Figure 1. Percentage of tests indicated by provider type for (A) the initial evaluation of distal symmetric polyneuropathy (scenario 1) and (B) the full evaluation of distal symmetric polyneuropathy (scenarios 1 and 2). EMG indicates electromyogram/nerve conduction study; GTT, glucose tolerance test; MRI, magnetic resonance imaging; SPEP, serum protein electrophoresis. * P < .001. † P < .05.
Figure 2. Percentage of American Academy of Neurology (AAN)–recommended tests (fasting blood glucose test, serum protein electrophoresis, B12 test, and glucose tolerance test) indicated by provider type for the full evaluation of distal symmetric polyneuropathy.
Callaghan BC, Kerber K, Smith AL, Fendrick AM, Feldman EL. The Evaluation of Distal Symmetric PolyneuropathyA Physician Survey of Clinical Practice. Arch Neurol. 2012;69(3):339-345. doi:10.1001/archneurol.2011.1735
Author Affiliations: Departments of Neurology (Drs Callaghan, Kerber, and Feldman and Ms Smith) and Internal Medicine (Dr Fendrick), University of Michigan, Ann Arbor.
Objective To define current clinical practice for evaluating distal symmetric polyneuropathy.
Design Using a modified Dillman method, we sent surveys to 600 internists, 600 neurologists, and 45 neuromuscular specialists selected from the American Medical Association Physician Masterfile. Survey questions pertained to which tests providers would order in the following 3 scenarios: (1) the initial evaluation of distal symmetric polyneuropathy, (2) the use of additional tests if the initial evaluation was unrevealing, and (3) patients with diabetes. The t test was used to compare the number of tests ordered by physician type, and the χ2 test was used to compare proportions of tests ordered.
Setting National survey of physicians.
Participants Internists, neurologists, and neuromuscular specialists.
Results The response rate was 35%. Overall, many tests were ordered for the full evaluation of distal symmetric polyneuropathy (mean [SD], 16.5 [7.2] tests), and there was substantial variation within and between provider types. Internists ordered fewer tests (mean [SD], 14.5 [6.1] tests) than did neurologists (mean [SD], 17.5 [7.9] tests) (P < .001). Regarding the glucose tolerance test, substantial differences were found between physician types, with neurologists and neuromuscular specialists ordering this test more frequently (28.6% and 72.3%, respectively) and internists ordering it less frequently (4.1%). A brain and/or spine magnetic resonance imaging scan was ordered by 19.8% of internists and 12.9% of neurologists.
Conclusions From the supporting evidence, current practice intent on evaluating distal symmetric polyneuropathy is highly variable and differs widely. For this disorder of the peripheral nerves, a high-yield test such as the glucose tolerance test is rarely used, whereas magnetic resonance imaging is likely overused. Research that defines the optimal evaluation of distal symmetric polyneuropathy has the potential to result in more efficient care.
Quiz Ref IDPeripheral neuropathy is a common and disabling condition that is diagnosed and evaluated by both internists and neurologists.1- 3 Distal symmetric polyneuropathy (DSP) is the most common type of neuropathy, accounting for the majority of cases.4,5 Many underlying disorders cause or are associated with DSP, with diabetes leading the list.4,6 Disappointingly, despite an exhaustive evaluation, many patients are left without a definitive diagnosis.7,8
Quiz Ref IDA systematic review9 performed by the American Academy of Neurology (AAN) found that the fasting blood glucose test, serum protein electrophoresis (SPEP), and the B12 test have the highest yield in the evaluation of DSP. The practice parameter statement accompanying this review9 also recommended that, if the fasting blood glucose level is normal, then the glucose tolerance test (GTT) may be considered, especially for those with a sensory and/or painful neuropathy. However, this review9 did not address the use of many other commonly ordered laboratory tests for this condition because the information is lacking on their utility.
In our study, we sought to investigate physician practices for evaluating DSP. One of the goals was to identify the degree of variation of practice because substantial variation offers compelling evidence of the opportunity to improve efficiencies in health care.10 Furthermore, we wanted to identify tests that are inappropriately used based on current evidence and to define which tests are in need of further study. The ultimate goal is to define the most efficient way to evaluate patients with this common, disabling condition.
