjacent even-Z and odd-Z elements simple plots of the type absolute abundance versus atomic number will often easily obscure small diflFerences in lanthanide distribution patterns. The method often used to remove the even-Z, odd-Z eflFect is to divide one distribution, element by element, by a known distribution, and plot the resulting ratios on a logarithmic scale against a linear scale of atomic number or ionic radius (5). If the two distributions are identical, all the ratios are the same and a horizontal line appears. Trends of diflFerences in the distributions appear as curves or sloped lines. 

---