Re: Margaret McCartney: The “breakthrough” drug that’s not been shown to help in Alzheimer’s disease
Margaret McCarthey’s article is correct to be cautious about the effectiveness of this drug for Alzheimer’s disease which she says might be a chance finding.. But this is not due to the relatively small size but due to the nature of testing for subgroups within an overall negative trial. If the overall trial showed no effect but a subgroup did then it is important to interpret these findings cautiously.
Firstly, if those with a mild disease benefited then it follows that those with severe disease got worse (as the overall trial did not show anything). Is this likely to be true? If not then it is likely to be a chance finding. A similar thing happened in another trial of a drug from Eli Lilly of raloxifine (1). In this large trial there was no impact of raloxifine on cardiovascular events. However, among a subgroup of women with increased risk of cardiovascular disease at baseline raloxifene appeared to show a 40% reduction in cardiovascular events. In a replication study of raloxifine solely among women with pre-existing cardiovascular disease there was no effect.
Whilst it is tempting to look for effectiveness among subgroups when the overall trial shows no effect these can often be misleading and, at best, are pointers to replication trials. Consequently, until a trial has been completed among patients with mild disease then the results are not believable.
(1) Barrett-Connor et al. JAMA. 2002;287(7):847-857. doi:10.1001/jama.287.7.847
(2) Barrett-Connor et al. N Engl J Med 2006; 355:125-137
Rapid Response:
Re: Margaret McCartney: The “breakthrough” drug that’s not been shown to help in Alzheimer’s disease
Margaret McCarthey’s article is correct to be cautious about the effectiveness of this drug for Alzheimer’s disease which she says might be a chance finding.. But this is not due to the relatively small size but due to the nature of testing for subgroups within an overall negative trial. If the overall trial showed no effect but a subgroup did then it is important to interpret these findings cautiously.
Firstly, if those with a mild disease benefited then it follows that those with severe disease got worse (as the overall trial did not show anything). Is this likely to be true? If not then it is likely to be a chance finding. A similar thing happened in another trial of a drug from Eli Lilly of raloxifine (1). In this large trial there was no impact of raloxifine on cardiovascular events. However, among a subgroup of women with increased risk of cardiovascular disease at baseline raloxifene appeared to show a 40% reduction in cardiovascular events. In a replication study of raloxifine solely among women with pre-existing cardiovascular disease there was no effect.
Whilst it is tempting to look for effectiveness among subgroups when the overall trial shows no effect these can often be misleading and, at best, are pointers to replication trials. Consequently, until a trial has been completed among patients with mild disease then the results are not believable.
(1) Barrett-Connor et al. JAMA. 2002;287(7):847-857. doi:10.1001/jama.287.7.847
(2) Barrett-Connor et al. N Engl J Med 2006; 355:125-137
Competing interests: No competing interests