Article type
Year
Abstract
Background: The number needed to treat (NNT) has gained much attention in the past years as a useful way of reporting the results of randomized clinical trials (Bender, 2005). Due to some undesirable statistical properties, the NNT statistic was criticized by some authors or even rejected (Hutton, 2000). Nevertheless, NNT measures have been applied to a number of research areas involving the development of more sophisticated techniques to calculate and interpret NNTs in medical research. Recently, different impact numbers were suggested, which offer a broader application of NNT statistics, especially in epidemiological research (Attia et al., 2002; Heller et al., 2003).
Objectives: This paper gives an overview about various impact numbers, comparing them with classical epidemiological effect measures and the original NNT statistic.
Methods and Results: Recent methodological suggestions regarding NNT statistics are summarized and compared. Relations to well-known effect measures of classical epidemiology are pointed out. The advantages and drawbacks of the new impact numbers are discussed. One disadvantage is that the new impact numbers are proposed without appropriate methods for interval estimation. Application areas are indicated, which require more research before useful NNT measures can be calculated.
Conclusions: As with the original NNT statistic, the new impact numbers are useful as reporting tool. However, more research is required to calculate meaningful confidence intervals. Moreover, in complex study situations the NNT statistics and impact numbers lose the advantage of simplicity so that deeper scientific knowledge is required to interpret the results correctly.
References:
Attia, J., Page, J., Heller, R.F. & Dobson, A.J. (2002): Impact numbers in health policy decisions. J. Epidemiol. Community Health 56, 600-605.
Bender, R. (2005): Number needed to treat (NNT). In: Armitage, P. & Colton,
T. (eds.): Encyclopedia of Biostatistics, Vol. 6, pp. 3752-3761. Wiley, Chichester.
Heller, R.F., Buchan, I.E., Edwards, R., Lyratzopoulos, G., McElduff, P. & St.Leger, S. (2003): Communicating risks at the population level: Application of population impact numbers. BMJ 327, 1162-1165.
Hutton, J.L. (2000): Number needed to treat: Properties and problems. J. R. Stat. Soc. A 163, 403-415.
Objectives: This paper gives an overview about various impact numbers, comparing them with classical epidemiological effect measures and the original NNT statistic.
Methods and Results: Recent methodological suggestions regarding NNT statistics are summarized and compared. Relations to well-known effect measures of classical epidemiology are pointed out. The advantages and drawbacks of the new impact numbers are discussed. One disadvantage is that the new impact numbers are proposed without appropriate methods for interval estimation. Application areas are indicated, which require more research before useful NNT measures can be calculated.
Conclusions: As with the original NNT statistic, the new impact numbers are useful as reporting tool. However, more research is required to calculate meaningful confidence intervals. Moreover, in complex study situations the NNT statistics and impact numbers lose the advantage of simplicity so that deeper scientific knowledge is required to interpret the results correctly.
References:
Attia, J., Page, J., Heller, R.F. & Dobson, A.J. (2002): Impact numbers in health policy decisions. J. Epidemiol. Community Health 56, 600-605.
Bender, R. (2005): Number needed to treat (NNT). In: Armitage, P. & Colton,
T. (eds.): Encyclopedia of Biostatistics, Vol. 6, pp. 3752-3761. Wiley, Chichester.
Heller, R.F., Buchan, I.E., Edwards, R., Lyratzopoulos, G., McElduff, P. & St.Leger, S. (2003): Communicating risks at the population level: Application of population impact numbers. BMJ 327, 1162-1165.
Hutton, J.L. (2000): Number needed to treat: Properties and problems. J. R. Stat. Soc. A 163, 403-415.