Simple tests to screen for diabetic peripheral neuropathy
Abstract
This is a protocol for a Cochrane Review (Diagnostic test accuracy). The objectives are as follows:
To determine the diagnostic accuracy of each simple test as triage to screen for diabetic peripheral neuropathy (DPN) involving limbs within different settings, or as replacement of nerve conduction studies (NCS) for the clinical diagnosis of DPN involving limbs, with NCS as the reference standard.
Background
Diabetes mellitus is a metabolic disorder resulting from a defect in insulin secretion, insulin action, or both. In 2000, more than 175 million people all over the world suffered from diabetes (Yach 2006), of which 5% to 10% had type 1 diabetes and 90% to 95% had type 2 diabetes (Creager 2003). It is estimated that the number of people with diabetes will reach around 360 million in 2030 (Wild 2004; Yach 2006). Diabetes can induce long‐term complications, including retinopathy, nephropathy, neuropathy and other vascular complications (American Diabetes Association 2013).
Target condition being diagnosed
Diabetic peripheral neuropathy (DPN)
DPN is one of the most common microvascular complications in both type1 and type 2 diabetes. DPN has been defined as "the presence of symptoms and/or signs of peripheral nerve dysfunction in people with diabetes after the exclusion of other causes" (Boulton 1998; Soliman 2002). It is the most common component in the causal sequence to foot ulceration (Reiber 1999). DPN can be broadly separated into generalised symmetrical polyneuropathy, and asymmetrical (focal and multifocal) neuropathy (Boulton 2004; Boulton 2005a; Dyck 2011a; Thomas 1997) (Table 1). Autonomic neuropathy can be either present or absent in DPN (American Diabetes Association 1996). A staging system, which encompasses four stages, has also been developed to provide a framework for diagnosis and management for DPN (Boulton 1998) (Table 2).
| Classificationa | Subgroup |
| Generalised symmetric polyneuropathies | Chronic sensorimotor (typical DPN) |
| Acute sensory | |
| Autonomic | |
| Focal and multifocal neuropathies | Cranial |
| Truncal | |
| Focal limb | |
| Proximal motor (amyotrophy) | |
| Co‐existing CIDP |
aAccording to Boulton et al. (Boulton 2005a);
DPN: diabetic peripheral neuropathy; CIDP: chronic inflammatory demyelinating polyneuropathy.
| Stage of diabetic peripheral neuropathy | Characteristics |
| Stages 0/1: no clinical neuropathy | |
| No symptoms or signs | |
| Stage 2: clinical neuropathy | |
| Chronic painful | Positive symptomatology (increasing pain at night): burning, shooting, stabbing pains ± pins and needles |
| Absent sensation to several modalities and reduced or absent reflexes | |
| Acute painful | Less common |
| Diabetes poorly controlled, weight loss | |
| Diffuse (trunk) | |
| Hyperaesthesia may occur | |
| May be associated with initiation of glycaemic therapy | |
| Minor sensory signs or even normal peripheral neurological examination | |
| Painless with complete/partial sensory loss | No symptoms or numbness/deadness of feet; reduced thermal sensitivity; painless injury |
| Signs of reduced or absent sensation with absent reflexes | |
| Diabetic amyotrophy | Muscle weakness and wasting |
| Sensory loss is slight, but pain at night common | |
| Subacute onset | |
| Stage 3: late complications of clinical neuropathy | |
| Foot lesions, e.g. ulcers | |
| Neuropathic deformity, e.g. Charcot joint | |
| Non‐traumatic amputation | |
Some evidence has shown that the prevalence of DPN among people with diabetes in the UK is estimated to be 50% (Sugimoto 2000), while the World Health Organization estimate for the UK is 29% (Wild 2001). A prospective study with 7.5% participants diagnosed with DPN at baseline showed that the prevalence increased to 45% after 25 years of follow‐up (Pfeifer 1995). In a large cohort of people with DPN in the UK, 7% developed a diabetic foot after one year (Abbott 1998).
DPN is largely concerned with the feet and lower limbs, although in some severe cases the hands may also be affected (Boulton 2005a; Boulton 2005b). Typically, it is a chronic, symmetrical and length‐dependent condition, compromising multiple nerves (Dyck 2011a; Tesfaye 2010). DPN of the limbs may involve large‐fibre nerves (more related to touch, vibration, position perception and muscle control), small‐fibre nerves (more related to thermal perception, pain and autonomic function) (Vinik 2004) or both. Most patients, however, have both large‐ and small‐nerve fibre damages in DPN of the limbs (Vinik 2004).
