LETTERS COMMENTS AND RESPONSES Improvement Interventions TO THE EDITOR: Davidoff's editorial (1) is absolutely correct in calling for the framework and statistics suitable for improvement activities rather than continuing to pretend that service delivery organization can be studied like a drug. If the question is how best can we learn to transition patients from hospital to home safely and with minimal need for short-term rehospitalization, quality improvement (QI) approaches are much more likely to succeed than research trials; they teach us much more about how the local system operates and what else might work, and they do so quickly. However, even if such a project dramatically decreased readmission rates and increased patient confidence and every other good thing that you would want to measure, it would not be published in Annals or any other major medical journal in the United States. It would not be funded by the National Institutes of Health and probably not by the Agency for Healthcare Research and Quality. It would fall entirely outside of the requirements of the Methodology Report for the Patient-Centered Outcomes Research Institute. In most academic institutions, dramatically reducing readmissions would not count toward promotion and tenure—the focus is only on getting those grants and publications. Davidoff's suggestions are a prescription for those concerned with public well-being and policymaker needs, many of whom have come to reject academic medicine with regard to its providing guidance for improving the delivery of health systems. If the point is to improve things, then the method is usually management guided by Shewhart statistics. The findings should be published if they are important for others who are embarking on this journey, especially if they are synthesized over many sites so that readers can understand the role of context and the pace of improvement and the gains can be sustained. Furthermore, the endeavor should be able to be funded and be important to promotion and standing in academic medicine.

Joanne Lynn, MD Center for Elder Care and Advanced Illness, Altarum Institute Washington, DC Disclosures: Authors have disclosed no conflicts of interest. Forms can be viewed at www.acponline.org/authors/icmje/ConflictOfInterest Forms.do?msNum=L14-0568.

Annals of Internal Medicine exactly opposite weaknesses. The RCT cannot adapt, and QI programs cannot conclude that observed outcomes are due to the program under investigation. Our team has an interesting vantage point from which to view this tension: We are simultaneously doing 2 tests of the same care model, CAPABLE (Community Aging in Place—Advancing Better Living for Elders). One is an RCT funded by the National Institutes of Health, and the other is a QI study funded by the Centers for Medicare & Medicaid Services Innovations Center. CAPABLE is designed to address unmet needs and functional difficulties experienced by older adults at home that are not addressed by our health care system despite substantial negative consequences. It includes $1300 worth of home repairs and modifications and up to 10 home visits by occupational therapists and nurses to enable older adults to carry out self-care and discretionary activities of most importance to them. A previous RCT pilot study of CAPABLE found that older adults in the intervention group decreased the number of difficulties associated with activities of daily living by 67% from baseline, whereas those in the control group who received home visits without a functional component decreased these difficulties by only 19% (2). The rapid-cycle evaluation plan of the test of CAPABLE funded by the Centers for Medicare & Medicaid Services Innovations Center requires that we examine quarterly primary outcomes resulting in study protocol changes, including more flexible times to follow-up and enhanced pain management protocols. However, the key components of an RCT—random assignment and a control condition—provide a stronger claim at causality than our QI test. Nonetheless, after completion, the intervention models of all RCTs are adapted to care contexts. Although both approaches have strengths and weaknesses, each could learn from the other. As statistical methods advance and we are better able to take advantage of propensity scoring, stepped wedge designs, and pragmatic trials in RCTs, we may be better able to understand practice contexts and develop adaptive intervention protocols to leverage QI strengths yet make causality claims. Quality improvement may benefit from more rigorous documentation of protocol adaptations and use of single-subject designs. Simultaneously engaging in both approaches also may have important advantages. Sarah L. Szanton, PhD, ANP Bruce Leff, MD Laura N. Gitlin, PhD Johns Hopkins University Baltimore, Maryland

Reference 1. Davidoff F. Improvement interventions are social treatments, not pills [Editorial]. Ann Intern Med. 2014;161:526-7. [PMID: 25285545] doi:10.7326/M14 -1789

TO THE EDITOR: Davidoff (1) highlights the critical tension in

approaches to evaluating health care improvement efforts— the tension between benefits from the rigorous comparison of static-protocol randomized, controlled trials (RCTs) versus those from QI programs that adapt to inevitable contextual challenges during implementation. These approaches have

Disclosures: Disclosures can be viewed at www.acponline.org /authors/icmje/ConflictOfInterestForms.do?msNum=L14-0567.

