Clinical care and other categories posters: Monitoring

A look at the association between metformin use and vitamin B12 deficiency

R Burrows-O'Donoghue, M Flynn, K Dahill, D Sivakumaran, J Wong

Endocrinology, Kingston Hospital NHS Trust, London, UK

Aims Metformin is the most commonly prescribed drug therapy for patients with type 2 diabetes. In spite of its popularity, the mechanism of action of metformin remains poorly understood. An underreported side-effect of long term metformin therapy is vitamin B12 deficiency. Currently no guidelines exist for checking B12 levels of patients on metformin.

This project aimed to review the records of a cross-section of patients under endocrinology outpatients, to study the correlation of long-term metformin therapy and low vitamin B12 levels, and to gauge the frequency of testing of B12 levels in this patient population.

Methods Records of 167 patients on metformin who were seen over a three year period were reviewed. The dose of metformin was noted from clinic letters, and vitamin B12 levels were found on the electronic hospital records.

Vitamin B12 deficiency was determined as a serum B12 concentration under 180 pg/ml.

Results 64 patients (38.7%) did not have their serum vitamin B12 checked at all since electronic blood results were recorded in 2001. Of those that had B12 levels checked (n=103), 15 (14.5%) had a serum B12 below 180 pg/ml compared with an estimated national prevalence of 6%.

Summary Our findings support the correlation between metformin use and vitamin B12 deficiency. They also highlight that many patients on long-term metformin therapy are not having serum B12 regularly checked. These findings support the need for a guideline on checking B12 levels of patients on metformin.

SUPPORTING INFORMATION The conference poster for this abstract is available online in the Supporting Information section at the end of this page.

Should you intensify therapy? A simple tool to quantify the chance of meeting HbA1c targets without adding medication

AP McGovern¹, JM Dennis¹, BM Shields¹, ER Pearson², AT Hattersley¹, AG Jones¹

¹University of Exeter Medical School, University of Exeter, Exeter, UK, ²Division of Molecular and Clinical Medicine, Ninewells Hospital and Medical School, University of Dundee, Dundee, UK

Aims Clinical inertia is a major problem in type 2 diabetes and is exacerbated by HbA1c variability. We developed a model to quantify the chance of HbA1c target attainment at six months without treatment intensification.

Methods We identified 61,161 people with type 2 diabetes from the UK primary care data (CPRD) who were on first, second, or third-line therapy and had a baseline and six-month HbA1c measurement on stable treatment. Baseline HbA1c measurement was at least six months after the last medication change. Using HbA1c and clinical characteristics at baseline, we developed a logistic regression model to assess the chance of being at a target HbA1c of 53mmol/mol six months later.

Results Baseline HbA1c was the strongest predictor of target attainment. Other factors which reduced the chance of target attainment were younger age, female gender, higher number of diabetes medications and the class of medication most recently added (lower chance with sulfonylureas, and DPP4-inhibitors compared with metformin). Model performance was excellent with an area under the receiver operating characteristic curve (AUC ROCC) of 0.887 (maximum 1.0). Deployment of the models shows people with an HbA1c at the nationally recommended intensification threshold of 58mmol/mol have only a 24.1% chance of being at the 53mol/mol target six months later without intensification.

Conclusions At the currently recommended intensification threshold just under a quarter of people achieve glycaemic control without intensification. The chance of achieving target without intensification can be individually quantified using this model. This approach could be used to support clinical decision-making.

Acknowledgement The MASTERMIND consortium

What do we know about the provision of flash glucose monitoring services? A survey of diabetes nurses in the UK

J Brake¹, S Bodman², N Milne³, H Rogers⁴, M Bannister⁵

¹Diabetes Nurse Consultant, Royal Liverpool University Hospital NHS Trust, Liverpool, UK, ²Diabetes Lead Nurse, Aneurin Bevan Health Board, Newport, UK, ³Community Diabetes Specialist Nurse, Manchester University FT, Manchester, UK, ⁴Nurse Consultant, Diabetes, King's College Hospital NHS FT, London, UK, ⁵Diabetes Nurse Consultant, Bradford Teaching Hospitals NHS FT, Bradford, UK

Aim The aim was to build up a picture of the provision of flash glucose monitoring services in the UK.

Methods A SurveyMonkey questionnaire was sent by email to members of the ‘At the 4-Front’ diabetes nursing education programme (n=168) in October 2019. Recipients were sent a reminder email about the survey to help increase the completion rate.

Results There were 33 respondents to the survey, all from diabetes nurses based in England (East Anglia, 9.0%; East Midlands, 3.0%; Greater London, 15.2%; North East, 3.0%; North West, 24.2%; South East, 21.2%; South West, 3.0%; West Midlands, 21.2%). Of the respondents, 66.7% were based in secondary or tertiary care and 33.3% were based in primary or community care. Regarding flash glucose monitoring, national and local guidelines were used in practice by 63.6% and 60.6% of respondents, respectively. Among other findings, education on flash glucose monitoring was reported to be offered through individual appointments by 24.2% of respondents, in groups by 42.4%, using both of these methods by 24.2%, via an online platform only by 3.0%, and through another service by 6.1%.