DPN of the limbs increases with both age and duration of diabetes, and seems more common in those with suboptimal glycaemic control and obesity (Boulton 2005b; Smith 2013). It often starts at the distal ends of the longest nerves with a stocking‐glove presentation and moves proximally (Boulton 2005b). Up to 50% of patients, however, may be asymptomatic (Boulton 2005a). Frequently reported symptoms in DPN could be positive (painful) symptoms or negative (non painful) symptoms (Boulton 2005b; Davies 2006; Melton 1999).
Reference standards
Electrodiagnostic findings provide a higher level of specificity for the diagnosis of polyneuropathy and should be included as part of the work‐up. Nerve conduction studies (NCS) are the most informative part of the electrodiagnostic evaluation which commonly include both NCS and needle electromyogram (EMG) (England 2005). For NCS, small pads are taped to the skin, deliver mild electric shocks and detect electric signals. Compared to the whole EMG, for which it may be necessary to insert thin needles into the muscles, NCS alone are relatively simple, noninvasive and timesaving. Further, due to the objectivity, reliability and sensitivity in the measurement of peripheral nerve function, NCS have long been a minimal criterion or a gold standard test for confirming the diagnosis of peripheral neuropathies (Buchthal 1957; Daube 1999; Donofrio 1990; Dyck 1988; Nasseri 1998).
Routine NCS include evaluation of motor function of the median, ulnar, peroneal, and tibial nerves, and sensory function of median, ulnar, radial, and sural nerves (Albers 1995). Recommended attributes encompass amplitude, distal latency, distance, conduction velocity, F‐wave latency and other measurements. It is important to decide how many and which nerves and parameters to assess when performing NCS (American Diabetes Association 1992). As different nerves and multiple attributes can be chosen in NCS, diagnostic criteria might vary in different studies (Dyck 2011a; Dyck 2011b). Despite many previous recommendations regarding NCS criteria of the diagnosis of polyneuropathy, no formal consensus exists (England 2005). In our review, we will accept the minimal diagnostic criteria as abnormality of one or more attributes (exceeding the normal limits between the 1st and 99th percentiles, or exceeding mean ± 2.3 standard deviation; variables, such as age, height, and temperature, should be considered when developing the reference range and interpreting the results) in two or more separate nerves to correctly define DPN (Dyck 1988; Dyck 2011a; Feldman 1994).
Results of NCS are vulnerable to many factors including filter setting, type of electrodes, the location of recording, limb temperature, qualification of examiner and other aspects. All these factors require meticulous attention to detail for reliable NCS (American Diabetes Association 1992). Applicable variables such as skin temperature, age, height, sex, and weight should be measured and accounted for when reporting a NCS as normal or abnormal (AAEM 1999).
In addition, two potential disadvantages must be acknowledged when NCS are considered in clinical and research settings. First, NCS have limits on the availability for routine diagnostic evaluation of DPN. Second, NCS are insensitive for the identification of small‐fibre neuropathy (Perkins 2003), although the clinical importance of small‐fibre neuropathy is likely to be insignificant in the context of DPN in which progressive loss of all nerve fibres is observed (Giannini 1999; Perkins 2003).
Index test(s)
Nowadays various simple neurological tests have been reported to be used for screening for DPN, some of which have also been combined into composite scoring systems to enhance the accuracy in the detection of DPN (Cornblath 2004; Perkins 2003). These tests whose accuracy we will evaluate mainly refer to the assessment of large‐fibre function, involving tendon reflex, pressure/touch sensation, vibratory sensation and protective sensation.
Ankle reflex test
While reflex tests are a conventional clinical examination in neurology, it is most common to test only ankle reflexes in the assessment of DPN (Cornblath 2004). The test is performed at both ankles. With the patient sitting or lying, the examiner dorsiflexes the foot and gently strikes the Achilles tendon with the reflex hammer. In the absence of reflex, the test can be repeated with reinforcement. Reflexes are typically scored as zero (absent with reinforcement), one (present but decreased), two (normal), three (increased), or four (increased with clonus) (Smieja 1999).
