Abstract
Background When monitoring long-term conditions, both over- and undertesting can risk patient harm and increase healthcare costs.
Aim To evaluate the evidence base for type 2 diabetes mellitus (T2DM) monitoring tests and develop methods for creating evidence-based testing strategies.
Design and setting Rapid reviews were conducted and a consensus process then used to evaluate the evidence base within primary care settings.
Method The authors identified tests that are recommended or used commonly to monitor T2DM. Filtering questions were created to examine the rationale for use of each test, which were answered by stepwise rapid reviews of evidence cited by guidelines, systematic reviews, and individual studies. A consensus group of patient representatives and clinicians voted whether tests should be included or excluded based on the evidence or whether further evidence was needed.
Results Of 15 tests, only haemoglobin A1c, to monitor disease progression and treatment response, and estimated glomerular filtration rate, to detect chronic kidney disease, have a strong evidence base. Based on available evidence and consensus group feedback, routinely testing for fructosamine to monitor disease progression; thyroid function, vitamin B12, ferritin, folate, clotting, bone profile, C-reactive protein, erythrocyte sedimentation rate, and B-type natriuretic peptide; and liver function for adverse treatment effects of metformin was deemed unnecessary. The study found insufficient evidence for lipids and haemoglobin to screen for secondary conditions, and for vitamin B12 to screen for adverse effects in those taking metformin.
Conclusion The study found that the evidence base for most T2DM monitoring tests is weak or absent. Clinicians should avoid non-evidence-based tests unless there are additional clinical indications for testing. Standardised evidence-based testing panels for T2DM and other long-term conditions could reduce unnecessary testing.
How this fits in
Guidelines are unclear about how to optimally monitor long-term conditions. Forty per cent of primary care tests are related to monitoring but which tests are used, and how often, varies substantially across the UK. This study identified a list of tests without strong evidence for the routine monitoring of people with type 2 diabetes mellitus (T2DM). Implementing the study's findings could reduce unnecessary testing for patients with T2DM; this could avoid costly consequences from overtesting, such as follow-up appointments, re-testing, referrals, invasive procedures, increasing clinician workload, and patient harm.
Introduction
Over 26 million people in the UK have ≥1 long-term condition.1 An estimated 4.7 million people in the UK have diabetes; of these, 90% have type 2 diabetes mellitus (T2DM).2 It is generally suggested that regular monitoring is beneficial for people with long-term conditions, but evidence for this is lacking. Earlier detection of disease progression or the development of secondary conditions allows for earlier intervention, which may improve patient outcomes. Most long-term conditions in the UK are monitored in primary care, where >40% of all tests ordered are used to monitor existing disease or medications.3
To ensure that monitoring improves patient outcomes, the number and frequency of tests need to be optimised because both over- and undertesting can be harmful. Undertesting may lead to delayed diagnoses, complications, patient harm, and litigation. Overtesting leads to an increased number of borderline results3 and false-positive results,4 which can result in unnecessary and possibly invasive follow-up tests, overdiagnosis, and unnecessary treatment. This cascade effect of medical interventions causes patient harm and anxiety.5 Previous research has estimated that 25% of pathology testing may be unnecessary.3,6
Currently, long-term condition monitoring is not optimal.7–10 Guidelines are unclear about what optimal monitoring looks like and are largely based on expert opinion.7 In the absence of clear national guidance, GP practices use local guidance that varies significantly from practice to practice.8 This has resulted in great variation in test use across the country.9,10 To avoid under- and overtesting, a good rationale is needed for each monitoring test underpinned by strong evidence. The aim was to develop a methodology to evaluate the rationale of monitoring tests and identify the evidence base for this rationale. This methodology was applied to blood tests that are currently used to monitor people with T2DM in primary care to highlight the gaps in the evidence base and identify a ‘minimal’ testing panel required for all patients with T2DM. This study is part of a larger programme of work, in which the authors aim to develop and test the cost-effectiveness of evidence-based testing strategies including optimal frequency of testing and information resources for patients and clinicians.
Method
Identifying candidate tests
The authors created a list of ‘candidate’ tests to consider that were identified through reviewing national guidelines including all tests that are currently recommended,7 a survey of clinicians and selecting tests ordered by the majority of the clinicians,11 and an analysis of routinely collected UK primary care data, including tests ordered >0.1 test per person per year for patients with T2DM.9
Tests were categorised under three reasons for testing: to 1) monitor disease progression and treatment response; 2) screen for secondary conditions; or 3) screen for adverse treatment effects for patients on stable drug treatment (that is, during the maintenance phase of monitoring; monitoring during treatment initiation and titration was not considered). For pragmatic reasons this study limited the investigation to the most common first-line treatments and known side effects listed in the British National Formulary.
