Revision History

© 2023 Clinical Data Interchange Standards Consortium, Inc. All rights reserved. 

1 Introduction

This document describes the CDISC implementation of the Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST 1.1) disease response criteria.

CDISC is not including an annotated case report form (CRF) for this disease response supplement as the primary purpose of the supplement is to show examples of how to implement the disease response criteria in CDISC data standards domains.

The representation of data collected for the disease response criteria is based on the Study Data Tabulation Model Implementation Guide: Human Clinical Trials (SDTMIG) Disease Response and Clin Classification (RS) domain model, which can be found at the CDISC website at: https://www.cdisc.org/standards/foundational/sdtmig.

These specific implementation details for the disease response criteria are meant to be used in conjunction with the SDTMIG. All CDISC QRS documentation packages can be found on the CDISC website at: https://www.cdisc.org/standards/foundational/qrs. 

The CDISC Intellectual Property Policy can be found on the CDISC website at: https://www.cdisc.org/about/bylaws.

1.1 Representations and Warranties, Limitations of Liability, and Disclaimers

This document is a supplement to the SDTMIG and is covered under Appendix F of that document, which describes representations, warranties, limitations of liability, and disclaimers. Please see Appendix F of the SDTMIG for a complete version of this material.

CDISC specifies how to structure the data that has been collected in a database, not what should be collected or how to conduct clinical assessments or protocols. CDISC disclaims any liability for your use of this material.

Although the United States Food and Drug Administration (US FDA) has provided input with regard to this supplement, this input does not constitute US FDA endorsement of any particular instrument. 

2 Copyright Status

The RECIST Working Group and European Journal of Cancer (EJC) own the copyright for the RECIST 1.1 guidelines and have made the guidelines available for non-commercial purposes exempt from copyright restrictions. All other rights are reserved. CDISC has included this supplement in the CDISC inventory of QRS data standards supplements. Hence, CDISC developed data standards for tumor identification, tumor results, and disease response, and applied these to RECIST 1.1.

The CDISC documentation of the criteria consists of: (1) controlled terminology and (2) standard data structures with examples.

Note: CDISC Controlled Terminology is maintained by National Cancer Institute (NCI) Enterprise Vocabulary Services (EVS). The most recent version should be accessed through the CDISC website at: https://www.cdisc.org/standards/terminology/controlled-terminology.

CDISC has developed this documentation at no cost to the copyright holder or any additional cost to users of the criteria beyond the normal license fees charged by the copyright holder.

References for RECIST 1.1:

• E.A. Eisenhauer, P. Therasse, J. Bogaerts, L.H. Schwartz, D. Sargent, R. Ford, J. Dancey, S. Arbuck, S. Gwyther, M. Mooney, L. Rubinstein, L. Shankar, L. Dodd, R. Kaplan, D. Lacombe, J. Verweij. New response evaluation criteria in solid tumours; Revised RECIST guideline (version 1.1). European Journal of Cancer 2009; 45:228-247.
  
• https://recist.eortc.org/recist-1-1-2/

3 The Oncology Disease Response and Supporting Domains Model for RECIST 1.1

RECIST 1.1 is a set of criteria used to evaluate the activity and efficacy of antineoplastic agents in solid tumors. The Oncology Disease Response and Supporting Domains Model for RECIST 1.1 encompasses the subset of SDTMIG domains used for the representation of RECIST 1.1 disease response and supporting tumor/lesion data in oncology studies. These SDTMIG domains include:

• Disease Response and Clin Classification (RS), a findings domain used to represent disease response findings and QRS clinical classifications instruments. The disease responses based on RECIST 1.1 are represented in RS. 
• Tumor/Lesion Identification (TU), a findings domain. The supporting tumor identification of RECIST 1.1 target, non-target, and new tumors are represented in the TU domain. 
• Tumor/Lesion Results (TR), a findings domain. The supporting tumor measurements and assessments of the RECIST 1.1 target, non-target, and new tumors are represented in the TR domain.

Note that the Procedures (PR) domain is not required for RECIST 1.1 but may be used to represent scan data including medical photography data.

