Table 2. Multivariable Binary Logistic Regression Models Examining the Association of Trial-Level Factors With Missing Visual Elements in Forest Plotsa.
| Characteristic | aOR (95% CI) | P value |
|---|---|---|
| Disease stage | ||
| Solid M0 | 0.89 (0.38-2.08) | .79 |
| Solid M1 | 0.82 (0.39-1.72) | .60 |
| Hematologic | 1 [Reference] | NA |
| Disease siteb | ||
| Breast | 1.42 (0.71-2.87) | .33 |
| Gastrointestinal | 0.92 (0.44-1.91) | .82 |
| Genitourinary | 1.84 (0.84-4.12) | .13 |
| Hematologic | 1.19 (0.58-2.48) | .63 |
| Thoracic | 0.70 (0.34-1.44) | .33 |
| Otherc | 1 [Reference] | NA |
| Treatment type | ||
| Systemic therapy | 1 [Reference] | NA |
| Supportive care | 0.68 (0.13-2.90) | .61 |
| Local therapy | 0.40 (0.05-2.72) | .35 |
| Cooperative study | ||
| Yes | 1.67 (0.81-3.47) | .16 |
| No | 1 [Reference] | NA |
| Industry funded | ||
| Yes | 0.60 (0.24-1.49) | .27 |
| No | 1 [Reference] | NA |
| Size of industry sponsor | ||
| Larger | 0.81 (0.39-1.69) | .57 |
| Smaller | 1 [Reference] | NA |
| Enrollment | 1.00 (1.00-1.00) | .89 |
| Publication date | 0.89 (0.83-0.95) | <.001 |
Abbreviations: aOR, adjusted odds ratio; M0, non-metastatic; M1, metastatic; NA, not applicable.
Each trial-level factor was evaluated in a structural causal model depicted on a directed acyclic graph to identify confounders unique to its association with missing visual elements (see Figure 2 for an example). For each multivariable model, only the effect estimate of the primary factor of interest is reported. The effect estimates of confounders are not reported here, as they are not representative of their adjusted association with missing visual elements.
The structural causal model for disease site showed no confounding factors for the association with missing visual elements; therefore, the adjusted estimate for the association of disease site with missing visual elements is the same as the unadjusted model.
Includes central nervous system, skin, endocrine, gynecologic, head and neck, and sarcoma.