An advisory group, including primary and secondary care clinicians and patient representatives, were invited to add (but not remove) tests and verify the categorisation.
Filtering questions to refine the list of candidate tests
A stepwise list of filtering questions was developed to evaluate the rationale for each test. For tests monitoring 'disease progression' or 'treatment response', these were:
Can you intervene to improve treatment response or to slow or prevent further disease progression?
Are there clear benefits of earlier detection or treatment?
For tests screening for 'secondary condition' and 'adverse treatment effect', these were:
Are people with the long-term condition at an increased risk of developing the secondary condition or side effect that the test is able to detect?
Is there anything the clinician can do to manage or treat this?
Are there clear benefits of earlier detection or treatment?
Rapid reviews to answer filtering questions
The authors conducted rapid reviews to answer each filtering question using a stepwise approach following PRISMA reporting guidelines where appropriate. If strong, up-to-date evidence was identified the authors considered the filtering question answered. The authors moved to the next step if insufficient evidence was identified to answer a filtering question (Figure 1). Owing to the nature and multitude of these rapid reviews they were not registered and separate protocols have not been published. The searches were developed by an information specialist (Supplementary Tables S1–S8 show the full search strategies). Screening was performed in Rayyan.12 Data were extracted in standardised extraction forms in Microsoft Excel, which were tested by two reviewers and amended, as necessary, before use. Data extraction and risk of bias (ROB) assessment were performed by one reviewer. ROB was assessed using standardised tools: ROBIS for systematic reviews,13 RoB 2 for randomised controlled trials,14 and ROBINS-E for cohort studies.15
First, the authors identified and appraised the evidence cited by National Institute for Health and Care Excellence (NICE) guidelines. Full texts of identified studies were retrieved, data were extracted, and ROB assessed.
Second, the authors identified and appraised evidence from studies that were part of the UK Prospective Diabetes Study (UKPDS), a pragmatic choice to rapidly identify high-quality evidence. The UKPDS is a landmark randomised, multicentre trial of glycaemic therapies in 5102 patients with newly diagnosed T2DM.16 The UKPDS study created a unique cohort with the longest prospective follow-up in a T2DM population.17 Manuscripts listed on the UKPDS website (www2.dtu.ox.ac.uk/UKPDS) were screened. All abstracts and full texts were screened independently by two reviewers and conflicts were discussed. Data were extracted on study design, population characteristics, number of included patients, and study results.
Third, the authors identified and appraised evidence from systematic reviews published in English since 2015. The authors searched the KSR Evidence database by combining terms for the long-term conditions and tests, secondary conditions, or side effects. Records were screened independently by two reviewers prioritising recent publications. Once a systematic review of sufficient quality (low ROB) was identified, the authors stopped screening. The inclusion criteria, number of included studies and patients, and summary estimates were extracted. The authors extracted the ROB judgement in KSR Evidence if available (based on the ROBIS tool).13
Fourth, the authors identified and appraised evidence from primary studies applying the same eligibility criteria as above, prioritising recent studies in UK populations. MEDLINE, Embase, and CENTRAL were searched, combining search terms for the long-term conditions, tests, secondary conditions, adverse events, and selected cohorts as appropriate. In total, 20% of abstracts and full texts were screened independently by two reviewers and conflicts were discussed. The remaining records were screened by one reviewer. Potentially relevant records were discussed with a second reviewer. Data were extracted on study design, population characteristics, number of included patients, and results for each included study.
Evidence report
The evidence identified through the rapid reviews was summarised in a report and circulated to an advisory group before a consensus meeting. Evidence was labelled as:
'good evidence' if the authors found high-quality systematic reviews or high-quality studies with large sample sizes;
'moderate evidence' if the authors found a systematic review with some quality concerns or several primary studies showing similar results but some studies having quality concerns; and
'weak evidence' if there was a single primary study with a small sample size with or without quality concerns.
For each test, the filtering questions were answered according to the available evidence as 'yes', 'no', or 'insufficient evidence'.
Consensus process
At a consensus meeting, the evidence for each test was presented, followed by a vote: 1) include test; 2) exclude test; or 3) insufficient evidence. A consensus of >80% was needed to include or exclude a test. A high consensus threshold was chosen, as this study's recommendations may influence long-term care for patients. If there was no consensus, the evidence was discussed followed by a second vote. If no consensus was reached after the second vote, the test was deemed to need further evidence.
Post-consensus meeting
One additional rapid review was conducted in response to questions raised at the consensus meeting following the same methods as above (step 4; see Figure 1). The review question was whether the patient knowing the results of lipid or liver function testing can influence patient behaviour and motivate them to implement lifestyle modifications. This was addressed with a series of systematic searches using an iterative approach. MEDLINE and Embase databases were searched.