References 1. Davidoff F. Improvement interventions are social treatments, not pills [Editorial]. Ann Intern Med. 2014;161:526-7. [PMID: 25285545] doi:10.7326 /M14-1789 2. Szanton SL, Thorpe RJ, Boyd C, Tanner EK, Leff B, Agree E, et al. Community aging in place, advancing better living for elders: a bio-behavioralenvironmental intervention to improve function and health-related quality of

596 © 2015 American College of Physicians

Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/933761/ on 07/07/2017

LETTERS life in disabled older adults. J Am Geriatr Soc. 2011;59:2314-20. [PMID:

References

22091738] doi:10.1111/j.1532-5415.2011.03698.x

1. Solberg LI, Mosser G, McDonald S. The three faces of performance measurement: improvement, accountability, and research. Jt Comm J Qual Improv. 1997;23:135-47. [PMID: 9103968] 2. Campbell DT, Stanley JC. Experimental and Quasi-experimental Designs for Research. Boston: Houghton Mifflin; 1963. 3. Hurley D. Chemical imbalance: our high-tech process of pharmaceutical research is broken—and the solution might be old-fashioned trial and error. The New York Times Magazine. 16 November 2014:61-73. 4. Craig P, Dieppe P, Macintyre S, Michie S, Nazareth I, Petticrew M; Medical Research Council Guidance. Developing and evaluating complex interventions: the new Medical Research Council guidance. BMJ. 2008;337:a1655. [PMID: 18824488] doi:10.1136/bmj.a1655 5. Parry GJ, Carson-Stevens A, Luff DF, McPherson ME, Goldmann DA. Recommendations for evaluation of health care improvement initiatives. Acad Pediatr. 2013;13:S23-30. [PMID: 24268081] doi:10.1016/j.acap.2013.04.007

IN RESPONSE: Many serious improvers share Dr. Lynn's frustration with the constraints that true experimental methods (mainly protocol-driven RCTs) impose on the study of improvement work. That frustration can be eased somewhat by remembering the many opportunities other than formal research that are available for developing new knowledge about improvement: exchanging and exploring improvement data with fellow improvers, care providers, and patient groups and reviewing the work with hospital governing boards, accrediting agencies, and funders (1). Improvers and clinicians alike need to take full advantage of the learning and leverage that can come from these less-than-academic pursuits. The failure in many quarters to recognize the professional and intellectual value of improvement work noted by Dr. Lynn is probably an even bigger problem; ditto its counterpart: improvers' resistance to using academic tools, such as theory, in their work. The research and improvement communities have enormous amounts to learn from each other; we are all poorer when each community is slow to accept what the other has to offer. The many undergraduate and graduate programs in improvement appearing rapidly in the United States and elsewhere suggest that considerable numbers of students, residents, and younger physicians are fortunately beginning to take improvement seriously as both an academic and a clinical discipline. Balancing the strengths of 1 method against the limitations of another is a thoroughly rational research strategy that Donald Campbell and Julian Stanley recommended 50 years ago in their classic monograph on research design (2). Dr. Szanton and colleagues have adopted this balanced approach in evaluating their improvement work. Their approach sounds promising and worth exploring further but also demands considerable extra time, effort, and funding. The power and visibility of RCTs makes them a dominant element in drug development. However, those trials are just the last step in a convoluted and uncertain process involving discovery, testing, and scale-up, most of which is hidden from public view. Ironically, this long, drawn-out process—with its mixture of theory and hunches, trial-and-error learning cycles, discovery and adaptation, and confusion and reorientation (3)—provides a useful model for new and more sophisticated ways to evaluate complex improvement interventions. In these nuanced and eclectic proposals, particular evaluation methods are tailored to the sequential phases of improvement efforts (4, 5). Randomized, controlled trials clearly play an important part in the evaluation of improvement work, but, as in drug development, they are only a part of the overall picture, necessary but not sufficient for reaching a deep understanding of the nature and effect of these complex interventions.

