Choi CT response criteria for GIST treated with targeted therapy

{
  "categories": [
    {
      "outcome_code": "CR",
      "label": "complete_response",
      "defining_state": "disappearance_of_all_lesions_and_no_new_lesions"
    },
    {
      "outcome_code": "PR",
      "label": "partial_response",
      "defining_state": "size_decrease_at_least_10_percent_OR_density_decrease_at_least_15_percent_without_progression_override"
    },
    {
      "outcome_code": "SD",
      "label": "stable_disease",
      "defining_state": "neither_CR_PR_nor_PD_and_no_symptomatic_deterioration_from_tumor_progression"
    },
    {
      "outcome_code": "PD",
      "label": "progressive_disease",
      "defining_state": "qualifying_size_increase_without_density_response_OR_new_lesion_OR_new_or_enlarging_intratumoral_nodule"
    }
  ],
  "applicability": {
    "use_for": "Response assessment of measurable gastrointestinal stromal tumor, originally metastatic GIST treated with imatinib, on serial contrast-enhanced CT.",
    "classification_unit": "one_patient_at_one_response_timepoint_using_a_prespecified_target_set",
    "required_inputs": [
      "confirmed_or_clinically_established_GIST",
      "systemic_targeted_therapy_and_start_date",
      "baseline_or_reference_CT",
      "comparable_follow_up_CT",
      "target_lesion_set",
      "sum_of_longest_diameters",
      "target_lesion_attenuation",
      "new_lesion_status",
      "intratumoral_nodule_status",
      "nonmeasurable_disease_status"
    ],
    "outside_scope": [
      "non_GIST_tumors_without_validation",
      "pathologic_response",
      "postoperative_recurrence_risk",
      "GIST_stage",
      "mitotic_or_molecular_risk_group",
      "treatment_selection_from_response_code_alone"
    ]
  },
  "acquisition_and_comparability_gate": {
    "baseline_staging_context": "A standardized contrast-enhanced CT establishes burden and target lesions; dedicated GIST protocols commonly include noncontrast, arterial liver and portal-venous abdominal-pelvic phases at baseline.",
    "response_phase": "Measure lesion size and attenuation on the same comparable portal-venous phase. During therapy, consensus protocols commonly use noncontrast plus portal-venous imaging.",
    "consistency_requirements": [
      "same_contrast_phase",
      "comparable_contrast_dose_and_timing",
      "comparable_scanner_and_reconstruction_when_possible",
      "same_target_lesions",
      "same_measurement_plane_and_ROI_method"
    ],
    "failure_rule": "If contrast phase, target identity or attenuation technique is not comparable, disable the density branch and state the limitation; do not manufacture a 15-percent density response.",
    "MRI_PET_boundary": "MRI can complement liver assessment and FDG PET/CT can clarify early or morphologically inconclusive response, but their signal or uptake changes are not inserted into the CT attenuation thresholds."
  },
  "target_and_measurement_model": {
    "size_metric": "Sum of the longest diameters of the prespecified target lesions, measured consistently on axial images according to the declared RECIST target-selection implementation.",
    "contemporary_target_set": "A common RECIST 1.1 implementation uses at most two target lesions per organ and five total. The original Choi publication predates RECIST 1.1; always record the target-selection version and never change the target set silently.",
    "attenuation_metric": "Target-lesion attenuation in Hounsfield units on portal-venous CT, measured at the level of the longest diameter with a reproducible ROI covering as much viable lesion as possible and including an enhancing rim when present while excluding adjacent normal structures.",
    "aggregation_rule": "Record each target lesion's diameter and attenuation plus the patient-level SLD and protocol-defined mean lesion density. Do not mix mean, median, single-lesion or summed attenuation across timepoints.",
    "size_decrease_percent_formula": "100 * (reference_SLD - follow_up_SLD) / reference_SLD",
    "size_increase_percent_formula": "100 * (follow_up_SLD - comparison_SLD) / comparison_SLD",
    "density_decrease_percent_formula": "100 * (reference_MLD - follow_up_MLD) / reference_MLD",
    "reference_rule": "State whether baseline, smallest prior burden or another trial-defined reference is used. The response code is not reproducible without the comparison date and reference convention.",
    "attenuation_QC": "The German consensus recommends considering an absolute attenuation difference of at least 10 HU because manual ROI variability can make a 15-percent change unstable; this is an implementation quality safeguard, not a replacement for the original 15-percent rule."
  },
  "original_Choi_2007_algorithm": [
    {
      "priority": 1,
      "if": "new_lesion_or_new_or_enlarging_intratumoral_nodule",
      "output_code": "PD"
    },
    {
      "priority": 2,
      "if": "all_lesions_disappeared_and_no_new_lesion",
      "output_code": "CR"
    },
    {
      "priority": 3,
      "if": "size_decrease_percent_at_least_10_OR_density_decrease_percent_at_least_15_AND_no_new_lesion_AND_no_obvious_progression_of_nonmeasurable_disease",
      "output_code": "PR"
    },
    {
      "priority": 4,
      "if": "size_increase_percent_at_least_10_AND_density_decrease_percent_less_than_15",
      "output_code": "PD"
    },
    {
      "priority": 5,
      "if": "does_not_meet_CR_PR_or_PD_AND_no_symptomatic_deterioration_attributed_to_tumor_progression",
      "output_code": "SD"
    }
  ],
  "boolean_and_override_rules": {
    "PR_OR_rule": "The original Choi PR threshold is size decrease of at least 10 percent OR attenuation decrease of at least 15 percent, provided progression exclusions are satisfied. Requiring both silently changes the system.",
    "size_growth_PD_AND_rule": "A size increase of at least 10 percent is PD only when the lesion does not simultaneously meet the density-response threshold; treatment-related myxoid or hemorrhagic change can enlarge a responding mass.",
    "nodule_within_mass_override": "A new or enlarging enhancing or hyperattenuating intratumoral nodule within an otherwise hypoattenuating treated mass is PD even when overall lesion size is stable or reduced.",
    "new_lesion_override": "A credible new GIST lesion is PD. Confirm an equivocal focus with prior imaging, another phase or follow-up rather than converting artifact or a benign lesion into progression.",
    "nonmeasurable_guard": "Obvious progression of nonmeasurable disease prevents PR. Describe the involved site instead of relying only on target measurements."