Touch sensation test
Semmes‐Weinstein monofilament test (SWMT)
The SWMT is a common screening tool for assessing the sensory function and the loss of pressure sensation (light touch perception) (Appendix 1). The size of the monofilament includes 0.5 g, 2 g, 10 g, 50 g, 200 g and other various types, which indicate the magnitude of the force on the monofilament when the monofilament is just bent. A 10 g monofilament test (also referred to the 5.07 monofilament) is the most common in practice (Boulton 2004; Valk 1997).
Neuropen
The Neuropen combines an interchangeable 10 g monofilament for cutaneous pressure assessment, and a calibrated sterile Neurotip for assessing pain sensation (Paisley 2002). The operation of the 10 g monofilament in the Neuropen is similar to that of SWMT (Appendix 1).
Ipswich touch test (IpTT)
As a simple, quick, and easily taught procedure, IpTT has been developed recently to screen DPN with the initial purpose of simplifying touch sensation test from SWMT (Appendix 1). The fact that IpTT necessitates little training may facilitate care assistants and nurses to obtain immediate feedback as to which patients require protection (Rayman 2011).
von Frey's hairs for testing touch perception thresholds
von Frey's hairs are a fast, novel and easy‐to‐perform procedure (Appendix 1). They are designed based on the similar principle of SWMT; however, touch perception thresholds can be assessed by buckling the hairs with the force ranging from 0.026 g on the first hair to 110 g on the last hair (Moharić 2012).
Vibratory sensation test
128‐Hz standard tuning fork
As an easy and traditional way to test vibratory sensation, the 128‐Hz standard (non‐graduated) tuning fork is a tool of screening for DPN (Appendix 1). An abnormal response is identified when the tested patients fail to perceive the vibration sensation while the examiner can. There are two general methods: on‐off method and timing method (Perkins 2001).
Graduated tuning fork
Unlike the standard tuning fork with limited capability to determine only the presence or absence of vibration perception, the graduated 128‐Hz tuning fork (Rydel‐Seiffer tuning fork) is able to determine the ability of patients to discriminate different vibration intensities (Appendix 1) (Garrow 2006; Kästenbauer 2004; Thivolet 1990). It relies on a threshold of vibration extinction estimated by the intersect between two virtual triangles that move exponentially on a scale from 0 to 8. In general practice, most physicians would rate a tuning fork vibration perception threshold of lower than four as abnormal (Liniger 1990)
VibraTip
VibraTip is a pocket sized, wipe clean device to test vibration perception for routine screening of DPN (Appendix 1). The product can overcome the limitations of using tuning forks by using a vibrating motor which provides consistent frequency and amplitude to allow a consistent intensity of vibration without pressure, coldness and sound. When activated, it provides a stimulus of 128 Hz.. VibraTip can be applied to the toe from any angle facilitating ease of testing, and can be magnetically attached to a specially designed neck lanyard facilitating ease of access (Bowling 2012; Bracewell 2012).
Electromechanical instruments for testing vibration perception thresholds (VPT)
VPT is one type of quantitative sensory test which is an extension of the sensory portion of the neurological evaluation with the ability to determine the absolute threshold in thermal perception, light touch perception, pain perception, cutaneous current perception, as well as vibration perception (American Diabetes Association 1992).
Electromechanical instruments for VPT include Biothesiometer, Neurothesiometer, Maxivibrometer, Vibrameter, Vibratron and the CASE IV system (Garrow 2006; van Deursen 2001) used on the basis of method of limits or method of levels (also called 'forced choice') (Appendix 1) (Cornblath 2004; Hansson 2007; Shy 2003). An average of three readings is recorded commonly from a test site (for example, at the end of the great toe). VPT cut‐off scores indicative of high or low risk for long‐term complications vary by the type of equipment utilised (Garrow 2006). Because of their ability to measure lower VPTs than a tuning fork, electromechanical devices have been recommended for community screening and routine clinical use, but the their expense can be higher (Garrow 2006).
Tactile circumferential discriminator (TCD)
The TCD is a new, portable sensory testing device used for a two‐point discrimination test which can reflect large‐fibre nerve function (two‐point discrimination) (Appendix 1). The device consists of a handheld disc with eight protruding rods of increasing circumference (numbered zero through seven). Rod zero rod is 12.5 mm in diameter, and rod seven is 40 mm. Scores are denoted as the lowest number of rods a patient can discriminate from rod zero and this is the threshold value of the TCD test. A score of six or higher is significantly correlated with neuropathy (Maser 1997; Vileikyte 1997).