Frank Davidoff, MD Geisel School of Medicine at Dartmouth College Hanover, New Hampshire Disclosures: The author has disclosed no conflicts of interest. The form can be viewed at www.acponline.org/authors/icmje/ConflictOf InterestForms.do?msNum=M14-1789. www.annals.org

Physician Drug Testing Is Unscientific and Is Unlikely to Achieve Stated Aims TO THE EDITOR: Pham and Pronovost (1) rightly criticize Cal-

ifornia's failed Proposition 46. Although their desire to identify impaired physicians and improve patient safety is laudable, it cannot be achieved through random, preemployment, notfor-cause, and possibly even sentinel event– based drug and alcohol testing. This last type of testing will miss alcoholics in withdrawal and, when done after the event, may produce false-negative results for many agents. The explosive growth of such testing in industry and health care has been fueled by popular misconceptions surrounding substance use and abuse, junk science, and business interests (Institute for a Drug-Free Workplace, pharmaceutical firms, and drug-testing companies) and by the public relations campaigns of a multibillion-dollar industry, whose entrepreneurial interest lies in magnifying the severity of workplace drug-related problems and extolling the benefits of drug testing as a solution (2, 3). Preemployment and random drug testing to find 1 otherwise-hidden drug abuser is estimated to cost between $700 000 and $1.5 million for the U.S. government's program (4). No solid data show that such testing deters drug use (4). Alcohol, narcotic, and sedative addiction are not as common among physicians as the general population (2). Their conclusion that physician drug testing has proven effective in health care (5) relies on 1 study, whose authors admitted, “We cannot conclude from our data whether there has been a decrease in the incidence of abuse” (6). The National Academy of Sciences concluded that frequently cited estimates of lost productivity and impaired performance from drug use are based on flawed data (7). Drug tests are subject to sabotage and to false-positive and falsenegative results (2). Testing damages workplace morale, reduces productivity, and hinders recruitment of skilled workers (2). No employer has been held legally liable for not having a drug-testing program; however, employers have incurred substantial legal costs in defending their programs against wrongful dismissal claims (2). Most existing drug-testing policies of hospitals are vague on procedural details and confidentiality. Substantial numbers of practicing physicians, residency program directors, and medical students oppose such Annals of Internal Medicine • Vol. 162 No. 8 • 21 April 2015 597

Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/933761/ on 07/07/2017

LETTERS testing (2). The Canadian Human Rights Commission recently disallowed preemployment and random drug testing of public employees on the grounds that such policies are human rights violations under the Canadian Human Rights Act. To improve patient safety and enhance quality of care, the medical profession should improve substance abuse education and training and encourage error reporting and analysis. Impairment testing (memory, vision, reflexes, and coordination) could uncover substance abuse, along with physical disabilities, mental illness, and sleep deprivation. Those found impaired or incompetent should be disciplined appropriately and referred for treatment (2). Martin Donohoe, MD School of Community Health, Portland State University Portland, Oregon Disclosures: The author has disclosed no conflicts of interest. Forms can be viewed at www.acponline.org/authors/icmje/ConflictOfInterest Forms.do?msNum=L15-0078.

References 1. Pham JC, Pronovost PJ. California's Proposition 46: a wolf in sheep's wool. Ann Intern Med. 2014;161:913-4. [PMID: 25265344] doi:10.7326/M14-2167 2. Donohoe M. Urine trouble: practical, legal, and ethical issues surrounding mandated drug testing of physicians. J Clin Ethics. 2005;16:85-96. [PMID: 15915849] 3. Lundberg GD. Mandatory unindicated urine drug screening: still chemical McCarthyism. JAMA. 1986;256:3003-5. [PMID: 3773220] 4. Maltby LL. Drug Testing: A Bad Investment. 1st ed. New York: American Civil Liberties Union; 1999. Accessed at www.aclu.org/drug-law-reform/drug -testing-bad-investment/ on 9 May 2013. 5. Pham JC, Pronovost PJ, Skipper GE. Identification of physician impairment. JAMA. 2013;309:2101-2. [PMID: 23629590] doi:10.1001/jama.2013.4635 6. Fitzsimons MG, Baker KH, Lowenstein E, Zapol WM. Random drug testing to reduce the incidence of addiction in anesthesia residents: preliminary results from one program. Anesth Analg. 2008;107:630-5. [PMID: 18633044] doi: 10.1213/ane.0b013e318176fefa 7. Normand J. Under the Influence? Drugs and the American Workforce. Washington, DC: National Academies Pr; 1994.