Conclusion While the sample size was not very large, it is possible to tentatively conclude, in England at least, that: national and local guidelines on flash glucose monitoring have a similar, and far from universal, uptake; and almost half of patients are offered a choice of group or individual education on flash glucose monitoring. Further work is needed to drive guideline uptake in this area.

Acknowledgement At the 4-Front

SUPPORTING INFORMATION The conference poster for this abstract is available online in the Supporting Information section at the end of this page.

The prevalence of incidental haematological conditions in patients with extremely low HbA1c levels (<20mmol/mol) identified through routine diabetes testing

G Sloan¹, K Lam², J Wright³, D Selvarajah⁴

¹Department of Diabetes & Endocrinology, Royal Hallamshire Hospital, Sheffield Teaching Hospitals NHS Foundation Trust, Sheffield, UK, ²Department of Haematology, Chesterfield Royal Hospital, Chesterfield, UK, ³Department of Haematology, Sheffield Teaching Hospitals NHS FT Sheffield, UK, ⁴Department of Oncology and Human Metabolism, University of Sheffield, Sheffield UK

Aim The HbA1c is the most common test to diagnose and monitor diabetes in adults. However, it may be falsely lowered in a number of medical conditions, reducing its accuracy. In the event of extremely low HbA1c levels (<20mmol/mol), other diabetes diagnostic tests should be performed e.g. glucose tolerance tests (OGTT). Moreover, extremely low HbA1c levels may be indicative of an underlying pathology. We aimed to quantify the number of HbA1c levels <20mmol/mol processed within Sheffield and determine their haematological and diabetes outcomes.

Methods We identified all patients with an HbA1c <20mmol/mol processed in Sheffield between November 2017 and 2018, and performed a retrospective case notes review.

Results Thirty-eight patients had HbA1c levels <20mmol/mol [mean 15.3 ± 4.2]. Nine patients (21%) had known disorders affecting red cell turnover and should not have had HbA1c testing. Ten patients (26%) were reviewed by haematology and 6 new haematological diagnoses were made (4 autoimmune haemolytic anaemia). Twenty-six patients (68%) had random serum glucose tests within one year of the abnormal HbA1c, 15 (58%) were abnormal. Only one patient subsequently had a definitive test for diabetes (OGTT). One patient had known type 2 diabetes and had their diabetes medications discontinued inappropriately.

Conclusions The HbA1c may be falsely lowered in a number of serious medical conditions which may require further investigation. This study showed that this is only performed on an ad-hoc basis. We recommend an automatic alert is generated with the HbA1c report to highlight patients who may require further investigations. Additionally, alternative glucose diagnostic/monitoring tests need to be implemented when the HbA1c is extremely low.

Digital Evaluation of Ketosis and Other Diabetes Emergencies (DEKODE): Indigenous automated system reliably predicted diabetic ketoacidosis (DKA) duration and its management

P Kempegowda^1,2, A Kolesnyk³, E Melson^1,2, L Thomas⁴, A Johnson⁴, S Ghosh², P Narendran^2,5

¹Institute of Metabolism and Systems Research, University of Birmingham, Birmingham, UK, ²Department of Diabetes and Endocrinology, University Hospitals Birmingham NHS FT, Birmingham, UK, ³Department of Health Informatics, University Hospitals Birmingham NHS FT, Birmingham, UK, ⁴College of Medical and Dental Sciences, University of Birmingham, Birmingham, UK, ⁵Institute of Immunology and Immunotherapy, University of Birmingham, Birmingham, UK

Aim To retrospectively validate DEKODE system for monitoring DKA management.

Methods DEKODE, developed in collaboration between the department of diabetes and health informatics at a University Hospital in West Midlands, identifies DKA episodes based on episodes where fixed rate intravenous insulin infusion (FRIII) was prescribed. To retrospectively validate the model, all episodes identified by DEKODE from September 2018 to August 2019 was manually verified for confirmation of diagnosis. DKA duration was defined as the difference in time between FRIII prescription time and end time for DEKODE and time from diagnosis to resolution as per standard criteria for manual data. Further, appropriateness of glucose and ketone measurements during entire DKA duration and fluids prescribed in the first 12h of diagnosis were compared between the two datasets. The difference between manual and automated data were analysed using Prism v6.0 (Graphpad Inc) and results are presented as the mean and standard error of mean (SEM).

Results A total of 150 episodes were identified by DEKODE during the study period. Of these, 147 had confirmed DKA. There was no significant difference in DKA duration between DEKODE and manual data (mean ± SEM, 16.0 ± 1.0h; 17.5 ± 0.9h; p=ns). Similarly, there was no difference FRIII appropriateness (98.3%±1.2%; 97.9%±1.1%; p=ns), hourly glucose (98.5%±2.6%; 105.6%±2.5%; p=ns) and ketone measurements (43.3%±2.1%; 47.1%±2.2%; p=ns) between the two systems. However, DEKODE over-predicted proportion of fluids prescribed (96.9%±3.2%; 84.4%±3.1%; p=0.0047).

Conclusion DEKODE system could help in monitoring DKA management cutting time to collect data, thus providing real-time performance results. Further prospective validation is currently underway.