Author: Amira Mandour
Biostatistician | Clinical Trials & Statistical Modeling
Expert
Objective:
In this study, we aim to quantify the association between different drug types and the number of fractures among patients, while accounting for age, and to investigate whether the type of drug influences the risk of fractures in patients. Given that the outcome of interest—the number of fractures—is count data, Poisson regression is an appropriate statistical approach for modeling these associations and estimating incidence rate ratios.
Results and Interpretation:
The Poisson regression model examined the association between treatment type and the number of fractures among patients. The model output indicates a significant difference in fracture risk between treatment groups. Specifically, the coefficient for Drug A was -1.288, which, when exponentiated, corresponds to an incidence rate ratio (IRR) of approximately 0.28. This means that patients receiving Drug A experienced about 0.28 times the number of fractures compared to the reference treatment, or equivalently, patients who take drug B had 3.6 times the risk of fractures compared to patients who take drug A. Controlling for age, Age is not a significant predictor in this study, suggesting that fracture risk is primarily associated with treatment type.
Poisson Regression Models Examining Fracture Counts by Treatment and Age:
| Table1. Poisson Regression: Treatment Only | |||
| Characteristic | IRR1 | 95% CI1 | p-value |
|---|---|---|---|
| Treatment Type |
|
|
|
| B | — | — |
|
| A | 0.28 | 0.20, 0.37 | <0.001 |
| 1 IRR = Incidence Rate Ratio, CI = Confidence Interval | |||
| Table2. Poisson Regression: Treatment + Age | |||
| Characteristic | IRR1 | 95% CI1 | p-value |
|---|---|---|---|
| Treatment Type |
|
|
|
| B | — | — |
|
| A | 0.25 | 0.18, 0.35 | <0.001 |
| Age (years) | 1.07 | 0.95, 1.19 | 0.3 |
| 1 IRR = Incidence Rate Ratio, CI = Confidence Interval | |||
The Poisson regression model examining treatment type only is presented in Table 1. The results indicate a significant difference in fracture risk between treatment groups. Patients on drug B experienced approximately 3.6 times higher fracture risk than patients on drug A.
Table 2 presents the Poisson regression model including both treatment type and age. Controlling for age, patients drug B still have nearly 4 times higher fracture risk compared to drug A. Age is not a significant predictor, suggesting that fracture risk in this study is primarily associated with treatment type. Including age does not substantially change the main finding from the previous model.
These results highlight that treatment type is the strongest predictor of fracture risk, and the effect remains robust after accounting for age.
Conclusion:
In this study, Poisson regression analysis was used to investigate the relationship between treatment type, age, and the number of fractures in patients. The results demonstrated that treatment type is significantly associated with fracture risk, with one treatment group exhibiting a higher incidence of fractures compared to the other. Age, while included in the adjusted model, did not show a statistically significant effect on fracture counts in this study.