Patient and public involvement
Two patient representatives reviewed the funding application and commented on the plain language summary. During the project, they actively participated in monthly project meetings as part of the project management group. They advised during the recruitment of further patient representatives. Four patient representatives contributed to the consensus meeting, where they had equal voting rights to the clinicians.
Results
Candidate tests
A total of 15 candidate tests were identified. Answers to the filtering questions are summarised in Tables 1
⇓-3 and evidence for each test is provided in more detail in Supplementary Tables S9–S11. See Supplementary Figures S1–S4 for PRISMA flow diagrams of each rapid review.
Table 1. Tests to monitor disease progression and treatment response — answers to filtering questions Table 2. Tests to screen for secondary conditions — answers to filtering questions Table 3. Tests to screen for adverse treatment effects of metformin for patients on stable treatment — answers to filtering questions Tests with strong evidence and rationale
Only HbA1c and eGFR, out of the 15 tests considered, had strong evidence (Tables 1 and 2).
Tests with unclear evidence or rationale
There was also insufficient evidence to justify routine liver function tests to detect non-alcoholic fatty liver disease, lipid tests for dyslipidaemia or cardiovascular risk, haemoglobin for anaemia, or vitamin B12 for metformin-related adverse effects. There was insufficient evidence on whether testing influences lifestyle changes. The effect may go in either direction (for example, a normal test result may increase or decrease the likelihood of behavioural change), so current evidence does not support the use of monitoring tests solely to promote behavioural change (Tables 2 and 3).
Tests that should not be routinely offered to patients with T2DM
Based on the evidence and a consensus meeting with an advisory group, 11 tests were identified that should not be routinely offered to people with T2DM. Fructosamine cannot replace HbA1c to monitor disease progression and should only be offered on a patient-by-patient basis if HbA1c cannot give correct results (because of genetic variant of haemoglobin). There is no rationale to offer thyroid function tests and bone profile tests, because evidence shows that people with T2DM are not at higher risk of related outcomes than the general population. Similarly, people on metformin are not at greater risk of abnormal renal or liver function, so these should not be offered routinely to people on stable metformin treatment. Also vitamin B12, ferritin tests, folate tests, clotting tests, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), and B-type natriuretic peptide (BNP) were excluded because of lack of rationale or evidence (Tables 2 and 3).
Discussion
Summary
This study found that the evidence base for most blood tests used to monitor T2DM is weak or absent. Having T2DM alone does not indicate the need for thyroid function tests, vitamin B12, ferritin tests, folate tests, clotting tests, bone profile, CRP, ESR, or BNP to screen for secondary conditions. However, if additional indications arise, these tests can and should be used. There is strong evidence supporting the use of HbA1c to monitor disease progression and treatment response. Fructosamine should only be considered if HbA1c cannot give correct results in patients with specific genetic variants. There was insufficient evidence on whether testing influences lifestyle changes. In the absence of strong evidence and the fact that the effect may go in either direction, tests should not be offered for the sole purpose of promoting behavioural change.
Strengths and limitations
To the authors’ knowledge, this is the first study to attempt to create evidence-based testing panels for long-term conditions. The main strength of this study is the use of a standardised approach and established methods to identify evidence.18 The authors of the current study developed new methods to determine the rationale of blood tests for T2DM management that can be used to optimise monitoring of other long-term conditions. Another strength is the involvement of a diverse advisory group that included clinicians and patient representatives. The advisory group helped to interpret the evidence and highlight additional factors to consider that were not captured by the filtering questions.
A weakness of this study is that the authors had to make some compromises in evidence collection for pragmatic reasons. Rapid review methods and a stepwise approach were used to identify evidence rather than full systematic reviews. As such, this study may have missed relevant evidence. However, full-text screening was stopped once strong evidence addressing the filtering question had been identified. Finally, accuracy or the natural variability of each test was not taken into account and optimal frequency of testing or optimal thresholds to use per test were not addressed.