IN RESPONSE: We thank Dr. Donohoe for sharing his views on physician drug testing. He suggests that impaired physicians cannot be identified through drug testing for various reasons, including misconceptions surrounding substance abuse, business interests, legal costs, and vague hospital procedures. Many of these points do not relate to the effectiveness of drug testing but rather a larger social context around it. Although it is true that drug testing will not detect all cases of abuse (for example, alcoholics in withdrawal and testing too long after sentinel events), these 2 examples hardly represent all of the situations in which drug testing can and will detect illicit substance use that could lead to impairment. Those that are detected can be further evaluated. If addiction is regarded as an illness that causes disability and impairment, developing screening tools to help detect it before overt impairment seems reasonable, similar to screening for other disorders before serious complications. For example, blood sugar testing is routine for early detection of diabetes, however, it is by no means foolproof. A high blood sugar level simply warrants further investigation. If, on the other hand, drug testing is considered a legal issue, then it

makes sense that a positive result from drug testing is not proof of impairment any more than a high blood sugar level is proof of diabetic retinopathy. Drug testing is no less accurate than commonly used laboratory tests for other illnesses (1). In fact, the same technologies are used as with other laboratory testing; significant drug testing results are followed by an evaluation by a physician medical review officer. All tests have a measurable sensitivity and specificity, and clinicians find them useful. Although drug testing is associated with costs, the benefits are numerous. Physicians with substance use disorders (and their patients) stand to benefit tremendously from treatment (2). Moreover, health care is an industry where “accidents” can potentially result in loss of patient well-being or life. Detection of a single impaired physician or prevention of a single adverse event might justify a decade's worth of testing. Finally, the tremendous value in building public trust is not easily quantified. Balancing the rights of individual freedom against the health and well-being of the public is difficult. At what point does a job become safety-sensitive enough to require screening for potentially impairing conditions? Does individual freedom necessitate that we wait for a disaster rather than impede a worker's rights to inappropriately consume alcohol or drugs in the workplace? Finally, Dr. Donohoe recommends substance abuse education, encouraging reporting and analysis of errors, and/or impairment testing. We agree with these recommendations, and we believe these 2 approaches are complementary and potentially synergistic. Gregory E. Skipper, MD Promises Treatment Centers Santa Monica, California Julius Cuong Pham, MD, PhD Peter J. Pronovost, MD, PhD Armstrong Institute for Patient Safety & Quality, Johns Hopkins University School of Medicine Baltimore, Maryland Disclosures: Disclosures can be viewed at www.acponline.org /authors/icmje/ConflictOfInterestForms.do?msNum=L15-0079. References 1. Stout PR, Bynum ND, Mitchell JM, Baylor MR, Ropero-Miller JD. A comparison of the validity of gas chromatography–mass spectrometry and liquid chromatography–tandem mass spectrometry analysis of urine samples for morphine, codeine, 6-acetylmorphine, and benzoylecgonine. J Anal Toxicol. 2009; 33:398-408. [PMID: 19874645] 2. Skipper GE, Specht T. State PHPS protect the public—and save careers. Physician Health News. 2013;18(February):4-5.

OBSERVATION Pregnancy After Spontaneous Coronary Artery Dissection: A Case Series Background: Spontaneous coronary artery dissection (SCAD) is an increasingly recognized and potentially fatal cause of myocardial infarction (MI) that predominantly affects

598 Annals of Internal Medicine • Vol. 162 No. 8 • 21 April 2015

Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/933761/ on 07/07/2017

www.annals.org

LETTERS Figure. Patient diagram indicating time frame of SCAD, subsequent pregnancy, and follow-up. Initial SCAD 22 mo