  },
  "variant_boundary": {
    "original_Choi": "Uses the OR rule for PR and the conditional size-growth rule for PD.",
    "modified_Choi_implementations": "Some studies or consensus implementations add an absolute 10-HU quality floor or require concurrent size and density change. Identify that protocol as modified Choi and never report its result as if it were the original 2007 definition.",
    "RECIST_boundary": "RECIST 1.1 is size-dominant and uses different PR and PD thresholds. Return Choi and RECIST separately when both are requested; do not merge their category logic.",
    "GIST_risk_boundary": "Choi response is unrelated to pathologic recurrence-risk systems based on primary site, size, mitotic count and rupture."
  },
  "morphology_and_pitfall_checks": [
    "Record transition from enhancing heterogeneous tissue to homogeneous hypoattenuation even when size changes little.",
    "Check for myxoid degeneration, necrosis, hemorrhage or cystic change that can alter size or attenuation and impair ROI comparability.",
    "Search every treated mass for a nodule-within-a-mass pattern and not only for global diameter growth.",
    "Check arterial-phase liver images when available because hypervascular metastases can be occult or isodense on portal-venous images.",
    "Use PET or short-interval expert adjudication for a technically valid but biologically discordant CT result when the answer will alter care."
  ],
  "management_and_risk_context": {
    "management_rule": "A Choi category is an imaging response assessment, not an instruction to continue, stop or change a TKI. Integrate treatment duration, adherence, toxicity, mutation, symptoms, resectability and multidisciplinary oncology review.",
    "progression_rule": "A technically confirmed PD pattern should be communicated promptly for oncology review, but the agent must not autonomously select a new drug, dose or operation.",
    "risk_rule": "The original study correlated the proposed response with FDG PET and time to progression in a limited single-institution cohort. The four labels do not provide an individual survival or resistance probability."
  },
  "output_contract": [
    "GIST_diagnosis_therapy_and_response_date",
    "baseline_and_comparison_dates",
    "CT_protocol_phase_and_comparability",
    "target_lesions_with_site_longest_diameter_and_HU",
    "SLD_and_MLD_reference_follow_up_and_percent_changes",
    "new_lesions_nonmeasurable_progression_and_nodule_within_mass",
    "original_or_modified_Choi_protocol",
    "CR_PR_SD_PD_or_unclassifiable_with_reason",
    "parallel_RECIST_result_only_when_separately_calculated",
    "no_autonomous_treatment_or_numeric_prognosis"
  ],
  "missing_input_behavior": [
    "If no pretreatment or valid reference examination exists, do not calculate a Choi response category.",
    "If portal-venous timing or attenuation ROI is not comparable, classify only from supported morphology and size while stating that the density-response branch is unavailable.",
    "If the target set or comparison reference is unknown, return unclassifiable and request those inputs rather than guessing percentage change.",
    "If a possible new lesion or intratumoral nodule is equivocal, flag possible progression and request confirmation instead of finalizing PD from an artifact-prone focus.",
    "If symptomatic deterioration is unknown in an otherwise SD-range scan, report radiographic non-CR/non-PR/non-PD and preserve the missing clinical qualifier."
  ],
  "interpretation_limits": [
    "Choi was developed in metastatic GIST treated with imatinib and should not be generalized to every tumor or therapy without validation.",
    "Attenuation is acquisition-dependent and a percentage alone can be unstable when baseline HU is low or technique differs.",
    "The criteria assess response at a timepoint and do not establish pathologic complete response, cure, resectability or long-term treatment benefit."
  ],
  "supporting_sources": [
    {
      "role": "primary_criteria",
      "citation": "Choi et al. J Clin Oncol. 2007;25:1753-1759",
      "doi": "10.1200/JCO.2006.07.3049",
      "pmid": "17470865"
    },
    {
      "role": "standardized_imaging_consensus",
      "citation": "Kalkmann et al. Cancer Imaging. 2012;12:126-135",
      "doi": "10.1102/1470-7330.2012.0013",
      "pmcid": "PMC3362866"
    },
    {
      "role": "current_GIST_guideline_context",
      "citation": "2023 GEIS Guidelines for gastrointestinal stromal tumors",
      "doi": "10.1177/17588359231192388",
      "pmcid": "PMC10467260"
    }
  ],
  "source_locator": "Choi et al. 2007, PMID 17470865, DOI 10.1200/JCO.2006.07.3049, proposed CT response criteria and CT-PET correlation; Kalkmann et al. 2012, PMC3362866, CT protocol, Therapy response assessment, Table 2, measurement technique and pitfalls; 2023 GEIS guideline, PMC10467260, diagnostic CT recommendations and current response-assessment context."
}