These domains are related using RSLNKGRP, TULNKID, TUREFID, TRLNKID, TRLNKGRP, TRREFID, and PRREFID. See Section 5, RELREC, for more information on the relationships and an example RELREC dataset.

CDISC biomedical concepts are available for RECIST 1.1 following the 2-layered approach:

• Conceptual/abstract layer that provides standards-agnostic, unambiguous semantic definition largely based on NCIt concepts
• Implementation layer based on valid CDISC dataset specializations that provide value-level metadata definitions that facilitate metadata-driven automation

SDTM dataset specializations have been defined for the TU, TR, and RS datasets. They are retrievable via the CDISC Library API as JSON output. See the CDISC Biomedical Concepts page (https://www.cdisc.org/cdisc-biomedical-concepts) for more information.

3.1 Assumptions for the Oncology Disease Response and Supporting Domains Model

All assumptions and business rules described in the SDTMIG are applicable to this supplement. Additional assumptions specific to RECIST 1.1 are listed below.

1. The CDISC Oncology Codetable (available at https://www.cdisc.org/standards/terminology/controlled-terminology) contains the valid value list for the tests in the TU, TR, and RS domains. 

   1. RSORRES is populated with the original result and RSSTRESC is populated with the standardized character result according to the associated controlled terminology. For example, RSORRES = "CR" and RSSTRESC = "CR" where "CR" is a valid value in the associated controlled terminology.

   2. TUORRES is populated with the original result and TUSTRESC is populated with the standardized character result according to the associated controlled terminology. For example, TUORRES = "TARGET" and TUSTRESC = "TARGET" where "TARGET" is a valid value in the associated controlled terminology.

   3. For the TR domain, TRORRES and TRSTRESC can contain numeric or character results and TRSTRESN is only populated for numeric results. There are situations when a character result in TRORRES is standardized to a numeric representation of the result in TRSTRESC and TRSTRESN.  

3.2 General Points on Representation of Data within the Oncology Disease Response Domains for RECIST 1.1

In RECIST 1.1, a subject's tumors are identified as target tumors, non-target tumors, and new tumors. At a specific disease assessment timepoint, the overall response in RECIST 1.1 is based on measurements of target tumors (target response), qualitative assessments of non-target tumors (non-target response) and the appearance of new tumors. Subjects with at least 1 target tumor can have overall response of CR (Complete Response), PR (Partial Response), SD (Stable Disease), PD (Progressive Disease) and NE (Not Evaluable). Subjects with only nonmeasurable disease (i.e., no target tumors) can have overall response of CR, NON-CR/NON-PD, PD, and NE.

Target tumors are measured in the longest diameter except when the location is a lymph node, when they are measured in the longest perpendicular. The examples in this supplement use TRTESTCD = "LDIAM" (Longest Diameter) for measurements of non-lymph node target tumors and TRTESTCD = "LPERP" (Longest Perpendicular) for measurements of lymph node target tumors. SDTMIG examples have previously used TRTESTCD = "DIAMETER" (Diameter) for measurements of both lymph nodes and non-lymph nodes. Either representation is acceptable for RECIST 1.1. The latter approach requires looking at the anatomical location in conjunction with the TRTESTCD to know whether the diameter was measured in the longest diameter or the longest perpendicular. The LDIAM and LPERP tests provide more specificity in the TRTESTCD itself and do not require looking at the anatomical location. In addition, the LDIAM and LPERP tests are used in other disease response criteria; the use of these 2 tests provides standardization across criteria for the same measurement. Some sponsors prefer the approach used within this disease response supplement due to these reasons. 

Target tumors may split (fragment) and/or merge (coalesce) during a study. See Section 4.2, Example 2: Split and Merged Tumors, for the preferred representation of measurements of split and merged tumors. Regardless of the representation of split and merged lesions, it is essential that the summation of the diameters is accurate.

In RECIST1.1, the following calculations are used in determination of the target response: sum of the diameters of target tumors (target sum), percentage change from baseline in the target sum, and percentage change from nadir in the target sum (smallest target sum at any assessment). These are referred to as summary values in the examples that follow. These summary values are not typically collected/calculated in investigator CRFs, but rather are usually provided by independent assessors. When not derived in EDC, summary values not in the collected data may be derived in the analysis datasets (i.e., ADaM).

