Gutierrez L, Booth H, Dedman D, Crellin E, Kaye JA, Franzoni C, Arana A, Schmitt-Egenolf M, Sundstrom A, Bystrom C. Case validation of cutaneous lymphoma to minimize protopathic bias. Poster presented at the 35th Annual ICPE Conference; August 27, 2019. Philadelphia, PA. [abstract] Pharmacoepidemiol Drug Saf. 2019 Aug 20; 28(S2):645.


BACKGROUND: Protopathic bias (PB) occurs when a drug is prescribed for initial signs/symptoms of an outcome not yet diagnosed, reflecting a reversal of cause and effect. PB may be of particular concern when studying outcomes with long latencies, such as cancer. It is commonly controlled by including a time lag in sensitivity analyses. In a multi-country study on the association of topical tacrolimus and pimecrolimus with cutaneous lymphoma (CL), PB was a concern because CL may present symptoms resembling atopic dermatitis and therefore may be treated with the study medication.

OBJECTIVES: To evaluate the potential for PB by acquiring further information on the medical history of CL cases, specifically to assess whether treatment with the study medications was initiated for symptoms that were compatible with early manifestations of CL.

METHODS: In Sweden, researchers reviewed hospital medical records of CL cases identified through the Swedish national cancer registry. In the United Kingdom, researchers from the Clinical Practice Research Datalink (CPRD) sent questionnaires to general practitioners (GPs) with CL cases. Information extracted included date of CL diagnosis; date of start of symptoms; location and extent of CL; biopsy results; and skin conditions such as psoriasis or atopic dermatitis, including location, extent, and date of onset.

RESULTS: In total, 29 (CPRD) and 73 (Sweden) potential CL cases were identified. In CPRD, GP questionnaires were sent for all 29 potential cases identified, and 19 have been returned to date. One did not provide additional information. Diagnosis was confirmed in 13 cases, and date of CL diagnosis was identified earlier by GPs than by case screening algorithm in four cases. Six cases had a prior history of skin condition on the same location as CL, one did not, and six did not provide enough information. In Sweden, it was possible to request information for 65 potential cases. To date, 48 medical records have been received.

CONCLUSIONS: Based on reviewed cases, the additional clinical information modified the outcome diagnoses or the date of symptoms initiation in half of the patients. Nearly half of the confirmed cases were previously diagnosed with benign skin conditions that share signs and symptoms with CL. Reverse causation could then introduce bias in the study results.

Share on: