7th March 2011
A report from the NHS Information Centre has assessed the variation in the use, by the NHS in England in 2009, of 18 groups of medicines considered by NICE in 29 technology appraisals. In the 12 groups where a comparison could be made, observed use was higher than predicted use for eight and lower for three, and in one case it was lower on one set of assumptions and higher on the alternative. This updates and expands a previous report, which looked at usage of a smaller range of medicines in 2008.
Although the information in this report is ‘experimental’ in status owing to limitations in the data (see below), it is likely to be useful to medicines managers, including the emerging GP consortia, and may highlight areas where careful examination of local usage of NICE-approved medicines and implementation of NICE guidance is appropriate. Feedback on the report is encouraged.
How are the results presented?
For each of the 12 groups where a comparison could be made, the report provides a comparison of national (England) and SHA (where appropriate) predicted versus observed uptake of medicines. The results are always presented at national level, with SHA and anonymised PCT analyses where possible. If the drug is mainly used for cancer then an analysis by cancer network is presented. Finally a chart shows the quarterly cost since 2001 (or since the drug was introduced if later than this). For the six groups where a valid comparison could not be made, these are presented separately with the reasons for this along with a series of questions inviting the reader to suggest improvements to the method or data.
What did the report find?
Out of the 12 groups where a comparison could be made, observed use by the NHS in England was higher than the predicted use for eight and lower for three. In the case of drugs used for acute coronary syndrome, the result was lower on one set of assumptions and higher on the alternative. The results are presented in the form of a ratio between predicted and observed use where possible. A ratio of 0.6 to 1 would indicate that observed use was 40 per cent less than expected while a ratio of 1.4 to 1 would indicate that use was 40 per cent higher than expected
The table below gives the national ratio and the SHA range: the range for PCTs (where available) is substantially greater.
|Drug or group||National ratio||SHA range|
|Ezetimibe||1.19||0.95 to 1.48|
|Acute coronary syndrome
(Abciximab, eptifibatide and tirofiban)
|0.60 to 1.02
(see note 1)
|0.31 to 1.87
(see note 1)
|Varenicline▼||1.17||0.7 to 2.1|
|Naltrexone||0.82||0.41 to 1.56|
|Ranibizumab▼||1.98||1.5 to 2.8|
|Insulin glargine and insulin detemir||1.12||0.94 to 1.14|
(Alendronate, Etidronate, Risedronate, Raloxifene, Strontium ranelate, Teriparatide)
|1.34||1.12 to 1.57|
|Alzheimer’s disease(Donepezil, Galantamine, Rivastigmine and Memantine)||1.46 or 3.43
(see note 2)
|1.14 to 1.92, or 2.69
to 4.52 (see note 2)
|Trastuzumab▼||0.86||0.74 to 1.12|
|Erlotinib▼||1.08||0.79 to 2.01|
Note 1 : Two figures are given to reflect different assumptions about the dose used. The range given for the SHAs encompasses both assumptions
Note 2: QOF data suggested a substantially smaller patient population (giving a higher ratio) than that estimated using public health observatory data (which gave a lower ratio)
What were the limitations to the data?
A report of this kind is limited by the available data. It is not yet possible to use routinely collected data to measure drug usage by indication or the stage of the disease at which it is used because of the limited information recorded, so predicted use (using the estimated number of eligible patients, the average dose and average length of treatment) was compared with observed use
For many medicines or groups of medicines, QOF incidence and prevalence data for the relevant clinical conditions could be used. However, for one of these (Alzheimer’s disease) the QOF data suggested a substantially smaller patient population than that estimated using public health observatory data. In addition, assumptions about the average length of treatment are central to producing predictions of use and there are difficulties in producing robust estimates of expected numbers of patients at a sub-national level.
With regard to volume of prescribing, there are significant limitations to the data. Data on medicines prescribed and dispensed in primary care can be obtained from NHSBSA Prescription Services, and these are widely considered to be generally complete and accurate. However the NHS does not centrally collect data on use of medicines in hospital. Instead, HPAI (Hospital Pharmacy Audit Index) data were used. HPAI is a database owned and maintained by IMS Health which contains data on use of medicines within hospitals based on issues from hospital pharmacies. There are limitations to these data, which are discussed in the report. One important limitation is the inability to obtain complete data on the supply of medicines through Homecare, where the medicine is administered to the patient at home as part of a package of treatment. The use of homecare as a means of medicines delivery has been growing in recent years. For this report pharmaceutical companies were invited to validate and improve data sets relating to secondary care medicines.
Variation in the use of medicines between cancer networks, SHAs and PCTs may be due to a number of factors including:
- Natural variation in populations, both in demographic profile and disease prevalence
- The national model used for estimation of eligible patients or assumptions of the average length of treatment being inappropriate or inaccurate (for example, due to changes in clinical opinion after the guidance was issued)
- Variation in presentation to the NHS by the relevant populations
- Variation in prescribing at the local level
- Variation in the use of alternative products or procedures
Nevertheless, the report contains further useful information which may help judge if NHS resources are being used in line with NICE recommendations, at a national, SHA, PCT or cancer network level. Feedback is requested from users on how best to estimate uptake to allow meaningful interpretation of any variation across local NHS organisations in the future.
The report demonstrates considerable variation in prescribing practices, and by implication, success at implementing NICE guidance, across individual PCTs. It indicates that there are considerable opportunities for improvements in this respect. For example, at PCT level the ratio for varenicline ranges from 0.20 (implying use 80% less than predicted) to 3.21 (implying use 221% higher than predicted). Notwithstanding the limitations to the data, this report will be useful to medicines managers, including the emerging GP consortia, and may highlight areas where careful examination of local usage of NICE-approved medicines and implementation of NICE guidance is appropriate.
Related PJ Online article: Patient access to NICE-approved drugs varies widely, report reveals
Make sure you are signed up to NPC Email updates — the free email alerting system that keeps you up to date with the NPC news and outputs relevant to you