RECIST 1.1 recommends using a standardized value of 5 mm in the calculation of sum of diameters when a target tumor is too small to measure. Note that this imputation should only be done when the lesion diameter is too small to measure. When a numerical value is given for the lesion diameter, then that value will be used even if the diameter is less than 5 mm. The original or collected value "TOO SMALL TO MEASURE" should be represented in the TRORRES variable and the standardized value should be represented in the TRSTRESC and TRSTRESN variables. For absent non-lymph node target tumors, 0 mm is used in the calculation of the target sum. The longest perpendicular measurements of lymph node target tumors which are pathological (>= 10 mm) or non-pathological (< 10 mm) are included in the target sum. 

The examples in this supplement use RSTESTCD = "NEWLPROG" (New Lesion Progression) to indicate whether a new lesion is equivocal or unequivocal. There are different methods of collecting data when there is an equivocal new lesion. Some sponsors update the overall response when unequivocal evidence of a new lesion has been later confirmed (i.e., documented as unequivocal). Other sponsors programmatically derive the new lesion progression date as the date when the new lesion was first identified during the analysis. RSTESTCD = "NEWLSIND" can have RSORRES = "Y" or RSORRES = "N" values only. This test can be used if the sponsor collects yes and no responses to indicate that new lesion(s) were identified at the disease assessment timepoint.

"NE" (Not Evaluable) is a standard result in the Oncology Response Assessment Result codelist (ONCRSR) which is applicable to RECIST 1.1. Therefore, in RS, "NE" is a valid RSORRES/RSSTRESC for RSTEST = "Overall Response". For RS, a supplemental qualifier for "Reason Response Not Evaluable" (RSREASNE) has been used in some examples. However, in TR, the current modeling for not evaluable results follows the approach where TRSTAT is populated with "NOT DONE".  TRSTAT = "NOT DONE" means that the result was missing and the reason that the result was missing is mapped to TRREASND. The TRREASND values cover reasons such as that the scan was not performed or other reason that the tumor was not evaluable. This modeling does not always explicitly or consistently identify the tumors which are not evaluable but some of the examples include "NOT EVALUABLE" in TRREASND.

In the case where there is an independent review and the investigator identified tumors at baseline but the independent assessor found no evidence of disease (NED) at baseline, this can be represented by an RS record with RSTEST = "Overall Response", RSORRES = "NED", and RSSTRESC = "NED".

Per the SDTMIG, "When a clinical classification result is based on multiple procedures/scans/images/physical exams performed on different dates, RSDTC may be derived." Some sponsors assign or derive the date of overall response (RSDTC) associated with the disease assessment timepoint, and some sponsors collect the date of overall response (RSDTC) associated with the disease assessment timepoint and provide instructions to the site/vendor on how to populate it when there is more than 1 evaluation date. In the RS examples, when there are multiple dates of evaluation performed as part of the disease assessment timepoint, RSDTC is populated using the following common convention (either entered by the investigator, assigned by the sponsor, or assigned by an independent assessor): For responses of CR (Complete Response) and PR (Partial Response), the latest date associated with the evaluation is used; for responses of PD (Progressive Disease), the earliest date associated with an evaluation is used. This convention is based on the following concepts: When evaluating measures of disease burden over time (e.g., duration of response, progression-free survival), it is generally considered conservative to assign a negative outcome (e.g., disease progression) to the date the assessment is initiated but not to assign a date to a positive response until the assessment is completed. For SD (Stable Disease), different conventions have been used in industry and the rationale varies. For the examples in this supplement, the convention used for SD is to assign the earliest date to RSDTC. However, picking the latest date is acceptable and a common approach. For NE (Not Evaluable), a date associated with the assessment is required but it generally does not impact the analysis, so either the earliest or latest date can be used.

In TR, TRDTC is populated with the date of tumor/lesion assessment/measurement (e.g., date of the scan, image, physical exam). In the examples, the summary results (e.g., Sum of Diameter) do not have TRMETHOD and TRDTC associated with them. However, if an appropriate TRDTC is available, it can be populated. In TR, scans which are not performed do not have TRDTC populated.

