HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

Correspondence to:

ARTICLES: B. A. Fallon, J. G. Keilp, K. M. Corbera, E. Petkova, C. B. Britton, E. Dwyer, I. Slavov, J. Cheng, J. Dobkin, D. R. Nelson, and H. A. Sackeim A randomized, placebo-controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy Neurology 2008; 70: 992-1003 [Abstract] [Full text] [PDF]

Correspondence: Submit a response to this article

Correspondence published:

A randomized, placebo-controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy

Adriana Marques, MD, Pamela Shaw, PhD (Bethesda, MD), Christopher H. Schmid, PhD (Boston, MA) Richard F. Kaplan, PhD (Farmington, CT), Afton Hassett, PsyD (New Brunswick, NJ) Eugene Shapiro, MD (New Haven, CT) Allen Steere, MD (Boston, MA) and Gary P. Wormser MD (Valhalla, NY   (4 June 2008)

Reply from the authors

Brian A Fallon, MD, Eva Petkova, PhD John G Keilp, PhD, Iordan Slavov, PhD   (4 June 2008)

A randomized, placebo-controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy

Elizabeth L. Maloney, MD   (6 May 2008)

Reply from the Editorialist

John J. Halperin   (6 May 2008)

Reply from the author

Brian A. Fallon   (6 May 2008)

