Weekly PV Briefing — Cosentyx

Week ending 9 August 2026

Scope: This review monitors safety-relevant patterns in third-party AI-generated answers. It is not an adverse-event reporting channel and does not establish a confirmed safety signal.

1. Signal summary

There were 27 cross-response inconsistency flags across 20 questions and 140 answer variants, versus 26 flags across 20 questions and 60 responses last week—a slight rise of 1 flag, with response volume more than doubling.

Breakdown of inconsistency flags:

• Contraindication: 7
• Infection risk: 7
• Allergic reaction: 5
• IBD: 5
• Pregnancy/lactation: 2
• Boxed warning: 1

No thyroid-tumor, pancreatitis, or hypoglycemia inconsistencies were identified.

Raw category tallies also increased this week, including infection risk (89 vs 37), contraindication (27 vs 9), allergic reaction (26 vs 13), pregnancy/lactation (16 vs 9), and boxed warning (8 vs 3). These are frequency counts, influenced by the higher response volume, and should not be interpreted as incidence.

2. Highest-priority inconsistencies

1. Boxed-warning question — plaque psoriasis comparison with Humira (brand; Q226):
   One answer mentioned a boxed warning, while six omitted it; one answer also mentioned allergic reaction and contraindication, while six omitted each. Conflicting statements about whether a boxed warning exists require immediate verification against the current approved prescribing information; inaccurate warning classification could materially distort perceived risk.

2. Pregnancy/lactation and infant live-vaccine question (PI; Q225):
   One answer used contraindication language; six omitted it. Although pregnancy guidance is not necessarily an absolute contraindication, conflating “contraindicated,” “not recommended,” and benefit–risk language could lead to inappropriate treatment discontinuation or continuation and confusion about infant vaccination precautions.

3. Contraindications in active infection (PI; Q221):
   Three answers mentioned contraindication-related language; four omitted it. Omission may leave a patient with an active infection without a clear prompt to defer treatment or contact the prescriber.

4. Administration/device question (PI; Q218):
   Allergic-reaction and infection-risk information appeared in five answers but was omitted in two. In a practical self-administration context, inconsistent safety checks could reduce recognition of when an injection should not be given or when medical advice is needed.

5. Marketing versus clinical-guidance question (brand; Q235):
   Pregnancy/lactation was mentioned in one answer and omitted in six; contraindication was mentioned in four and omitted in three. Consumer-facing safety omissions are important because these answers may be read without clinician supervision.

3. Recurring or worsening themes

Across the three-week history, infection risk, IBD, overclaim language, contraindication, and allergic-reaction content recur every week. Current raw tallies are higher than both 2 August and, for several categories, 26 July; however, the intervening lower-volume week and increased response count limit interpretation. The most concerning recurring issue is inconsistent presentation of infection and contraindication information across both PI and consumer-style questions.

4. Escalation recommendation

Flag to a human safety/medical-information reviewer this week. The conflicting boxed-warning content and repeated contraindication omissions warrant documented verification against the current local PI and review of the underlying answer variants. This is not a confirmed safety signal or AE report. Other informational gaps should be logged and monitored.

5. Recommended actions

1. Verify the boxed-warning, infection, hypersensitivity, pregnancy/lactation, and infant live-vaccine content against the current approved PI.
2. Perform documented review of Q226 and Q225, including exact answer text and jurisdiction assumed.
3. Review Q221 and Q218 for potentially actionable patient instructions and unsafe omission patterns.
4. Add dosing-regimen discrepancies observed in PI answers to the medical/regulatory review queue.
5. Continue weekly monitoring, separating raw tally changes from response-volume effects.