Steel ball‐bearing test
The steel ball‐bearing test is also a novel test, invented to exam the protective sensory (Appendix 1). Five ball‐bearings numbered 1 to 5 correspondingly designed with a diameter of 1.5, 2.0, 2.5, 3.0 and 3.5 millimetres are used in this test. The score range of the ball‐bearing test is 1 to 6, which indicates the smallest ball‐bearing that the patient can feel. A score of six indicates that the patient can not feel any of the ball‐bearings. Both feet will be examined and physicians begin with the right one. The higher result of the two feet will be recorded as the ball‐bearing score (Papanas 2006).
Clinical pathway
It is recommended that all patients should be screened for DPN at the diagnosis of type 2 diabetes and five years after the diagnosis of type 1 diabetes and should receive one or more of the following tests annually: pinprick, temperature, ankle reflex, and vibration perception (128‐Hz tuning fork) or pressure sensation (10 g monofilament test) (American Diabetes Association 2013; Boulton 1998; Boulton 2005a). Combinations of more than one test may help to detect DPN more sensitively (Boulton 2005a). In such screening, any history of neuropathic symptoms should be elicited and a careful clinical examination of the feet and lower limbs should be performed (Boulton 2005a); in fact, NCS and exclusion of other causes are rarely needed except when the diagnosis of DPN needs to be confirmed (American Diabetes Association 2013).
As recommended by the 1988 consensus statement from the San Antonio conference on diabetic neuropathy, multiple assessments, including clinical symptoms, clinical signs, electrodiagnostic studies, quantitative sensory testing and autonomic function testing, should be applied for the diagnosis and classification of DPN (American Diabetes Association 1988). The report from the American Academy of Neurology in conjunction with the American Association of Electrodiagnostic Medicine and the American Academy of Physical Medicine and Rehabilitation define distal symmetric polyneuropathy (of which DPN of limbs is a member) and suggests that patients with abnormal NCS have a relatively high likelihood of this condition (England 2005). Recently, it was proposed that an abnormality of NCS combined with symptom(s) or sign(s) is essential to confirm the diagnosis of DPN since NCS appears to be the first objective and quantitative indication. Symptoms, signs or both without abnormal NCS contribute to the diagnosis of possible clinical DPN or probable clinical DPN, while abnormal nerve conduction alone without symptoms or signs may support the diagnosis of subclinical DPN (Dyck 2011a; Tesfaye 2010). The procedure of screening and diagnosis of DPN in clinical care has been summarised in a flow chart (Figure 1).
Flow chart of screening and diagnosis of DPN of limbs in clinical practice
Screening tests could be tests to identify symptoms and/or signs
Prior test(s)
Type 1 or type 2 diabetes should be confirmed by diagnostic tests for diabetes. Information on the duration of diabetes, history of foot ulcer, glycaemic control and complaints related to peripheral neuropathy should be obtained.
Role of index test(s)
Relatively traditional tests such as ankle reflex, 10 g monofilament test and 128 Hz tuning fork have been recommended as screening tests for DPN. They can be used solely or jointly as triage tests in clinical practice for physical examination to assess the signs of DPN, which may also contribute to clinical diagnosis of DPN. Their results require confirmation by more objective measures such as electrodiagnostic, quantitative sensory and autonomic function tests, which can help establish the confirmed diagnosis and classification of DPN (American Diabetes Association 1992; Boulton 2005a). However, in some population‐based epidemiological studies, these tests have been used jointly in the replacement of NCS to identify DPN.
VPT potentially offers a quick and accurate screening instrument to evaluate DPN in the clinic; however, selection of the suitable instrument may depend on the availability of resources, the number of clinicians involved in neuropathy testing and frequency of use (Garrow 2006). It has also replaced NCS to detect patients with DPN in epidemiological studies.
Although TCD and the steel ball‐bearing test are new tests for the assessment of large‐fibre function, wider use may be expected in the near future (Papanas 2011).
Alternative test(s)
Some other simple tests, which mainly assess small‐fibre function are also available to screen for DPN, including conventional tests ‐ temperature sensation test (e.g. Tip‐therm) (Viswanathan 2002) and superficial pain test (e.g. Neurotip) (Perkins 2001), and innovative tests ‐ NeuroQuick (Ziegler 2005) and Neuropad (Papanas 2005). However, there is no acknowledged 'gold standard' for the evaluation of tests in diagnosing small‐fibre neuropathy (skin biopsy seems relatively preferred) (Devigili 2008). It is still insufficient to evaluate the accuracy of these tests using NCS as a reference standard, which primarily indexes large‐fibre dysfunction, and therefore, we will not include these tests in our review.