Note:

1. Examples in this supplement are based on assumptions about data collection forms. For example, the collection of split and merged tumors may be collected in different ways depending on the form design and system limitations. The preferred SDTM data representation is shown in the examples.
2. Examples in this supplement assume that no protocol modifications were made to RECIST 1.1.

4 Examples for the Oncology Disease Response Domains Model for RECIST 1.1

4.1 Example 1: Investigator Assessment of Subject with Lymph Nodes as Target Lesions

This RECIST 1.1 example shows investigator response (RS) data and the underlying tumor identification (TU), tumor results (TR), and procedure (PR) data. In this example, lymph nodes are selected as target lesions. The tumor identifier is in TULNKID and matches TRLNKID in the TR domain. The image identifiers are in TUREFID and TRREFID in the TU and TR domains, respectively, and they match the image identifiers in PRREFID in the PR Domain.

4.1.1 RS Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the RS domain. This example shows the data for 1 subject collected at the week 6, week 12, and subsequent 8-week follow-up visits. Note: Sponsors may include the new lesion indicator test (RSTESTCD = "NEWLIND") at every timepoint if the new lesion yes/no question is part of the data collection.

4.1.2 TU Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the TU domain. This example shows the data for 1 subject at screening and at weeks 36 and 44, where new lesions are identified. In this example, lymph nodes are selected as target tumors. In RECIST 1.1, target lymph node tumors are measured in the longest perpendicular (TRTESTCD = "LPERP") in the TR domain. The image identifier is in TUREFID and matches a PRREFID in the PR domain.

4.1.3 TR Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the TR domain. This example shows the data for 1 subject collected at screening, week 6, week 12, and subsequent 8-week follow-up visits. This example shows measurements (i.e., longest perpendicular) of lymph nodes as well as measurements of other non-lymph node target tumors (i.e., longest diameter). In this example, when TRTEST = "Lymph Node State" and TRORRES = "NON-PATHOLOGICAL", it indicates that the lymph node tumor is less than 10 mm. When a target lymph node tumor has longest perpendicular greater than or equal to 10 mm, the assessment is represented with TRTEST = "Lymph Node State" and TRORRES = "PATHOLOGICAL". The image identifier is in TRREFID and matches a PRREFID in the PR domain.

4.1.4 PR Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the PR domain. This example shows the data for 1 subject collected at screening, week 6, week 12, and subsequent 8-week follow-up visits. The example shows an MRI of the head and neck (PRTRT = "MRI" and PRLOC = "HEAD AND NECK"), a CT scan of the chest (PRTRT = "CT SCAN" and PRLOC = "CHEST"), and medical photography of the skin of the trunk (PRTRT = "PHOTOGRAPHY" and PRLOC = "SKIN OF THE TRUNK") at each disease assessment timepoint. The image identifier is in PRREFID.

4.2 Example 2: Split and Merged Tumors

This example illustrates investigator response (RS) data and the underlying tumor identification (TU), and tumor results (TR) data. This example shows the preferred representation of split (fragmented) and merged (coalesced) tumors. The tumor identifier is in TULNKID and matches TRLNKID in the TR domain. 

4.2.1 RS Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the RS domain. This example shows response data for 1 subject collected at the week 8 and subsequent 8-week follow-up visits: target response (RSTESTCD = "TRGRESP"), non-target response (RSTESTCD = "NTRGRESP"), and overall response (RSTESTCD = "OVRLRESP"). RSLNKGRP is used to link the overall response to the underlying assessments in the TR domain which support the overall response at the assessment.

4.2.2 TU Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the TU domain. This example shows the data for 1 subject collected at screening and subsequent visits. The example shows the preferred representation of split (fragmented) and merged (coalesced) tumors. This example also shows TUREFID used to represent the image identifier and TUSPID used to represent a sponsor's internal tracking identifier.  

4.2.3 TR Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the TR domain. This example shows the data for 1 subject collected at screening, week 6, and subsequent 8-week follow-up visits. The example shows the preferred representation of split (fragmented) and merged (coalesced) tumors. This preferred approach provides transparent traceability to parent lesions. This example shows how to represent data on a single row when a target tumor becomes too small to measure. Per RECIST 1.1, the original results would be "TOO SMALL TO MEASURE" and the standard results (TRSTRESN and TRSTRESC) would be "5". In addition, this example represents a situation where a result was not provided because the image was obscured and the tumor was not assessable.

