Evaluating Bidirectional Toxicity

Introduction

If an analysis is assessing low toxicity separately from high toxicity, then bidirectional toxicity variables should be used in the ADaM dataset. Toxicity variable ATOXGR may be used in addition to the bidirectional toxicity variables to capture shifts from low to high or high to low within a parameter. The low and high toxicity grade variables may be copied directly from SDTM as predecessor variables or they may be imputed or assigned based on the statistical analysis plan and sponsor requirements (e.g., unidirectional toxicity variables).

Advice/Instructions/Guidance

Mapping bidirectional lab toxicity for parameters such as hemoglobin would include analysis toxicity variables for both low and high toxicity (ATOXGRL/ATOXGRH and BTOXGRL/BTOXGRH). ATOXGR/BTOXGR would be added to analyze shifts in toxicity that can go from low to high or high to low. 

• ATOXGR: toxicity grade of AVAL or AVALC that may be based on --TOXGR in SDTM, imputed, or assigned
• ATOXGRL: low toxicity grade of AVAL or AVALC that may be based on --TOXGR in SDTM, imputed, or assigned 
• ATOXGRH: high toxicity grade of AVAL or AVALC that may be based on --TOXGR in SDTM, imputed, or assigned 
• BTOXGR: ATOXGR of the baseline record indicated by ABLFL 
• BTOXGRL: ATOXGRL of the baseline record indicated by ABLFL
• BTOXGRH: ATOXGRH of the baseline record indicated by ABLFL 

If unidirectional parameters like platelets, with toxicity grading only performed in the low direction, are included in the dataset, then only BTOXGRL and ATOXGRL would be populated for those parameters. All high-toxicity variables would be null for this parameter. Similarly, parameters with toxicity grading defined only in the high direction would have null low-toxicity parameters and only BTOXGRH and ATOXGRH would be defined. 

The analysis toxicity for the record (ATOXGR) may be represented as "1", "2", "3", "4", and "5" for high toxicity and "-1", "-2", "-3", "-4", and "-5" for low toxicity with "Grade" prepended. ATOXGR captures both low and high toxicities for special cases where shifts can occur from low toxicity grades to high toxicity grades and high toxicity grades to low toxicity grades. Alternatively, overall toxicity may be represented as "Mild", "Moderate", "Severe", "Life-threatening", and "Death" with Low or High appended to signify the direction. The examples in this article detail the former method; values can be assigned or imputed based on reporting needs. 

SHIFTy can be used to represent change in toxicity of paired baseline and post-baseline values including (BTOXGR, ATOXGR), (BTOXGRL, ATOXGRL), (BTOXGRH, ATOXGRH) for programming shift tables. For example, SHIFT1 can represent the shift from baseline in the low direction between BTOXGRL and ATOXGRL and SHIFT2 can represent the shift from baseline in the high direction between BTOXGRH and ATOXGRH. SHIFT3 can represent the overall shift from baseline between BTOXGR and ATOXGR for cases where high and low toxicity are analyzed together. Conflicting toxicity grades may result if the toxicity ranges in the SAP are provided for multiple units, since borderline values can have different toxicity levels depending on the units. If LBORRESU and LBSTRESU represent different units and result in different toxicity grades, then the Define-XML should clearly document which variable was used to determine toxicity grades.

Examples

These examples illustrate how bidirectional and unidirectional parameters in the same dataset may be handled. In the first example, because sodium is bidirectional it is graded in both low and high directions, but low toxicity is analyzed separately from high toxicity. Since they are analyzed separately, ATOXGR is not needed and bidrectional variables are sufficient. Albumin is unidirectional and graded only in the low direction; all high-toxicity variables will be null for this parameter. SHIFT1 represents the change from baseline in the low direction and SHIFT2 is the change from baseline in the high direction.

Example 1

In Example 2, the sponsor wants to analyze low toxicity and high toxicity together. This requires adding ATOXGR and BTOXGR to the ADaM dataset. The records for USUBJID 10001 are the same as in Example 1 with the addition of these variables. ATOXGR is the higher toxicity of ATOXGRL and ATOXGRH, while BTOXGR is the higher toxicity of BTOXGRL and BTOXGRH. SHIFT3 is the overall change from baseline analyzing both low and high toxicities together. 

Sodium has a value of 158 mmol/L corresponding to an ATOXGRH of grade 3 at visit 1 and a baseline value of 152 mmol/L corresponding to a BTOXGRH of grade 2. The low toxicity variables are grade 0 since this example represents an increase in sodium in the high direction. SHIFT1 shows the change from baseline in the low direction which is "0 to 0" and an increase in toxicity of "2 to 3" in the high direction. SHIFT3 is the overall shift including both directions. It is included in Example 2 for ease in programming of analysis tables that require analyzing both low and high toxicity shifts together for parameters that are graded in both directions. Albumin is 3 g/dL at this visit, which is a low toxicity of grade 1, and is 4.2 g/dL at baseline (grade 0). The high-toxicity parameters and SHIFT2 are null since there is not a grading in the high direction for this parameter. SHIFT3 is represented as "0 to -1", showing the worsening in grading. For USUBJID 10002, sodium has a value of 126 mmol/L. corresponding to an ATOXGR of grade -3 at visit 1, and a baseline value of 150 mmol/L, corresponding to a BTOXGR of grade 1. SHIFT3 shows this worsening in grading from grade 1 (high direction) to grade -3 (low direction).

Example 2

References and Resources 

ADaMIG v1.3, Sections 3.3.7, 4.9 

US Department of Health and Human Services Cancer Therapy Evaluation Program. Common Terminology Criteria for Adverse Events (CTCAE) v6.0 (MedDRA 28.0). Published July 22, 2025. Accessed August 19, 2026. 

US Food and Drug Administration. Pilot OCE/OOD Standard Safety Data Requests v1.3. Published February 2021. Accessed August 19, 2026. 

Xie L, Wang J, Sun J, Lai R. Making Lab Toxicity Tables Less Toxic on Your Brain. PharmaSUG 2019 Paper SS-306. Accessed August 19, 2026.