Rationale
DPN places a large burden on healthcare budgets. Of all the complications of diabetes mellitus, lifetime expenditures on DPN ranks third after macrovascular disease and diabetic nephropathy (Caro 2002). If patients with DPN progress to diabetic foot, any foot lesion occurring as a result of diabetes and its complications (Boulton 2008), makes the costs of long‐term treatment much heavier. Curing one case of diabetic foot without requiring amputation would cost 17,500 US dollars (13,075 EUR, September 2012 conversion), while the cost of an amputation is 30,000 US dollars to 35,000 US dollars (23,075 EUR to 26,920 EUR, September 2012 conversion) (Ragnarson 2004). But if DPN could be detected in the early stage, enhanced glucose control might prevent the development of clinical neuropathy and reduce nerve conduction and vibration threshold abnormalities (Callaghan 2012).
From the perspective of clinical practice, screening for DPN in community and outpatient settings successfully predicts those at risk of ulceration (Abbott 2002; Adler 1997). Hospitalised patients with diabetes, who are likely to be older, bed bound and with more co‐morbidities, should also be screened so that foot protection can be targeted, because 3.3% of people with diabetes in hospital acquired a foot lesion (Rayman 2010). Although some tests have been recommended in related clinical guidelines for the diagnosis or screening of DPN (Boulton 2003; Boulton 2005a; Boulton 2005b; Vijan 1997), the development of these recommendations was based more on expert consensus than sound evidence. So far, there is no agreement which standardised screening tool should be applied in clinical practice.
As for epidemiological research, index tests used for the assessment of the prevalence of DPN varied in different studies and thus, the studies resulted in different estimates ranging from 17% to 60% (Adler 1997; Davies 2006; Gregg 2004; Liu 2010; Tesfaye 1996; Won 2012; Young 1993). Not only were the varied estimates attributed to different populations but also to the different screening tools (Davies 2006).
There are three related systematic reviews published: two focus on the SWMT while another involves SWMT, tuning fork, NSS, NDS and MNSI (Dros 2009; Feng 2009; Kanji 2010). They all prefer to use NCS as the reference standard. In the two studies that are only relevant to SWMT, variation in both of the diagnostic values and the accuracy was found (Dros 2009; Feng 2009). However, both reviews solely evaluated the accuracy of SWMT, failing to provide the whole spectrum of tests in this field. Another review found that abnormal results on monofilament testing and vibratory perception (alone or in combination with the appearance of the feet, ulceration, and ankle reflexes) are the most helpful signs (Kanji 2010). However, this review limited the setting to the bedside, where the accuracy of tests may differ from that in community due to possibly different disease spectra. In the Kanji 2010 review, only two databases were searched and the language of studies was restricted to English. Further, the authors only provided limited rather than integral detailed information on the methodological quality for each included study.
With reference to the problems mentioned above, this review will therefore further assess the accuracy of all potential simple tests for screening DPN to supply more comprehensive evidence.
Objectives
To determine the diagnostic accuracy of each simple test as triage to screen for diabetic peripheral neuropathy (DPN) involving limbs within different settings, or as replacement of nerve conduction studies (NCS) for the clinical diagnosis of DPN involving limbs, with NCS as the reference standard.
Secondary objectives
To estimate the relative accuracy of simple tests for screening DPN involving limbs, with NCS as the reference standard.
To assess the impact of potential sources of heterogeneity on the performance of simple tests for DPN involving limbs: (1) related to the study population (spectrum of the disease: with versus without other vascular complications; symptoms of DPN: people with no neurological symptoms versus neurological symptoms (if available, positive versus negative neurological symptoms); duration of diabetes; level of glycosylated haemoglobin A1c (HbA1c) in adults: < 7% versus ≥ 7%; body mass index (BMI) in adults: < 25 versus ≥ 25 kg/m²; types of diabetes: type 1 versus type 2 diabetes mellitus; age: < 18 years old versus ≥ 18 years old); (2) related to the simple tests (different thresholds; body sites tested; numbers of body sites tested; types of instrument; examiner's expertise: specialists in diabetes or neurology versus other healthcare professionals); (3) related to the reference standard (numbers of body sites tested with NCS; examiner's expertise: specialists in electrodiagnosis versus other healthcare professionals); (4) related to the healthcare setting (community versus outpatient setting versus inpatient setting); (5) related to the methodology based on the QUADAS‐2 items (risk of bias for patient selection, index test, reference standard, and flow and timing; concerns regarding applicability of patient selection, index test, and reference standard).
