Continuing in treatment as a form of selection bias
Panic patients who continued treatment (N = 12) did not have symptom scores significantly different from those of patients who declined or discontinued treatment (N = 12) but they did have lower pretreatment MHPG levels. Continuing in treatment may itself cause selection bias for biochemical variables under study
Edlund, M., & Swann, A. C. (1989). Continuing in treatment as a form of selection bias. American Journal of Psychiatry, 146(2), 254-256.