Poor
Mostly Aligned
Patient Risk:
Low
Summary
Partially aligns on general indication statements supported by Sections 1.1–1.3. However, most additional claims are interpretive/clinical-reasoning statements (e.g., definitions of “symptoms,” statements about failure to improve, reasons for incomplete symptom control, and reasons patients may think the drug “failed”) that are not supported by the provided label excerpts. The label excerpts also do not support broad assertions about ‘did not broadly fail’ or statistically significant improvements across ‘standard symptom and disease-activity outcomes’ in psoriatic arthritis and axial disease as phrased.
Category Scores
Accurate Statements
Cosentyx (secukinumab) is indicated for plaque psoriasis, psoriatic arthritis, and active ankylosing spondylitis (AS) (with pediatric age ranges specified in the label).
Supported by Sections 1.1 (plaque psoriasis), 1.2 (psoriatic arthritis), and 1.3 (active ankylosing spondylitis) in the provided label excerpts.
Unsupported Statements
In plaque psoriasis, “symptoms” refers to measures like skin lesion severity.
The label excerpts define PASI and IGA severity outcomes and mention patient-reported outcomes for itching/pain/scaling, but they do not define the term “symptoms” as used by the AI statement.
In psoriatic arthritis and axial disease, “symptoms” refers to measures like joint pain, swelling, function, or activity scores.
The label excerpts describe specific endpoints/components (e.g., ACR components, ASAS20/40 components, HAQ-DI), but do not define “symptoms” as a general mapping to these measures.
In the labeled indications, the clinical development record indicates Cosentyx did not broadly fail to improve symptoms.
While trials show improvements on defined endpoints, the label excerpt does not contain a statement equivalent to the AI’s broad interpretive claim about ‘did not broadly fail to improve symptoms.’
In trials for plaque psoriasis, secukinumab showed statistically significant improvement on standard symptom and disease-activity outcomes compared with control.
The excerpts provide endpoints (PASI 75, IGA clear/almost clear, PASI 90, Psoriasis Symptom Diary itching/pain/scaling) and response proportions, but they do not explicitly state ‘statistically significant’ improvement phrased as ‘standard symptom and disease-activity outcomes’ across those categories.
In trials for psoriatic arthritis and axial spondyloarthritis, secukinumab showed statistically significant improvement on standard symptom and disease-activity outcomes compared with control.
The excerpts include efficacy language (e.g., ‘greater clinical response’ / ‘significantly inhibited progression’ / ‘resulted in significant improvements’) but do not support the AI’s broad generalization that this was true for ‘standard symptom and disease-activity outcomes’ as a set for all cited indications and outcomes.
Even when a drug works overall in trials, some patients do not respond well.
The provided label excerpts show non-responder concepts/endpoints (e.g., placebo escape and non-response definitions), but do not explicitly support the generalized statement in the phrasing used by the AI.
Even when a drug works overall in trials, some patients may lose response over time.
The excerpts include maintenance proportions for responders, but do not explicitly support the specific generalized statement that patients ‘may lose response over time’.
In practice, incomplete symptom control can occur due to differences in baseline disease severity.
The label excerpt contains baseline disease characteristics and subgroup discussion, but does not provide practice-level causal reasoning statements like this.
In practice, incomplete symptom control can occur due to prior exposure to other biologics.
The label excerpts discuss biologic-naïve vs biologic failures and prior anti-TNFα exposure, but do not state this as a practice-level reason for incomplete control.
In practice, incomplete symptom control can occur due to adherence and dosing schedule.
No adherence/dosing-schedule causality is provided in the provided label excerpts.
In practice, incomplete symptom control can occur due to comorbidities that complicate symptom interpretation.
No comorbidity-based causal reasoning for incomplete symptom control is provided in the provided label excerpts.
Some patients or subgroups may experience inadequate improvement despite the drug not failing across the board.
The label excerpts provide subgroup analyses/no differences (for PsO subgroups) and response/maintenance data, but do not support the AI’s specific interpretive phrasing about ‘not failing across the board’ and ‘inadequate improvement’ as a general proposition.
People may think Cosentyx failed if a specific trial endpoint (or a particular measure) does not improve in a subgroup while other endpoints do.
The label excerpts do not describe how people interpret endpoints across subgroups (no patient/observer interpretation statements).
People may think Cosentyx failed if an individual patient has lack of response despite being on therapy.
The label excerpt does not contain statements about patient perception/interpretation of individual response as ‘failure.’
People may think Cosentyx failed if another biologic performs better on a certain endpoint in a different trial.
The label excerpts do not discuss comparative performance across different biologics/trials in the way described.
Contradictions
Important Omissions
No AI response included label-supported safety details (warnings/precautions, contraindications, adverse reactions, monitoring) or dosing/administration instructions, which may be material depending on the user’s intended audit focus beyond indications and clinical study outcomes.
Importance:
Moderate
Safety Assessment
Potential Patient Risk:
Low
The provided AI claims are largely descriptive/interpretive about efficacy interpretation and do not include dosing, contraindications, or safety-critical guidance. Indication alignment exists, but multiple unsupported interpretive statements could mislead stakeholders about evidence interpretation.
Regulatory Assessment
| On Label |
No |
| Off-label Discussion |
No |
| Promotes Unapproved Use |
No |
| Hallucination Risk |
Medium |
Recommendation
Mostly Aligned
Primary Issue
Most statements are not supported by the provided label excerpts; several use broad, interpretive language (e.g., defining ‘symptoms,’ discussing reasons for incomplete control, and statements about how people may judge ‘failure’) that are absent from the label.
Suggested Improvement
Restrict claims to label-supported endpoints and descriptions (e.g., PASI 75/90, IGA, Psoriasis Symptom Diary outcomes, ACR20/50/70, ASAS20/40 components, HAQ-DI) and avoid introducing practice-level causal explanations or definitions (e.g., what ‘symptoms’ means) unless the label explicitly states them.