Methods
Criteria for considering studies for this review
Types of studies
Prospective and retrospective single‐gate studies (that is 'cohort type accuracy studies') (Deeks 2009; Rutjes 2005) and studies with fully paired or randomised comparison design (Bossuyt 2008) will be eligible, regardless of language of publication.
Participants
People with type 1 or type 2 diabetes, who are being screened for neuropathy, regardless of age and gender. We will exclude those who already have overt neuropathy with foot ulcers and other related manifestations.
Index tests
Any of the following (but not limited to) simple tests including ankle reflex test, light touch sensation tests (SWMT, Neuropen, IpTT, von Frey's hairs), vibratory sensation tests (128‐Hz standard tuning fork, graduated tuning fork, VibraTip, electromechanical instruments for VPT), TCD and steel ball‐bearing test.
Target conditions
We will focus on DPN that involves limbs. Any stage of DPN will be dichotomised as ’no DPN’ versus ’DPN of any stage’, which may include mild, moderate and severe DPN. Where different classifications were used in primary studies, we will require and convert the data according to our stage dichotomous classification.
Reference standards
We will include studies in which nerve conduction studies (NCS) have been applied solely as the reference standard. We will specify and critically consider the possible differences of reference standards among all the eligible studies in our review.
Search methods for identification of studies
Electronic searches
We will use the following sources from inception to present time for the identification of studies.
-
The Cochrane Library
-
MEDLINE (Ovid)
-
EMBASE (Ovid)
-
MEDION
-
ARIF
-
Web of Science: Science Citation Index Expanded
-
Trials registers:
-
The metaRegister of Controlled Trials (www.controlled‐trials.com/mrct/)
-
The EU Clinical Trials register (www.clinicaltrialsregister.eu/)
-
The US National Institutes of Health trials register ClinicalTrials.gov (www.clinicaltrials.gov)
-
The Australian New Zealand Clinical Trials Registry (www.anzctr.org.au)
-
The WHO International Clinical Trials Registry Platform Search Portal (apps.who.int/trialsearch/)
-
-
Grey literature:
-
The Grey Literature Report (www.greylit.org/)
-
OpenGrey (www.opengrey.eu/)
-
OAISTER (www.oaister.org/)
-
For detailed search strategies please see Appendix 2. We will use Web of Science for forward citation tracking of early key publications. We will continuously apply PubMed's 'My NCBI' (National Center for Biotechnology Information) email alert service for identification of newly published studies using a basic search strategy (see Appendix 2). Four weeks before we submit the final review draft to the Cochrane Metabolic and Endocrine Disorders Group (CMED) for editorial approval, we will perform a complete update search on all specified databases. Should we detect new studies for inclusion, we will evaluate these and incorporate findings in our review before submission of the final review draft.
If we detect additional relevant key words during any of the electronic or other searches we will modify the electronic search strategies to incorporate these terms and document the changes. We will place no restrictions on the language of publication when searching the electronic databases or reviewing reference lists in identified studies.
We will send results of electronic searches to Cochrane Metabolic and Endocrine Disorders Group for databases that are not available at the editorial office.
Searching other resources
We will try to identify other potentially eligible studies or ancillary publications by searching the reference lists of retrieved included studies, (systematic) reviews, meta‐analyses, and health‐technology assessment reports.
We will try to contact manufacturers of related index tests to identify studies.
Data collection and analysis
Selection of studies
To determine the studies to be assessed further, two review authors (ZY, YZ) will independently scan the abstract, titles or both sections of every record retrieved. All potentially relevant articles will be investigated as full text. Differences will be resolved by a third party (YH). If resolving disagreement is not possible, the article will be added to those 'awaiting assessment' and authors will be contacted for clarification. An adapted PRISMA (preferred reporting items for systematic reviews and meta‐analyses) flow‐chart of study selection (Figure 2) will be attached (Liberati 2009; Moher 2009).