We developed an 8-question survey that pertains to the diagnostic evaluation of DSP. A comprehensive list of tests ordered for this condition was created on the basis of an extensive literature review. These 47 tests included the following categories: hematology/chemistry, diabetes, vitamins, paraprotein, rheumatology, infections, immunology, radiology, electrodiagnosis, and pathology. The survey presented physicians with 3 common clinical scenarios (http://jdrf.neurology.med.umich.edu/Brian/NeuropathySurvey.doc) and asked them to select the tests that they would order for each scenario. The first scenario was with regard to the initial evaluation of DSP with classic description of the condition, no clear cause from the clinical history, and the label of DSP. The second scenario was with regard to additional testing if the initial evaluation was unrevealing. The third scenario was with regard to the evaluation specifically in patients with DSP and a history of diabetes. For each test, we also asked physicians to report which tests they considered reliable and valid.
The American Medical Association Masterfile was used to select physicians for the survey. We first excluded retired physicians, those with addresses outside the United States, and those with pediatric specialties and subspecialties. A random sample of 600 internists, 600 neurologists, and 45 neuromuscular specialists was selected. However, self-identification of specialty by responders was used for classification of provider type. For our power calculation, we assumed a 35% response rate. Anticipating 210 measurable responses per group would provide 95% CIs of ±4% to 6% for binary response variables with 86% power to detect 15% point differences between physician subgroups.11
The survey was mailed to each selected physician. A modified Dillman method was used that entailed 4 mailings separated by 2-week intervals.12 The first mailing was a pre–notice letter followed by the survey with a $2 incentive. The third mailing was a reminder postcard followed by a replacement survey. Demographics on nonresponders were obtained from the American Medical Association Masterfile.
Statistical analyses comparing internist vs neurologist responses were based on t tests (unpaired, 2-sided) and χ2 tests. Negative binomial regression models were used to evaluate for associations between provider characteristics and number of all tests ordered. Additionally, logistic regression models were designed to examine the association between provider characteristics and those who ordered 3 or 4 of the AAN-recommended tests compared with those who did not. All statistical analyses were performed with SAS version 9.2 (SAS Institute Inc, Cary, North Carolina). The University of Michigan institutional review board approved our study.
The overall response rate was 34% (429 of 1245 physicians). After excluding those with returned mailings due to wrong address, the participation rate was 39%. Neurologists were more likely to respond than internists (P < .001) (Table 1). All other demographic variables were similar between nonresponders and responders, including age, sex, medical degree, years since graduation, region of the country, and metropolitan statistical area size.
The mean (SD) age of responders was 47.4 (11.6) years, and 29.2% were women (Table 1). The mean (SD) number of years since graduation from medical school was 19.8 (12.1) years. Of the 429 returned surveys, 210 (49.0%) were completed by internists, 172 (40.1%) by neurologists, and 47 (11.0%) by neuromuscular specialists (19 physicians were classified on the basis of the American Medical Association Masterfile, and 28 physicians were classified on the basis of survey self-identification). Most respondents (94.6%) had doctor of medicine degrees. The distribution between regions of the country was evenly distributed; however, the vast majority of physicians practiced in areas with a metropolitan size of greater than 250 000 people. Most respondents (61.1%) reported that patients with neuropathy account for 1% to 10% of their patients. Another 35.1% of respondents reported that more than 10% of the patients they see have neuropathy. Regarding practice setting, 43.7% of respondents work in a private practice setting, 28.6% at a university, and 11.0% in a nonuniversity teaching hospital.
