Oncology is one of the main indications in clinical trials nowadays. One specific method to analyse the oncology data is to review the lab results for safety reports. Mostly the lab results are analysed based on the reference ranges. But in oncology studies, it adds another complexity level to review the lab results using toxicity grading. The NCI Common Terminology Criteria for Adverse Events (CTCAE) is a set of criteria that is used to populate toxicity grades using statistical programming in the SDTM LB domain. In this paper, we will demonstrate the implementation of toxicity grading using CTCAE version 5.0 and the potential challenges.
DESCRIPTION
The first version of CTCAE for implementing toxicity grading was incorporated in 1999. In that version, the derivation of toxicity grading was simple, and it depends only based on Upper Limit Normal (ULN) and Lower Limit Normal (LLN). But later, it adds more complexity to the derivation over time. Some of them are
- Change from baseline becomes a part of the derivation.
- Introduction to MedDRA term associated with the toxicity.
- The Clinical Assessments become a part of the derivation.
- Reference ranges for baseline results become a part of the derivation.
Here the toxicity derivation is straightforward since, if the result is below the lower limit of normal, we should assign toxicity from Grade 1 to Grade 3 based on different limits of values in each column. We should also make sure that the unit is the same as it is mentioned in the criteria. For example, We have the result as 2.5 with unit g/dl, then the toxicity grade will be 2 as it is less than 3 and greater than or equal to 2. For Grade 4 and Grade 5, we will not derive the grades. Usually, we will ignore the assessments based on clinical signs or symptoms.
Toxicity Grading for Bilirubin (based on the upper limit of normal and baseline reference range)
Here, we need to check 2 conditions for deriving the toxicity grading. One is the result based on the ULN and one is the result based on the baseline value. If the baseline range is normal, the derivation will go with ULN and if not, the derivation will go with the baseline result. For the baseline record, we need to derive the toxicity grade based on ULN only.
Toxicity Grading for Creatinine (based on the upper limit of normal and baseline reference range)
In this scenario also, we need to populate the toxicity grading based on both baseline and ULN. But here, we don’t need to check the reference range for the baseline, and we will just calculate the grades based on both baseline result and ULN and assign the maximum toxicity. Note that “;” stands for an “or” condition and we should always keep in mind to assign the maximum toxicity grade if multiple conditions are satisfied.
Toxicity Grading for Lymphocytes (based on the upper limit of normal and lower limit of normal)Here, the calculation is slightly different from what we have seen before. We must calculate the toxicity grades based on 2 scenarios. One is if the result is greater than ULN and one is the result is less than LLN. It is also recommended to keep a variable in SUPPLB to understand whether the toxicity grade is populated based on the increased value or decreased value.
Importance of assigning Grade 0 (zero)Even though we are assigning toxicities from Grade 1 – Grade 4 using CTCAE, it is also very important to keep Grade 0 (zero) as well in the dataset for analysis purposes.
If the lowest grade defined in the scale has a lower limit of ULN as per the below diagram and the CTCAE grading is based on increased values of the corresponding lab test, then results that are below ULN are assigned as Grade 0 (zero).
If the lowest grade defined has a higher limit of LLN as per the below diagram and the CTCAE grading is defined based on decreased values of the corresponding lab test, then results that are above LLN are assigned as Grade 0 (zero).
CHALLENGES AND ISSUE HANDLING
In the CTCAE tracker, we may find some difficulties for the toxicity derivation if it is not only based on the numeric values in the dataset. Some of the scenarios and decision handling are mentioned below.
| Scenario | Decision |
| Involves mutually exclusive numeric thresholds, and clinical signs or symptoms are included in the criteria | Assign a grade to the lab value based on the thresholds only and do not consider the clinical signs or symptoms information within the CTCAE v5.0. The summaries of AEs by grade will be used to review the definitive grade |
| Involves mutually inclusive numeric thresholds when depending on the assessment of clinical signs or symptoms are included | The lab test will not be assigned a grade programmatically |
| Involves multiple set of criteria for lab tests | Worst case grade will be assigned. |
| Involves character lab results | We do not standardize character lab results at a global level. The grading will be based on numeric values only |
CONCLUSION
CTCAE criteria were defined by NCI to standardize grading of Adverse Events in clinical trials using lab data. If we are populating lab toxicities by using the CTCAE criteria, then it is crucial to consider the challenges described such as baseline handling, conversions, etc. which could be subjective. We also recommend that these assumptions made are documented in an appropriate location i.e. Statistical Analysis Plan, sponsor-driven CTCAE grading guidance document, and/or footnotes in the summary tables. The data collection process in CRF should also be streamlined in order to collect the required information for toxicity grading so that the programming can be done with minimal effort. When not sure about handling a scenario for toxicity, the sponsor should reach out to the NCI-CTEP help desk to clarify the questions. When the lab toxicities are analyzed, appropriate documentation should be prepared to explain the sponsor process in grading these values. Even if an additional complexity level has been added to the CTCAE version, its implementation can be done easily and consistently with proper planning. We can also implement some SAS macros to populate these grades since the manual effort involved in this process is huge.