Flow chart of study inclusion
Data extraction and management
Two review authors (ZY, YH) will independently extract data concerning details of study design, study population, comparator test(s), index test(s) and their performance using standard data extraction templates (Table 3; Appendix 3; Appendix 4; Appendix 5; Appendix 6) with any disagreements to be resolved by discussion, or if required by a third party (LJ).
| Characteristic Study ID | Eligible | Recruited into the study | Received index test | Received reference standard | Included in the analysis | Lost to follow‐up |
| Study 1 | ||||||
| Study 2 | ||||||
| Study 3 | ||||||
| Study 4 | ||||||
| Total | ... | ... | ... | ... |
"‐" denotes not reported.
The following items will be included.
General information: published/unpublished, title, authors, country, language of publication, year of publication, sponsoring, setting, prevalence in study centre(s).
Participants: sampling (consecutive/convenience), inclusion criteria, exclusion criteria, total number and number in comparison groups, sex, age, ethnicity, BMI, type of diabetes mellitus, duration of diabetes mellitus, glycaemic control, antihyperglycaemic treatment, symptoms, severity of target condition.
Index test: type of test, type of device, diagnostic criteria, sites investigated, additional sources of clinical material, qualification of assessor.
Reference test: type of test, type of device, diagnostic criteria, sites investigated, additional sources of clinical material, qualification of assessor.
Results: number of true positives, false positives, true negatives, false negatives, adverse events.
We will send an email request to contact persons of included studies to enquire whether authors are willing to answer questions regarding their studies. The results of this survey will be published in Appendix 7. Thereafter, we will seek relevant missing information on the study from the study author(s) of the article, if required. For example, when we find studies with a limb rather than a patient as the unit of analysis, we will email the authors to ask whether they have summary data on the individual, as the unit of analysis in our review is the whole person.
Dealing with duplicate publications and companion papers
In the case of duplicate publications and companion papers of a primary study, we will try to maximise the yield of information by simultaneous evaluation of all available data but we will not include the same group of patients more than once in any given analysis.
Assessment of methodological quality
In the QUADAS‐2 (quality assessment of diagnostic accuracy studies) instrument, quality is defined as both the risk of bias and applicability of a study, i.e. "1) the degree to which estimates of diagnostic accuracy avoided risk of bias, and 2) the extent to which primary studies are applicable to the review's research question" (Whiting 2011). We will complete the assessment in four phases with QUADAS‐2 as required (Whiting 2011).
We will assess the applicability of a study and risk of bias. Two review authors (ZY, RC) will independently rate each of the four key domains (patient selection, index test(s), reference standard, flow and timing) using signalling questions (Appendix 8). Possible disagreement will be resolved by consensus, or with consultation of a third author (SZ) in case of disagreement.
We have followed the process for tailoring QUADAS‐2 to our systematic review as described in the publication (Whiting 2011) by omitting a signalling question in the domain of patient selection and adding one signalling question respectively in the domain of index test, and flow and timing. We have also developed preliminary review‐specific guidance on how to assess each signalling question to judge risk of bias. We will pilot the tool and apply criteria in a small number of studies by at least two authors (ZY, RC). If agreement is not good, we will add further refinement to the tool. We will use our guidelines to judge risk of bias as 'low', 'high' or 'unclear'.
We will primarily analyse studies at low risk of bias, low concern regarding applicability or both for all or specified domains. We will explore the influence of individual criteria in a sensitivity analysis.
We will present a 'Risk of bias and applicability concerns' figure and a 'Risk of bias and applicability concerns summary' figure.
Statistical analysis and data synthesis
The unit of analysis is a patient rather than a limb or a part of limb. Data for the true positive, true negative, false positive and false negative values for each study will be tabulated. Test results will be treated as positive or negative for the cut‐off values of the index tests as described above. Forest plots showing pairs of sensitivity and specificity, with 95% confidence intervals (CI) will be constructed for each study. The sensitivity and specificity pairs will be visualised in the receiver operator characteristic (ROC) space for each test.
Our primary analyses will compare each simple test with the reference standard. As we expect that, for each simple test, the number of investigated sites may vary and thus various thresholds will have been used across studies, we will consider using the hierarchical summary receiver operating characteristics (HSROC) model (Rutter 2001) to estimate a summary ROC curve. In case of multiple thresholds in an individual study, we will report accuracy estimates for all the thresholds. For the HSROC model, we will give priority to the pre‐specified threshold, while in the absence of the pre‐specified threshold we will then choose the most common one across the studies. We will use SAS software to fit the HSROC model. Results from the hierarchical models will be input into Review Manager 5.2 to provide plots of the estimated curve(s), or summary point(s) and confidence region(s).