Of the 47 tests presented in the survey for the evaluation of DSP, a mean (SD) number of 10.6 (4.5) tests were ordered by internists, compared with a mean (SD) number of 12.2 (5.4) tests ordered by neurologists (P = .002) and a mean (SD) number of 13.3 (5.4) tests ordered by neuromuscular specialists (all comparisons are between internists and neurologists) (Table 2). Tests that were frequently ordered by all 3 physician groups included those that measured B12 levels (93.0%), complete blood cell count (85.6%), thyrotropin (85.6%), and hemoglobin A1C (71.8%) and the panel 7 test (84.9%) and liver function test (62.0%). There was a significant disparity regarding the use of the GTT, with 2.3% of internists indicating intent to order this test compared with 16.7% of neurologists (P < .001) and 34.0% of neuromuscular specialists (Figure 1A). Similarly, intent to order electromyograms/nerve conduction studies was substantially higher among neurologists and neuromuscular specialists (74.8% and 89.4%, respectively) than among internists (28.5%) (P < .001). Neurologists ordered SPEP and immunofixation more frequently than did internists (68.1% vs 28.5% and 36.7% vs 7.6%, respectively; P < .001). Internists ordered the following tests more frequently than did neurologists: fasting blood glucose test (76.7% vs 58.1%), urinalysis (37.2% vs 10.0%), and chest radiography (14.0% vs 7.6%). Neurologists ordered tests for rheumatologic and infectious diseases more frequently than did internists, with the exception of human immunodeficiency virus and hepatitis B and C viruses, but only tests for the erythrocyte sedimentation rate and antinuclear antibody were indicated in more than 50% of patients (62.9% of neurologists indicated that they would order an erythrocyte sedimentation rate test, and 51.9% of neurologists indicated that they would order an antinuclear antibody test). Very few physicians of any type ordered lumbar puncture (1.9%), skin biopsy (1.6%), sural nerve biopsy (1.2%), magnetic resonance imaging (MRI) of the brain (1.6%) or spine (3.7%), genetic testing (0.5%), or immunologic tests (anti-GM1 [3.3%], antisulfatide [1.9%], anti–myelin-associated glycoprotein [4.2%], paraneoplastic antibodies [4.4%], and celiac screen [3.7%]) for the initial evaluation of DSP.
When the initial workup of DSP is negative, internists indicated that they would order a mean (SD) total of 14.5 (6.1) tests for the full evaluation of DSP, and neurologists indicated that they would order a mean (SD) total of 17.5 (7.9) tests for the full evaluation of DSP (P < .001); neuromuscular specialists indicated that they would order a mean (SD) total of 18.7 (6.4) tests (Table 2). In this scenario, the disparity in use of the GTT by physician type increased because only 4.1% of internists would order the test compared with 28.6% of neurologists (P < .001) and 72.3% of neuromuscular specialists. Nearly 100% of physicians would order a fasting blood glucose test, a hemoglobin A1C test, or a GTT (Figure 1B).
Nearly 1 in 5 internists (ie, 19.8%) indicated intent to order an MRI scan of the brain and/or spine for the full evaluation of DSP (scenarios 1 and 2) (8.1% ordered MRI brain scans, and 15.7% ordered MRI spine scans). Use of MRI was less frequent among neurologists (12.9% total, with 3.3% ordering MRI brain scans and 11.9% ordering MRI spine scans) and neuromuscular specialists (8.5%). Electromyograms/nerve conduction studies were ordered by 67.4% of internists, 92.9% of neurologists (P < .001), and 100% of neuromuscular specialists. The tests that were rarely ordered as part of the initial evaluation but that were much more frequently ordered after the initial test result was normal were as follows: heavy metal tests (36.4% of all physicians), all immunologic tests (13%-30% of neurologists), genetic tests (10.0% of neurologists), skin biopsies (15.2% of neurologists), sural nerve biopsies (13.8% of neurologists), and lumbar puncture (12.9% of neurologists).
Considering the AAN-recommended tests (fasting blood glucose test, SPEP, B12 test, and GTT), only 2.3% of internists and 17.1% of neurologists indicated they would order all 4 tests (P < .001). Approximately 40% of physicians ordered 3 tests, and an additional 30% to 40% would order 2 (Figure 2).
