A: Daley BM, Shuster S. Effect of aspirin on pruritus.

BMJ 1986;293:907

1. Identify the criteria for patient selection. It is unclear how they were selected, but the 13 patients included suffered from pruritus (itching) due to skin disorders such as atopic or chronic discoid eczema, psoriasis, primary biliary cirrhosis. No patient had a history of peptic ulceration.

2. Identify the treatment(s) being evaluated 900 mg oral aspirin at night vs no treatment.

3. Identify the main measure(s) of patient outcome Itch (measured objectively as nocturnal scratch using limb movement meters, and subjectively on visual analogue scale). The measures are poorly explained.

4. How were patients allocated to treatment? There was no random allocation of treatment. Looks as if each patient was acting as own control (cross-over design), with systematic allocation by days of the week.

5. Was appropriate blinding/masking/allocation concealment used? No blinding at ‘trial’ entry, no blinding at assessment.

6. Identify the main statistical finding(s) on which conclusions are based As in (inadequate) Table.

7. Overall, discuss whether any potential biases in the trial’s design are sufficient to affect the trial’s conclusions In general, a terrible study, rendering authors’ conclusions invalid. (See also correspondence).

B: Bloom SL, et al. Lack of effect of walking on labor and delivery.NEJM 1998;339:76-79

1. Identify the criteria for patient selection Women in spontaneous labour with uncomplicated pregnancies 36-41 weeks gestation. More details in 1st paragraph of 2nd column of page 76.

2. Identify the treatment(s) being evaluated Walking during first stage of labour compared to usual care (confined to bed). More details in 2nd paragraph of 2nd column of page 76.

3. Identify the main measure(s) of patient outcome Obstetric outcomes, not specified in detail under methods (so don’t know if they are just reporting outcomes selected post-hoc), but from ABSTRACT, ‘patients’ discomfort and outcomes’ and from RESULTS, length of labour, use of pain relief, mode of delivery, neonatal outcomes, and women’s choice for a future labour.

4. How were patients allocated to treatment? Other than being told in ABSTRACT and in 2nd paragraph of 2nd column of page 76, that it was a randomised trial with women randomly assigned, no further details are given.

5. Was appropriate blinding/masking/allocation concealment used? Although clearly women and care-givers could not be blind to allocated management after random allocation, some outcomes could have been blinded.

6. Identify the main statistical finding(s) on which conclusions are based As given in Tables and RESULTS section. Only p values. Confidence intervals would have been helpful.

7. Overall, discuss whether any potential biases in the trial’s design are sufficient to affect the trial’s conclusions Assuming ‘proper’ random allocation has taken place, conclusions seem ok.

C: Salam,MA, et al. Randomised comparison of ciprofloxacin suspension and pivmecillinam for childhood

shigellosis.Lancet 1998;352:522-527

1. Identify the criteria for patient selection Children aged 2-15 with dysentery but who had not received any suitable anti-microbial therapy. Further details in 1st of PATIENTS on page 522.

2. Identify the treatment(s) being evaluated Ciprofloxacin suspension vs pivmecillinam tablets.

3. Identify the main measure(s) of patient outcome Dysentery on day 3; 6 or fewer stools, no bloody-mucoid stools and no fever on day 5; and other outcomes as stated under Outcomes on page 523.

4. How were patients allocated to treatment? Computer-generated list of random numbers; double-dummy design. Drugs packaged by pharmaceutical company.

5. Was appropriate blinding/masking/allocation concealment used? Patients (and their families), and medical, nursing or laboratory staff, and researchers remained blinded during the study.

6. Identify the main statistical finding(s) on which conclusions are based As in text on pages 524-5, Tables 2 and 3 and Figure 2.

7. Overall, discuss whether any potential biases in the trial’s design are sufficient to affect the trial’s conclusions Unlikely to be serious biases such as to affect validity of conclusions.