Secondly, we will focus on the comparative accuracy of the simple tests with the reference standard. We will use the HSROC model within SAS software to conduct indirect and direct comparisons separately (Rutter 2001), in which case from each included study we will also choose the same threshold as is used in our primary analyses. We will assess the fit of model by likelihood ratio tests comparing models with and without the covariates for shape and accuracy parameters successively (Macaskill 2010). All studies will be included in each pair‐wise indirect comparison. Only those studies that make a direct fully paired or randomised comparison will be included in the direct comparisons.
Investigations of heterogeneity
We will investigate heterogeneity by visual inspection of forest plots and ROC curves. Given adequate amount of data (10 or more studies for one index test), we will investigate heterogeneity within SAS environment by adding the covariates specified under Secondary objectives as potential determinants or sources of heterogeneity to the HSROC model to identify statistically significant covariates.
For individual patients' characteristics such as age, metabolic control, effect of treatment, and duration of diabetes, we will first extract and analyse stratified accuracy results (for example, results separately of the subgroup of less than 18 years old versus the subgroup 18 years and older), if available within a study. However, if not available, we will convert the covariates age, effect of treatment to percentages and the covariate duration of diabetes to mean and then add the converted numerical covariates to the model.
Sensitivity analyses
We will perform sensitivity analyses to explore the impact of study quality on the meta‐analytic results.
-
Restricting the analyses to the studies with either low risk of bias or low concerns regarding applicability in each domain of the QUADAS‐2 instrument.
-
Restricting the analysis to the studies with prospective design.
-
Restricting the analysis taking account of three individual quality items: blinding of reference standard results, blinding of index test results and interval of less than two months between index tests and reference test.
-
Restricting the analysis on the data sources (published versus unpublished).
Assessment of reporting bias
We will not undertake any formal assessment of reporting bias in our review due to current uncertainty about how to assess reporting bias in diagnostic test accuracy reviews, especially in the presence of heterogeneity (Macaskill 2010).
Flow chart of screening and diagnosis of DPN of limbs in clinical practice
Screening tests could be tests to identify symptoms and/or signs
Flow chart of study inclusion
| Classificationa | Subgroup |
| Generalised symmetric polyneuropathies | Chronic sensorimotor (typical DPN) |
| Acute sensory | |
| Autonomic | |
| Focal and multifocal neuropathies | Cranial |
| Truncal | |
| Focal limb | |
| Proximal motor (amyotrophy) | |
| Co‐existing CIDP | |
| aAccording to Boulton et al. (Boulton 2005a); | |
| Stage of diabetic peripheral neuropathy | Characteristics |
| Stages 0/1: no clinical neuropathy | |
| No symptoms or signs | |
| Stage 2: clinical neuropathy | |
| Chronic painful | Positive symptomatology (increasing pain at night): burning, shooting, stabbing pains ± pins and needles |
| Absent sensation to several modalities and reduced or absent reflexes | |
| Acute painful | Less common |
| Diabetes poorly controlled, weight loss | |
| Diffuse (trunk) | |
| Hyperaesthesia may occur | |
| May be associated with initiation of glycaemic therapy | |
| Minor sensory signs or even normal peripheral neurological examination | |
| Painless with complete/partial sensory loss | No symptoms or numbness/deadness of feet; reduced thermal sensitivity; painless injury |
| Signs of reduced or absent sensation with absent reflexes | |
| Diabetic amyotrophy | Muscle weakness and wasting |
| Sensory loss is slight, but pain at night common | |
| Subacute onset | |
| Stage 3: late complications of clinical neuropathy | |
| Foot lesions, e.g. ulcers | |
| Neuropathic deformity, e.g. Charcot joint | |
| Non‐traumatic amputation | |
| Characteristic Study ID | Eligible | Recruited into the study | Received index test | Received reference standard | Included in the analysis | Lost to follow‐up |
| Study 1 | ||||||
| Study 2 | ||||||
| Study 3 | ||||||
| Study 4 | ||||||
| Total | ... | ... | ... | ... | ||
| "‐" denotes not reported. | ||||||