Internists indicated that they would order a mean (SD) number of 6.5 (4.4) tests, neurologists a mean (SD) number of 6.6 (4.8) tests (P = .94), and neuromuscular specialists a mean (SD) number of 9.1 (5.6) tests (Table 2). All physicians frequently indicated that they would order a complete blood cell count, a panel 7 test, a liver function test, and a thyrotropin test. Neurologists and neuromuscular specialists (42.9% and 76.6%, respectively) were much more likely to order an electromyogram/nerve conduction study than were internists (22.1%) (P < .001). Neurologists were also more likely than internists to order a B12 test (72.4% vs 57.0%), SPEP (31.0% vs 8.1%), and immunofixation (18.6% vs 3.5%). Internists were much more likely than neurologists to order a hemoglobin A1C test (91.9% vs 68.1%), a urinalysis (25.6% vs 6.7%), and/or a chest radiography (6.4% vs 1.9%). The remainder of tests were ordered only rarely.
More internists than neurologists considered the GTT to be unreliable (12.0% vs 5.2%) and invalid (6.4% vs 2.6%). Other tests for which more than 15% of physicians had significant concerns about reliability and validity included those that measured the erythrocyte sedimentation rate, C-reactive protein level, vitamin E level, angiotensin-converting enzyme, and the presence of Lyme disease and MRI of the brain and spine. Physicians were also concerned with the validity of the rheumatoid factor and antinuclear antibody tests. More internists than neurologists indicated concern regarding the validity of skin biopsy (10.2% vs 4.6%) and sural nerve biopsy (9.6% vs 4.7%).
To determine which provider characteristics are associated with the number of tests ordered and the number of AAN-recommended tests ordered, 2 regression frameworks were used (Table 3). Variables included age, years since graduation, sex, metropolitan statistical area size where the physician practices, region of the country, specialty, medical degree, practice setting, and frequency of patients with neuropathy in the physician's practice. For total number of tests ordered for the full evaluation, there was a significant association with specialty because neurologists were 17% more likely to order tests (odds ratio, 1.17 [95% CI, 1.02-1.34]) in fully adjusted regression models. Similarly, for the number of AAN-recommended tests ordered, the only significant association was physician specialty, with more neurologists (odds ratio, 1.73 [95% CI, 1.08-2.75]) and neuromuscular specialists (odds ratio, 5.46 [95% CI, 1.69-17.68]) than internists ordering 3 or 4 AAN-recommended tests. None of the other provider characteristics were significantly associated with the total number of tests or the total number of AAN-recommended tests ordered.
Many tests are ordered for the evaluation of patients with DSP, and there is substantial variation in care within and between physician groups. Quiz Ref IDDespite the high numbers of tests being ordered, the tests with the highest diagnostic yield are frequently omitted. Furthermore, the GTT (a high-yield, low-cost test) is rarely used, whereas MRI (a low-yield, high-cost test) is likely overused.13- 16 More research is needed to define the most effective and efficient approach for the evaluation of DSP.
Considerable variation within physician groups is demonstrated by the large standard deviations for the numbers of tests ordered under each scenario. There is also substantial variation between physician groups, as revealed by the difference in test use between internists and neurologists. A recent report10 from the Institute of Medicine identified scientific uncertainty as one of the key drivers of rising health care costs in the United States and suggested consistency of treatment and/or evaluation as a possible solution. Furthermore, high utilization of resources does not necessarily lead to better health outcomes; in fact, the opposite has been observed.17 Therefore, efforts to identify the reasons for the variation in the care of these patients are essential.
In all 3 clinical scenarios presented, neurologists ordered significantly more tests than did internists. In fully adjusted regression models, the only variable significantly and independently associated with ordering more tests was specialty, with neurologists ordering more tests compared with internists. The fact that neurologists order more tests is not surprising because specialists tend to order more tests and because the patients they evaluate tend to be more complicated with more severe symptomatology.