This example shows a situation where the investigator provided the summary values for target lesions as a group (TRGRPID = "TARGET"). The TRTESTCD = "SUMDIAM" is used for the assessments. Note that sponsors should not derived these values if they were not part of the data capture.

4.3 Example 3: Scans Either Not Done or Not Evaluable

This example shows investigator response (RS) data and the underlying tumor identification (TU) and tumor results (TR) data.

4.3.1 RS Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the RS domain. This example shows the data for 1 subject collected at week 8 and subsequent 8-week follow-up visits. The overall response (RSTESTCD = "OVRLRESP") has RSORRES = "NE" because the scans of the target and non-target lesions were not done.

4.3.2 TU Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the TU domain. This example shows the identification of target and non-target tumors at screening. The tumor identifier is in TULNKID and matches TRLNKID in the TR domain. This example shows a subject with 4 target lesions (TULNKID = "T01"-"T04") and 3 non-target lesions (TULNKID = "NT01"-"NT03") identified at screening. The subject did not have any new lesions. 

4.3.3 TR Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the TR domain. This example represents a situation where a result was not provided because the scan was not performed.

4.4 Example 4: Accepted Record Flag for Multiple Independent Assessors

This example shows independent assessor response (RS) data and the underlying tumor identification (TU), and tumor results (TR) data. This example shows the use of the acceptance flag to identify records when multiple assessments are performed by independent assessors (radiologist 1 and radiologist 2) for the same timepoint.

4.4.1 RS Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the RS domain. This example shows the data for 1 subject collected at week 8 and week 16. 

The acceptance flag (RSACPTFL) is used to identify records when multiple assessments are performed by independent assessors (e.g., radiologist 1, radiologist 2) for the same timepoint. RSACPTFL indicates the "selected" response when more than 1 reading is performed. In some instances, a vendor may only supply a single response record for the selected response, either because there was adjudication or because the vendor selected the appropriate response. This alternative approach would result in only the selected response records in the RS domain. It is not expected that the RSACPTFL flag would be populated by the sponsor; this type of record selection or censoring would be part of the analysis dataset.

This example also includes the situation where the best response assessment is provided by the data capture rather than derived.

This example uses RSNAM (name of the vendor), RSEVAL (role of the person providing the assessment; e.g., "INDEPENDENT ASSESSOR"), and RSEVALID (specific type of medical reviewer who performed the independent assessment). In this case, 2 radiologists independently reviewed the scans (RSEVALID = "RADIOLOGIST 1", "RADIOLOGIST 2"). The acceptance flag (RSACPTFL) indicates that the "RADIOLOGIST 2" records are the accepted records (rows 9-16). The best overall response (RSTESTCD = "BESTRESP") is provided by the vendor in the data transfer in rows 4, 8, 12 and 16.

4.4.2 TU Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the TU domain. This example shows the use of the acceptance flag (TUACPTFL) to identify records when multiple assessments are performed by independent assessors (e.g., radiologist 1, radiologist 2) for the same timepoint. The acceptance flag indicates records which are chosen by an adjudicator. Not all sponsors have both sets of independent reviews exported (i.e., only accepted records are transferred). It is not expected that TUACPTFL would be populated by the sponsor; this type of record selection or censoring would be part of the analysis dataset. The tumor identifier is in TULNKID and matches TRLNKID in the TR domain. 

4.4.3 TR Domain Model

The following table shows the terminology used to implement RECIST 1.1 in the TR domain. This example shows the use of the acceptance flag to identify records when multiple assessments are performed by independent assessors (e.g., radiologist 1, radiologist 2) for the same timepoint. It is not expected that TUACPTFL would be populated by the sponsor; this type of record selection or censoring would be part of the analysis dataset.

In this example, the independent assessor provided the summary values for target lesions as a group (TRGRPID = "TARGET"). The TRTESTCDs used for the assessments are: "SUMDIAM", "ACNSD", "PCBSD", and "PCNSD". Note that sponsors should not derive these values if they were not part of the data capture or not provided in the transfer from the vendor.