OBJECTIVES:  The 21^stCentury Cures Act calls for the incorporation of real-world evidence (RWE) into the drug labeling process. Critics cite deficiencies of RWE as compared to randomized controlled trials (RCT). We aimed to assess whether subjectively reported best response to treatment differed from responses based on radiographic measurements of target lesions. METHODS:  Electronic case report forms (eCRF) were fielded to assess clinical response to systemic therapy in a rare malignancy. A comparison of subjectively reported best response with radiographic measurements of target lesions was performed to assess responses. RESULTS:  Fifteen physicians with experience treating this malignancy participated in a CRF-based analysis of patterns of care. Treatment response for each line of therapy was collected for each of the 59 patients via CRF. For 9 patients (15%) with reported partial remission (PR), CRFs were augmented with bi-dimensional measurements of sentinel tumors from pre-treatment and best response radiographs (reports). Treatment response was calculated using sum of diameters according to RECIST and compared to the corresponding physician-reported response. Measurement-based response revealed 4 PR, 2 stable disease, and 3 progressive disease, and a concordance of 44.4% (95% CI: 15.3%, 77.4%) between response assessment methods. Reasons cited for variance include availability of previous scans for comparison, inconsistency in target lesions imaged, bone metastases difficult to follow, and use of diameter vs SUV (standardized uptake value) on positron emission topography (PET) scans. CONCLUSIONS:  Subjective assessments of response collected via manual chart extraction, electronic medical record (EMR) review via natural language processing, or CRF may be problematic and limit the potential role of RWE in drug label expansion. The collection of target lesion measurements by CRF presents an attractive alternative to elevate the quality and accuracy of clinical response assessment in oncologic patients, but limitations exist.