help bciiandhelp bcibVersion 1.1 31-05-2009-------------------------------------------------------------------------------

Title

bcii-- Calculates the number needed to treat (NNT) and confidence intervals for patients improving (comparison between intervention and control group in a randomised controlled trial)bcib-- Calculates NNT and confidence intervals for patients benefiting (either improvements gained or deteriorations prevented) in a randomised controlled trial

Syntax

bcii#a #b #c #d[,options]

bcib#a #b #c #d #e #f #g #h[,options]

optionsDescription ------------------------------------------------------------------------- Mainlevel(#)set confidence level; default is prevailing setting (see creturn)

Descriptions

bciiestimates absolute risk reduction and number needed to treat for for the difference between the proportion of improving patients in the intervention group and the proportion of improving patients in the control group. It uses a method that is considered to have superior coverage properties to the conventional Wald method for calculating confidence intervals (Newcombe's Method 10); this may be important when reciprocally transforming absolute risk reductions to obtain numbers needed to treat (NNTs) as described by Bender, 2001. estimates NNTs for benefit; the difference between the proportion of improving patients minus the proportion of deteriorating patients in the intervention group and the proportion of improving patients minus the proportion of deteriorating patients in the control group. Newcombe's method 10 has been modified to incorporate these extra variance terms as described by Froud et al, 2009.

Options+------+ ----+ Main +-------------------------------------------------------------

levelset confidence level; it must lie between 0.1 and 99.9%, inclusive; default is the Stata level setting.

Remarks

bciicalculates confidence intervals for the NNT by reciprocal transformation of risk difference confidence intervals, from differences between two independent proportions (patients improving in the treatment group and in the control group)using Newcombe's Method 10. The confidence intervals are considered to have better coverage than the conventional Wald confidence interval and are less prone to aberrations following reciprocal transformation to NNT limits.The sequence of

#a,#b,#cand#dinbciicome from a contingency table of 'improvements' x group (see epitab). The additional#a,#b,#cand#dinbcibcome from a similar table of 'deteriorations' (see epitab).

Examples

. bcii 82 62 143 194

. bcii 82 62 143 194, level(90)

. bcib 82 62 143 194 5 12 220 243

ReferencesR. Froud, S. Eldridge, R.Lall, M.Underwood, Estimating NNT from continuous outcomes in randomised controlled trials: Methodological challenges and worked example using data from the UK Back Pain Exercise and Manipulation (BEAM) trial ISRCTN32683578.

BMC Health ServicesResearch2009 IN PRESS.R. Bender, Calculating confidence intervals for the Number Needed to Treat

Controlled clinical trials22-102-110, 2001.

R. G. Newcombe, Interval estimation for the difference between independent proportions: comparison of eleven methods.

Statistics inMedicine17:873-90, 1998.

AcknowledgementsCode for calculating Wilson intervals of the individual proportions (used in calculation of Newcombe's Method 10 confidence intervals) is adapted from

rdciiby Joseph Coveney, who adapted code fromciwiby Nicholas J. Cox. Thanks to my colleagues, S. Eldridge, R.Lall, and M.Underwood for their help and comments on the modifications to Newcombe's Method 10, and to Gordon Guyatt, Thomas Kottke, and Kamshwar Prasad, for their very useful comments on the accompanying paper; and to Mandy Hildebrandt and Michal Vaillant for their comments on an earlier version of the paper.

AuthorRobert Froud r.j.froud@qmul.ac.uk

Also seeManual:

[ST] epitab