Comparison with existing literature
This study showed that for eGFR and HbA1c there is a good rationale underpinned by strong evidence to test in people with T2DM and this is in line with current guidelines.19,20 Similarly, a study assessing the clinical value and cost-effectiveness of screening programmes for kidney disease using simulation methods concluded that annual monitoring of people with type 2 diabetes to identify the development of early kidney disease was cost-effective and beneficial for patients.21
This research suggests that a lot of current testing is unnecessary. For instance, this study showed that there is no rationale to routinely test people with T2DM with liver function or thyroid function tests. However, liver function tests are the third most commonly ordered tests for people with T2DM, with the average patient receiving 1.5 tests per year, and thyroid function tests are the sixth most commonly ordered, with many patients receiving annual testing (0.8 per person per year).9
This research identified tests lacking strong evidence that are still routinely used and recommended by guidelines. For example, lipid profile tests are used routinely in patients with T2DM (1.2 tests per person per year),9 but increasing evidence suggests that most lipid monitoring is unhelpful. The signal–noise ratio in cholesterol monitoring is weak, which means that frequent monitoring leads to more false-positive than false-negative results. It has been estimated that it takes at least 3 years,22,23 or even >5 years,24 before the signal exceeds the noise. In other words, it can take at least 3 years before plausible changes in true cholesterol levels exceed analytical and biological variation. In line with this, the minimum difference between two consecutive lipid test results that must be exceeded before the change is considered statistically significant (also known as reference change value) is high: 11%–20% in total cholesterol and 21%–30% in high-density lipoprotein and low-density lipoprotein.25
There was also no consensus on whether people on metformin should be tested regularly for vitamin B12 levels, to screen for adverse treatment effects, even though good evidence was found that metformin use increases the risk of B12 deficiency. Since metformin-induced B12 deficiency depends on dosage and length of use, and it takes time to develop B12 deficiency and related issues, it may not be necessary to regularly monitor B12 levels in every patient taking metformin.
Implications for research and practice
Future research is needed to estimate the optimal frequency of testing, which will be addressed as part of the wider programme of work, which uses economic modelling to estimate the cost-effectiveness of different testing frequencies. Annual testing is often chosen for pragmatic reasons, but these are not evidence based and can therefore lead to adverse consequences related to overtesting.26,27 The authors are currently evaluating the clinical and cost-effectiveness of this evidence-based testing panel to monitor T2DM in primary care in a cluster randomised controlled trial ('Test Smart'). Future research should also take the diagnostic accuracy of these routine tests into account and determine the optimal threshold to use, as these are essential parts of a testing strategy.22
Clinicians should avoid non-evidence-based tests such as thyroid function tests or bone profile unless there are additional clinical indications for testing. For tests where the evidence is uncertain, the final decision to monitor, what to monitor, and how often should still be made on an individual basis and in discussion with the patient.
Although tests are generally inexpensive, unnecessary testing leads to costly consequences such as follow-up appointments, re-testing, referrals, invasive procedures, increasing clinician workload, and patient harm. The methodology presented here can be applied to other long-term conditions to optimise the use of monitoring tests. Standardising the monitoring of long-term conditions based on the latest evidence can reduce unwanted variation in testing and has the potential to prevent patient harm and increased healthcare costs related to both over- and undertesting.
Notes
Funding
This study was supported by National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research (reference: NIHR201616). This research was supported by the NIHR Applied Research Collaboration West. The views expressed in this article are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. The funder was not involved in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.
Ethical approval
Not applicable.
Contributors
Martha MC Elwenspoek drafted the manuscript, developed the protocol, conducted abstract screening, full-text screening, data extraction, and performed the risk of bias assessment. Rachel O’Donnell led the behavioural review, and contributed to abstract screening, full-text screening, data extraction, and risk of bias assessment. Joni Jackson participated in abstract screening, full-text screening, data extraction, and risk of bias assessment. Sarah Dawson was responsible for developing all search strategies, and Katie Charlwood identified evidence gaps. Alastair Hay and Jessica Watson contributed clinical expertise and secured funding for the project. Penny Whiting developed the methods, secured funding, and provided project oversight.All authors had full access to the data in the study and can take responsibility for the integrity and accuracy of the data analysis. All authors have read and approved the final manuscript.
Provenance
Freely submitted; externally peer reviewed.
Data
The data that support the findings of this study are available within this article and supplementary files.
Acknowledgements
The authors would like to express their gratitude to Aisling O’Rourke for her invaluable administrative support throughout the project and for organising the consensus meeting. The authors also extend their sincere thanks to the members of the consensus group, including patient representatives Bob Cottis, Jean Palmer, Malcolm Turner, and Patricia MacCalla; GPs Sam Merriel, Rachel Johnson, Alan McFarlane, Charlie Kenward, and Darunee Whiting; renal consultant Dominic Taylor; and primary care nurse Jennifer Charlewood. The authors are particularly grateful to their topic experts Rafael Perera, Tom Lewis, and Martin Myers for their insights and expertise. Additionally, the authors wish to acknowledge the contributions of the wider team, including Mary Ward, Howard Thom, Alice Malpass, Clare Thomas, Hayley Jones, and Jonathan Sterne, as well as patient representatives Francesco Palma and Christina Stokes. The authors would like to acknowledge and remember the late Jonathan Banks, whose valuable contributions and thoughtful collaboration greatly enriched this work.
Competing interests
The authors have declared no competing interests.