In addition to ordering a higher number of overall tests, neurologists also ordered more of the high-yield tests (fasting blood glucose test, SPEP, B12 test, and GTT), which are supported by recent AAN diagnostic recommendations for DSP. However, only a small number of physicians order all 4 tests, with 57.6% of internists and 41.0% of neurologists ordering 2 or less. Of these tests, the GTT showed the most disparity between provider types. Although only a few internists indicated that they would order this test, almost a third of neurologists and more than 70% of neuromuscular specialists include the GTT as part of the full workup of DSP.9 This low level of use of the GTT may, in part, be due to the controversy surrounding impaired glucose tolerance (IGT) as a cause of neuropathy.14,16,18- 20 Furthermore, although the Diabetes Prevention Program demonstrated that an intense diet and exercise regimen helped prevent progression to diabetes in this population, this intervention is not easily available to most patients, and there is no definitive data on prevention or treatment of neuropathy.21Quiz Ref IDOn the other hand, 2 studies14,16 have shown that patients with idiopathic neuropathy have a much higher prevalence of IGT than expected based on literature-based controls. These results, in conjunction with work demonstrating improvement in nerve fiber density in patients with IGT and neuropathy who follow a diet and exercise regimen,22 support IGT as a potentially common and treatable cause of neuropathy. However, more definitive studies are needed to firmly establish IGT as a cause of neuropathy and to make more conclusive recommendations regarding the use of the GTT as a screening test in patients with DSP.
Surprisingly, a significant number of physicians indicated that they would order an MRI of the brain and/or spine in the full workup of a patient with neuropathy. In the survey scenarios, physicians were presented with a standard description of DSP with the diagnosis, and still MRIs were ordered by 1 out of 6 physicians. Internists were much more likely to indicate ordering MRI scans, which may, in part, result from a lack of confidence in their ability to localize symptoms to the peripheral nerves. However, a higher-than-expected proportion of neurologists also ordered MRI scans. Although structural lesions of the brain and/or spine can mimic distal symmetric neuropathy, the survey scenarios were designed to specifically discourage concern for any disorder other than DSP. Because DSP affects the peripheral nervous system, imaging of the central nervous system does not seem to be indicated for this patient population and thus may represent a target for future efficiency improvements.
Electromyography was a frequently ordered test by all physician types in the evaluation of DSP. Neurologists were more likely to order this as part of the initial workup and for patients with diabetes, but most physicians believe that this test should be a part of the full evaluation of DSP. Its high level of use is likely because electrodiagnostic studies can provide more objective evidence of neuropathy and can identify those with demyelinating features. However, given that this test has important downsides (eg, cost, time, and patient discomfort), further research is needed to define the role of this diagnostic test in routine DSP presentations.
Serum protein electrophoresis and immunofixation are also ordered at differing rates depending on specialty. Neurologists are much more likely to order both tests, whereas internists rarely order an immunofixation test. By itself, SPEP is much less sensitive than when performed in conjunction with an immunofixation, and, therefore, both tests should be performed when there is suspicion of an underlying bone marrow disorder.23- 25 Evidence exists that patients with neuropathy have an increased prevalence of abnormal SPEP results, especially patients with demyelinating neuropathy.26 However, neuropathy is a common condition, and current evidence is unclear regarding the use of routine SPEP and immunofixation testing in all patients with DSP.
Despite variations in clinical practice within this patient population, there are many tests that most physicians order. These include a complete blood cell count, a panel 7 test, a liver function test, a thyrotropin test, and a B12 test and at least 1 test to evaluate for diabetes (fasting blood glucose test, hemoglobin A1C test, and/or GTT). Although no evidence exists to support the use of a complete blood cell count, a panel 7 test, a liver function test, or a thyrotropin test, these tests have seemingly become the standard of care for patients with neuropathy. Future investigations are needed to confirm the value of these tests. In contrast, there is little consensus regarding rheumatologic, immunologic, infectious disease, genetic, and vitamin level testing or regarding skin and sural nerve biopsies. Most physicians indicate that these are not needed as part of the routine initial evaluation of patients with DSP, but there is still substantial use of these tests in the full workup. Future investigations are needed to clarify the role of these tests.
