Trends and variation in unsafe prescribing of methotrexate: a cohort study in English NHS primary care

Summary

At least 21 people died from methotrexate poisoning in England and Wales between 1993 and 2017. The prescribing of methotrexate 10 mg tablets remain, common, but practice is extremely variable: most practices prescribed none, but 697 NHS GP practices in England prescribed ≥14.3% of all methotrexate as 10 mg tablets; the percentage for 66 practices was ≥52.4%. Breaches of existing guidance were more common in urban practices and those with a worse composite OpenPrescribing quality score. CCG membership explained 25.5% of prescribing variation, suggesting that CCGs exert a substantial influence on clinical practice regarding methotrexate.

Strengths and limitations

OpenPrescribing data includes all prescribing in all typical practices in England, thereby minimising the potential for obtaining a biased sample. Real prescribing and spending data, which are sourced from pharmacy claims, were used; as such, these did not need to rely on surrogate measures. This was complemented with aggregated patient-level data to identify patterns of co-prescribing, which, again, were sourced from pharmacy claims. Using primary data, rather than survey data, eliminates the possibility of recall bias. Ideally, hospital prescribing would also have been included; the authors have advocated for these data to be more widely shared but, at present, they are only available for pharmaceutical industry marketing and a limited range of unpublished analyses at NHS Improvement.

Data on deaths related to methotrexate were obtained from ONS, the most robust dataset available on drug poisonings. This dataset is comprehensive but figures are for deaths related to those registered in each calendar year, rather than those that occur in each year. In addition, a coroner’s inquest can take months, or even years, to complete.

The data on legal claims obtained from NHS Resolution have limited structure and do not give any information on the reason for the claim: deaths data may include deaths not related to primary care prescribing, for example, after prescriptions that were given in hospital or adverse reactions not caused by excessive dosing.

Comparison with existing literature

The authors are aware of no prior work on the prevalence of breaches on methotrexate guidance; however, incomplete implementation of the important national NHS safety alert is consistent with extensive prior work showing incomplete or slow adoption for other national prescribing guidance.^18,19 A 2008 survey of 376 members of the British Association of Dermatology reported 49 deaths of patients taking methotrexate; of these, one was caused by confusion between the 2.5 mg and 10 mg doses, and two were caused by daily rather than weekly dosing.²⁰ However, these survey data rely on recall over many years of a doctor’s career unlike the present study, which used data from death certificates where poisoning related to methotrexate was mentioned.

One article, published in 2006, reported 137 patient-safety incidents related to methotrexate in England over the previous 10 years;²¹ however, these figures are likely to include a wide range of issues, including inconsistent documentation by clinicians regarding dose (such as using 2.5 mg tablets only), prescribing of concomitant folate, and use of blood tests.

National organisations in France, Canada, and Australia have also issued advice to prescribers, similar to the NHS; also similar to the findings presented here, deaths and other errors associated with methotrexate prescribing continue to occur in these countries.^22–24

Implications for research and practice

Previous work across several countries has resulted in a variety of suggested approaches to minimising the risks involved with methotrexate prescribing,^22,23,25 and the EMA has recently issued further guidance.^4,5 NHS Improvement has classed overdoses associated with methotrexate as a Never Event;⁶ however, the authors found no evidence of any action taken by any national NHS body when anomalous prescribing has been detected in a region or practice; this is concerning but may, in part, be explained by the fact that the responsibility for enacting change is unclear. The Never Event framework was initially managed by the NPSA; this body closed in 2012 and became part of NHS Improvement, which does not have any remit over primary care; this may change as NHS Improvement merges with NHS England.

Locally, the NHS has invested extensively in medicines optimisation activity, in which teams of pharmacists in every CCG monitor prescribing behaviour and advocate for change with individual clinicians. Previous work has shown that CCG membership is associated with prescribing patterns for practices,^19,26,27 and, in the present study, a statistically significant relationship was found between the CCG to which a practice belonged and the variation in prescribing of methotrexate 10 mg tablets. It is very concerning that a number of CCGs exhibit minimal change-related activity in response to an important safety alert. In the authors’ view, there is room for substantial improvement in local staff training, alongside open data monitoring by NHS England, and appropriate action for those failing to implement change.

Practices with a higher OpenPrescribing score were also more likely to have a higher proportion of 10 mg tablets prescribed than those with a lower score. It is unlikely that methotrexate causes poor prescribing in other areas, or poor prescribing in other areas cause breaches of safety guidance when prescribing methotrexate. The authors propose that both aspects of prescribing are linked by more fundamental issues, such as individual clinicians’ skills on evidence-based medicine, or the extent to which the practice team works together to review prescribing behaviour as indicated by their own practice’s data, identify areas in which they are outliers or exhibit unusual prescribing, and take action collectively to address these issues.

The scale of breaches of existing guidance for methotrexate is clear; however, it is likely that many other safety issues exhibit similarly prevalent breaches. Given the low cost and high impact of data analyses in identifying individual practices and the scale of national problems, it is suggested that this should be a high priority for research. Although many deaths have been attributed to methotrexate, data from death certificates are scant. Accessing and reviewing the text of coroners’ reports for all deaths associated with methotrexate would establish the role the drug played in these deaths and help identify preventive strategies; although this would take time, this level of review seems consistent with the prioritisation of correct methotrexate dosing as a Never Event.

Decision-support tools and pop-ups that feature in a clinician’s EHR software at the point of care may offer an important opportunity to block unsafe prescribing; the methotrexate safety alert from the NPSA specifically highlights that EHRs should include alerts and prompts.³ The NHS makes significant investments in EHRs²⁸ and it is imperative that their user interfaces help health professionals to prescribe safely. It was found that one of the EHR systems used in the NHS (Vision) was associated with higher prescribing of 10 mg tablets; this may reflect weaker preventive measures in this system. This finding should be investigated promptly to understand whether the design choice of the user interface in Vision increases the likelihood that a patient is prescribed 10 mg tablets.

More generally, the authors have been repeatedly blocked from researching the impact of pop-ups on prescribing, as there is no national framework or data on which pop-ups are implemented in each setting. This also means that NHS commissioners and leaders cannot routinely identify which pop-ups are implemented across the NHS. In order to realise the often-cited potential of technology to improve safety, the NHS needs better oversight of technology and the ability to ensure system providers make modifications quickly if, or when, shortcomings are identified.^29,30

In summary, the prevalence of unsafe methotrexate prescribing has gradually reduced, but it remains common and with substantial variation between GP practices. This is unlikely to be a unique problem. The authors recommend that the NHS invests in better strategies around audit and targeted dissemination of safety information, and identifies named individuals and roles with responsibility for implementing the recommendations of safety alerts.