1 G2P2 4 wk 2 *MS G4P3

To term

Pregnant aged 35 y

18 mo

Induced VD

FMD

Last follow-up

Postpartum SCAD

4 mo Pregnant aged 42 y

19 mo

3 G4P4

To term

Pregnant aged 40 y

18 mo

4 G4P4

6.5 y

Miscarriage

Pregnant aged 34 y

Last follow-up

2y

Elective C-section

To term

2 mo

SCAD recurrence

NSVD

5 TIA, SM G2P2

28 mo

Pregnant aged 36 y

6 G4P4

26 mo

Pregnant aged 36 y

34 wk

To term

Last follow-up

Elective C-section

Induced VD

14 mo

4.7 y

3 mo

Last follow-up

Last follow-up

Last follow-up

3 mo 7 G4P3

9 mo

Pregnant aged 35 y

Elective C-section

To term

Last follow-up

6 wk 8 FT, MCT G3P1

23 mo

Pregnant aged 39 y

Miscarriage

3.6 y

Last follow-up

C-section = cesarean section; FMD = fibromuscular dysplasia; FT = fertility treatment history; G = gravida; NSVD = normal spontaneous vaginal delivery; MCT = mixed connective tissue disease; MS = multiple sclerosis; P = para; SCAD = spontaneous coronary artery dissection; SM = systemic mastocytosis; TIA = transient ischemic attack; VD = vaginal delivery.

young women (1, 2). Pregnancy is considered a contributory factor in approximately 12% to 30% of events (1, 3, 4) and recurrence is now recognized, with 10-year Kaplan–Meier estimated rates as high as 29% (1). Current recommendations to avoid pregnancy after SCAD are based on the presumed increased risk for recurrence with pregnancy, although, to our knowledge, assessment of pregnancy after SCAD has not been described. Objective: To determine the risk for subsequent events in patients who became pregnant after SCAD. Methods: After approval by the Mayo Foundation Institutional Review Board, we reviewed records of all women in the Mayo Clinic SCAD Registry who had become pregnant after SCAD (as described elsewhere [5]) and had been enrolled in the registry between 1 July 2011 to 1 April 2014. Follow-up was determined by last clinical visit or study correspondence. Findings: Of 363 women, 8 (2.2%) reported pregnancy after SCAD (Figure). Mean age at the time of SCAD MI was 36 years (SD, 3), 4 of which were peripartum. Elapsed time to pregnancy was 18 months (SD, 9). Five of the 8 women (63%) knew their MI was due to SCAD before pregnancy, and 3 (38%) were initially misdiagnosed with coronary vasospasm. Six pregnancies resulted in live births: 3 vaginal deliveries and 3 elective cesarean sections. One patient with placenta www.annals.org

previa and antepartum hemorrhage required preterm cesarean section. No other significant delivery complications were reported. Two pregnancies resulted in miscarriage at 9 and 15 weeks. During follow-up (median, 36 months [interquartile range, 24 to 70 months]), 7 of the 8 women did not have complications, but 1 had recurrent SCAD 9 weeks after delivery with ST-segment elevation MI involving the left main coronary artery. Percutaneous coronary intervention was unsuccessful, so she had emergent coronary artery bypass grafting (CABG). Her initial SCAD was unrelated to pregnancy. Despite normalization of left ventricular ejection fraction within 1 year, she developed posttraumatic stress disorder. No other patients had recurrence. Discussion: The current study is the first, to our knowledge, to describe the potential outcomes of future pregnancy after SCAD. Although the cohort and observed recurrence were relatively small (1 in 8), the consequences were substantial. The woman with recurrence had a large MI that required emergent CABG and subsequent posttraumatic stress disorder. Whether pregnancy independently increases risk for SCAD recurrence is uncertain, but this possibility is certainly concerning. Because there are no established strategies to Annals of Internal Medicine • Vol. 162 No. 8 • 21 April 2015 599

Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/933761/ on 07/07/2017

LETTERS prevent recurrent SCAD, avoidance of subsequent pregnancy is advised. For those who proceed with pregnancy despite risk, we recommend dedicated care with cardiology and maternal fetal medicine specialists. An unexpected finding was that more than one third of the women were initially misdiagnosed with vasospasm as the cause of MI. We recommend that all women of childbearing age who have had a MI be assessed to confirm the cause with consideration of SCAD. This study is limited by the small number of women and selection bias that occurs with a registry. Regardless, no other reports of pregnancy after SCAD exist, and this study provides important information regarding pregnancy risk for a woman who has had SCAD or MI of uncertain cause. Marysia S. Tweet, MD Sharonne N. Hayes, MD Rajiv Gulati, MD, PhD Carl H. Rose, MD Patricia J.M. Best, MD Mayo Clinic College of Medicine Rochester, Minnesota

characteristics and clinical implications. Circ Cardiovasc Interv. 2014;7:656-62. [PMID: 25138034] doi:10.1161/CIRCINTERVENTIONS.114.001676