Quiz Ref IDThe most common cause of DSP is diabetes, and physicians clearly understand the need for testing for this condition. However, given that DSP and diabetes are both common conditions, the question remains which tests to order for patients with both conditions. In the scenario of patients with DSP and known diabetes, more neurologists than internists indicated that they would order SPEP and the B12 test more frequently. As stated earlier, they also believe that electrodiagnostic tests are valuable in this patient population. Determining the appropriate evaluation for this population is important given that diabetes is by far the leading cause of DSP.4,6
Our study has important limitations. Our findings are based on the self-reported intent to order tests, not claims data. Thus, we cannot account for the actual actions of physicians. However, the advantage of our study was the use of typical and standardized scenarios, which substantially reduces the potential confounding that can occur when assessing claims data for use in DSP. On the other hand, the limited clinical information in our survey scenarios may not have adequately accounted for the diverse presentations that occur in routine care. The brief scenarios also did not allow us to assess the effect of other important clinical aspects on test use. Our survey response rate was less than 40%, which limits the generalizability of our findings. However, we did have a large number of responders, and responders and nonresponders were similar in our available measures. The low number of neuromuscular specialists makes inferences about this physician group tenuous. We also investigated many comparisons without corrections. However, these results are meant to generate hypotheses, and many of our results revealed P values of less than .001. Finally, despite the recent summary research on the value of using these tests to diagnose the cause of DSP, no current research uses an optimal methodology to define the value of these tests by the effect of these tests on meaningful health outcomes.
In conclusion, we found significant variation in the evaluation of DSP, and the tests with the best evidence to support their use are often not ordered by physicians. The GTT is rarely ordered, particularly by internists, despite being one of the tests with the highest yield. This finding may be the result of the fact that there is controversy about whether IGT causes DSP.14,16,18- 20 In contrast, MRI scans are used frequently without a clear indication. More research is needed for physicians to be able to evaluate patients with DSP in an effective and efficient manner.
Correspondence: Brian C. Callaghan, MD, Department of Neurology, University of Michigan, 109 Zina Pitcher Pl, 4021 BSRB, Ann Arbor, MI 48104 (firstname.lastname@example.org).
Accepted for Publication: August 23, 2011.
Published Online: November 14, 2011. doi:10.1001/archneurol.2011.1735
Author Contributions:Study concept and design: Callaghan, Kerber, Fendrick, and Feldman. Acquisition of data: Callaghan and Smith. Analysis and interpretation of data: Callaghan, Kerber, Fendrick, and Feldman. Drafting of the manuscript: Callaghan. Critical revision of the manuscript for important intellectual content: Kerber, Smith, Fendrick, and Feldman. Statistical analysis: Callaghan and Kerber. Obtained funding: Feldman. Administrative, technical, and material support: Smith. Study supervision: Callaghan, Fendrick, and Feldman. Dr Callaghan was also involved in development and distribution of the survey.
Financial Disclosure: Dr Kerber received speaker honoraria from the American Academy of Neurology 2010 and 2011 annual meetings and performed consulting work for the American Academy of Neurology. Dr Fendrick serves as a consultant to Abbott, ActiveHealth Management/Aetna, AstraZeneca, Avalere Health, BlueCross BlueShield Association, Blue Shield of California, the Center for Medicare and Medicaid Services, GlaxoSmithKline, the Health Alliance Plan, Hewitt Associates, Highmark BlueCross BlueShield, Integrated Benefits Institute, MedImpact HealthCare Systems Inc, Merck, the National Business Coalition on Health, the National Pharmaceutical Council, Perrigo, Pfizer, Regence BlueCross BlueShield of Oregon, sanofi aventis, the State of Indiana, Thomson Reuters, TriZetto, UCB, WebMD, and zanzors. He is also on the Speaker's Bureau for Merck. He is also involved in research with Abbott, AstraZeneca, Eli Lilly, Genentech, GlaxoSmithKline, Merck, Novartis, Pfizer, and sanofi aventis.
Funding/Support: Drs Callaghan and Feldman and Ms Smith are supported by the Taubman Medical Institute, the Katherine Rayner Program, and the Program for Neurology Research and Discovery. Dr Callaghan is also supported by a National Institutes of Health T32 training grant (NS007222) and an American Diabetes Association Junior Faculty Award. Dr Kerber was supported by National Institutes of Heatlth/National Center for Research Resources grant K23 RR024009 and by Agency for Healthcare Research and Quality grant R18 HS017690.