Generalizing from a controlled trial: the effects of patient preference versus randomization on the outcome of inpatient versus outpatient chronic pain management

Patients accepting randomization in a randomized controlled trial (RCT) may not be representative of the clinical population from which they are drawn, calling into question the generalizability of study findings. Comparison of randomized and non-randomized inpatient and outpatient samples at baseline and in treatment outcomes up to one year was made to determine whether the findings of the RCT generalized to non-randomized patients in the same treatment program. One hundred and twenty one patients with intractable pain, randomized between inpatient, outpatient and waiting list control, were compared with 128 who elected for either inpatient or outpatient treatment. Treatment was a group-based multidisciplinary cognitive-behavioral treatment program aimed at enabling patients to return to more normal function despite persistent pain, delivered to mixed groups of randomized and elective patients, and outcome was measured by physical performance, pain impact on function, mood, and drug use. Agreement to randomization was a function of travelling distance from home to hospital. Non-randomized patients largely resembled their randomized counterparts before and after treatment. In order to indicate the clinical significance of results, analyses were conducted using numbers needed to treat (NNTs). NNTs estimate the number of patients required in the treatment condition for one of them to achieve the specified outcome who would not have achieved it in the comparison condition. Across a range of measures at one month follow-up, comparison of inpatients with outpatients gave NNTs between 2.3 and 7.5, and comparison of inpatients with waiting list controls gave NNTs between 2.3 and 3.6. At one year inpatients showed greater likelihood than outpatients of maintaining these treatment gains.