CORRECTIONS Correction: Pharmacologic Interventions for Painful Diabetic Neuropathy A recent article (1) was missing the estimates for pregabalin versus placebo in the Data Synthesis section of the abstract and in the section, Meta-analysis by Individual Drug. The values are as follows: pregabalin (standardized mean difference, ⫺0.55 [CrI, ⫺0.94 to ⫺0.15]). This has been corrected in the online version. Reference 1. Griebeler ML, Morey-Vargas OL, Brito JP, Tsapas A, Wang Z, Carranza Leon BG, et al. Pharmacologic interventions for painful diabetic neuropathy: An umbrella systematic review and comparative effectiveness network meta-analysis. Ann Intern Med. 2014;161:639-49. [PMID: 25364885] doi:10.7326/M14-0511

Acknowledgment: The authors thank the patients participating in the Mayo Clinic SCAD Registry and acknowledge Jill Boyum; Michelle Schoeppner; Diane Vrieze; and Sue Ward, RN, for their registry work.

Correction: Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): The TRIPOD Statement

Financial Support: This study was funded in part by the Mayo Clinic

A recent article (1) was erroneously published with the American College of Physicians copyright symbol. The ACP does not hold copyright on this manuscript. This has been corrected in the online version.

Division of Cardiovascular Diseases, Chicago Mercantile Exchange Foundation, and SCAD Research, Inc.

Disclosures: Disclosures can be viewed at www.acponline.org /authors/icmje/ConflictOfInterestForms.do?msNum=L14-0446.

References 1. Tweet MS, Hayes SN, Pitta SR, Simari RD, Lerman A, Lennon RJ, et al. Clinical features, management, and prognosis of spontaneous coronary artery dissection. Circulation. 2012;126:579-88. [PMID: 22800851] doi:10.1161 /CIRCULATIONAHA.112.105718 2. Tweet MS, Eleid MF, Best PJ, Lennon RJ, Lerman A, Rihal CS, et al. Spontaneous coronary artery dissection: revascularization versus conservative therapy. Circ Cardiovasc Interv. 2014;7:777-86. [PMID: 25406203] doi:10.1161 /CIRCINTERVENTIONS.114.001659 3. Thompson EA, Ferraris S, Gress T, Ferraris V. Gender differences and predictors of mortality in spontaneous coronary artery dissection: a review of reported cases. J Invasive Cardiol. 2005;17:59-61. [PMID: 15640544] 4. Mortensen KH, Thuesen L, Kristensen IB, Christiansen EH. Spontaneous coronary artery dissection: a Western Denmark Heart Registry study. Catheter Cardiovasc Interv. 2009;74:710-7. [PMID: 19496145] doi:10.1002/ccd.22115 5. Eleid MF, Guddeti RR, Tweet MS, Lerman A, Singh M, Best PJ, et al. Coronary artery tortuosity in spontaneous coronary artery dissection: angiographic

Reference 1. Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. 2015;162:55-63. [PMID: 25560714] doi:10.7326/M14-0697

Correction: The State of Research Funding From the National Institutes of Health for Criminal Justice Health Research For the following sentence in a recent article (1): “More than 20 million Americans, or 9% of the U.S. adult population, are currently or have been incarcerated (2)”, references 3 and 4 should be cited in addition to reference 2. This has been corrected in the online version. Reference 1. Ahalt C, Bolano M, Wang EA, Williams B. The state of research funding from the national institutes of health for criminal justice health research. Ann Intern Med. 2015;162:345-52. [PMID: 25732276] doi:10.7326/M14-2161

600 Annals of Internal Medicine • Vol. 162 No. 8 • 21 April 2015

Downloaded From: https://annals.org/pdfaccess.ashx?url=/data/journals/aim/933761/ on 07/07/2017

www.annals.org

Correction: pharmacologic interventions for painful diabetic neuropathy.

Correction: pharmacologic interventions for painful diabetic neuropathy. - PDF Download Free
98KB Sizes 2 Downloads 19 Views