As authors of QFracture we would like to respond to some of the issues raised in the rapid responses to our paper. We agree with Taguchi that next steps could include an RCT to determine whether use of screening tools such as QFracture can actually reduce incidence of osteoporotic fractures. We would also like to see an external validation of QFracture by an independent academic team. We also agree with Bazelier regarding the importance of starting follow up after a period of registration with the practice and in the methods for our study we clearly state "We included patients in the analysis only once they had a minimum of one year's complete data in their medical record". We hope this addresses her concerns.
Several correspondents suggest additional risk factors which might have been considered for inclusion in QFracture. Some of these (such as COPD and SSRIs) could be tested in future versions of the algorithm whilst others (such as vitamin D deficiency) are not routinely recorded in GP electronic health records so would not be possible to include. We are aware of the large number of predictors already incorporated in QFracture and recognise the need to keep it relatively simple to use in everyday clinical practice. We therefore propose that we will only include additional risk factors if there is a significant improvement in the performance of the algorithm in an independent validation sample or a clinically significant reallocation of patients into high or low risk categories.
The FRAX developers mention their website has had over a million hits as evidence of its widespread use in primary care. However, in order to conduct our study, we hit their website a million times which would account for a large part of their hits. More recently (in 2010), we have checked the FRAX calculations using our validation cohort and found a significant change in the scores calculated by the FRAX website compared with the scores in our paper. We do not know whether this represents a bug in the FRAX software, an error in the underlying algorithm or an intended change in the algorithm. Since the FRAX algorithm(s) is unpublished we contacted the FRAX developer. Disappointingly they have not been able to comment. In the meantime, it might be safest to assume that the validation of FRAX reported by us in the BMJ paper is a historical validation of a previous version of the FRAX algorithm rather than a validation of the current FRAX algorithm.
Competing interests:
Authors of the original paper and develpers of QFracture. JHC is director of ClinRisk and CC is statisticial consultant.
Rapid Response:
QFracture - authors response
Several correspondents suggest additional risk factors which might have been considered for inclusion in QFracture. Some of these (such as COPD and SSRIs) could be tested in future versions of the algorithm whilst others (such as vitamin D deficiency) are not routinely recorded in GP electronic health records so would not be possible to include. We are aware of the large number of predictors already incorporated in QFracture and recognise the need to keep it relatively simple to use in everyday clinical practice. We therefore propose that we will only include additional risk factors if there is a significant improvement in the performance of the algorithm in an independent validation sample or a clinically significant reallocation of patients into high or low risk categories.
The FRAX developers mention their website has had over a million hits as evidence of its widespread use in primary care. However, in order to conduct our study, we hit their website a million times which would account for a large part of their hits. More recently (in 2010), we have checked the FRAX calculations using our validation cohort and found a significant change in the scores calculated by the FRAX website compared with the scores in our paper. We do not know whether this represents a bug in the FRAX software, an error in the underlying algorithm or an intended change in the algorithm. Since the FRAX algorithm(s) is unpublished we contacted the FRAX developer. Disappointingly they have not been able to comment. In the meantime, it might be safest to assume that the validation of FRAX reported by us in the BMJ paper is a historical validation of a previous version of the FRAX algorithm rather than a validation of the current FRAX algorithm.
Competing interests: Authors of the original paper and develpers of QFracture. JHC is director of ClinRisk and CC is statisticial consultant.