A randomized, placebo-controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy 4 June 2008
Adriana Marques, MD, NIAID/NIH 10/11N234 10 Center Dr., Bethesda MD 20892, Pamela Shaw, PhD (Bethesda, MD), Christopher H. Schmid, PhD (Boston, MA) Richard F. Kaplan, PhD (Farmington, CT), Afton Hassett, PsyD (New Brunswick, NJ) Eugene Shapiro, MD (New Haven, CT) Allen Steere, MD (Boston, MA) and Gary P. Wormser MD (Valhalla, NY Send Correspondence to journal: Re: A randomized, placebo-controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy amarques{at}niaid.nih.gov Adriana Marques, MD, et al.	For transparent results, randomized, placebo-controlled clinical trials optimally evaluate a primary outcome of treatment efficacy with a simple, well-defined statistical test. Fallon et al. chose a linear mixed model with factors of group, time and cognitive domain as their primary analysis. [1] Model fitting involved selecting correlation structure and assessing significant two- and three-way interactions. Model selection procedures are, by definition, data-driven and are generally considered appropriate for exploratory analyses rather than for evaluating treatment efficacy in confirmatory randomized trials. [2] Interpreting p-values as if the selected model had been hypothesized a priori is frequently overly optimistic and underestimates p-values. [3,4] Using the same data for model selection and parameter estimation can also substantially bias parameter estimates, particularly if the number of model parameters is large relative to the number of subjects. [3] The primary conclusion involves the interaction between treatment and time, interpreted as a beneficial effect of IV ceftriaxone at 12 weeks that disappears by 24 weeks. Given the potentially optimistic error rates, too much emphasis is given to the borderline p value of 0.053 for the comparison in cognitive improvement between antibiotic and placebo at 12 weeks. Attaching a strong conclusion to them particularly when the results are not clear-cut is unwarranted as they were post-hoc comparisons. Inclusion of a group of healthy controls who do not directly inform the drug vs. placebo comparison also raises the question of how the overall tests of significance might reflect differences between healthy subjects and patients, rather than drug vs. placebo. The lower index score in the antibiotic group at baseline raises the question whether part of the difference in gains between the two groups was introduced by a “regression towards the mean” effect. That is, whether by chance the antibiotic group had more individuals with worse test scores that drifted closer to the overall mean at the next time point. The authors do not mention the significant interaction between group and cognitive domain in the primary omnibus analysis. The randomized, placebo-controlled trial design allows for straightforward statistical analysis of the primary outcome. Fallon at al. made it difficult to interpret the results by choosing as their primary analysis a complicated model chosen by a data-driven model selection process. [1] Nevertheless, the Fallon study joins other randomized, double-blind trials that failed to provide convincing evidence that additional antibiotic therapy is a beneficial strategy for patients with post-Lyme disease symptoms. [5] Other management approaches need to be studied. References 1. Fallon BA, Keilp JG, Corbera KM, et al. A randomized, placebo- controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy. Neurology 2008 Mar 25;70:992-1003. 2. Guidance for Industry. E9 Statistical principles for clinical trials. U.S. Department of Health and Human Services Food and Drug Administration Federal Register on September 16, 1998 (63 FR 49583), Accessed online February 7, 2008. http://www.fda.gov/cder/guidance/index.htm 3. Burnham KP. Anderson DR. Model Selection and multimodel inference. A practical information-theoretic approach (1998). Springer, New York. 4. Ye J. On measuring and correcting the effects of data mining and model selection. Journal of the American Statistical Association 1998:120-131. 5. Wormser GP, Dattwyler RJ, Shapiro ED, et al. The clinical assessment, treatment, and prevention of Lyme disease, human granulocytic anaplasmosis, and babesiosis: clinical practice guidelines by the Infectious Diseases Society of America. Clin Infect Dis. 2006 Nov 1;43:1089-1134. Disclosure: The authors report no conflicts of interest.
Reply from the authors 4 June 2008
Brian A Fallon, MD, Columbia University 1051 Riverside Drive, Unit 69, NYC, NY 10032, Eva Petkova, PhD John G Keilp, PhD, Iordan Slavov, PhD Send Correspondence to journal: Re: Reply from the authors baf1{at}columbia.edu Brian A Fallon, MD, et al.	Marques et al. call for the use of a “simple well-defined statistical test” without specifying their test of preference. The analysis of clinical trials requires statistical approaches that adequately handle missing data. Mixed effects models (MEM) produce valid inferences under the “missing at random” (MAR) assumption in contrast to the more restrictive “missing completely at random” assumption required for simple tests. [6,7,8, 9] The validity of the likelihood approach (MEM) under MAR is based on correctly specifying the covariance structure. The small sample size prevented the specification of the completely unrestricted covariance [10], which would have circumvented modeling it. Testing whether the effect is the same for all cognitive domains (three-way interaction) before assessing an overall treatment effect (two-way interaction) is an established procedure that is not post-hoc. Patient groups did not differ significantly on baseline cognition. In addressing regression to the mean (RTM), the authors do not specify the reference population. Is it (1) the healthy control or (2) the study patient population? In either case, RTM is not selective and should impact both drug and placebo groups. In case (1), RTM should affect patients in each group similarly, since each had cognitive index scores well below the control mean at baseline. The placebo group exhibited gradual cognitive improvement consistent with a practice effect at a rate parallel to the controls (figure 1) and no evidence for RTM. The antibiotic group’s trajectory for improvement steeply increased to week 12 and then, contrary to RTM, decreased to week 24. In case (2), since the placebo group was above the overall patient mean, RTM should have brought the placebo mean down, which it did not. The logical explanation for the pattern of improvement followed by worsening in the antibiotic group is a treatment effect in the first 12 weeks and loss of the effect after treatment discontinuation. Healthy controls were included in the study design to enable parcellation of the practice, placebo, and drug effects. The significant group by cognitive domain interaction was not surprising, as patients with Lyme encephalopathy have a differing pattern of cognitive strengths and weaknesses than healthy controls. Because of the small sample size, we emphasized confidence intervals rather than p- values. Finally, although acute benefits were seen, the risks of IV treatment and lack of sustained benefits in cognition underline the need for other treatments for persistent Lyme encephalopathy that are safe, effective, and durable. References 6. Little RJA. A class of pattern mixture models for normal incomplete data. Biometrika 1994;81:471-483 7. Siddiqui O and Ali MW. A comparison of the random-effects pattern-mixture model with last-observation-carried-forward (LOCF) analysis in longitudinal clinical trials with dropouts. J. Biopharmaceutical Statistics 1998;8:545-563. 8. Everitt BS. Analysis of longitudinal data: Beyond MANOVA. British Journal of Psychiatry 1998;172:7-10. 9. Molenberghs G and Kenward MG. Missing Data in Clinical Studies. Wiley & Sons Inc.: 2007 Chichester, England. 10. Mallinckrodt, CH, Clark WS, Stacy RD. Accounting for dropout bias using mixed effects models. J. Biopharmaceutical Statistics 2001;11: 9-21. Disclosure: Disclosure listed on Correspondence to which this article refers - Roche Pharmaceuticals supplied ceftriaxone free of charge for this study but were not involved in any other aspect of the study. Dr. Fallon has given expert testimony at hearings related to Lyme disease and its treatment. The other authors report no disclosures.
A randomized, placebo-controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy 6 May 2008
Elizabeth L. Maloney, MD, none 25611 W. Comfort Dr Wyoming, MN 55092 Send Correspondence to journal: Re: A randomized, placebo-controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy bettymal2003{at}yahoo.com Elizabeth L. Maloney, MD	Fallon el al. report results from a carefully constructed ceftriaxone trial for Lyme encephalopathy demonstrating that patients responded well to additional treatment but failed to sustain the response when treatment was withdrawn. [1] The study makes several important points that contradict the conventional wisdom about Lyme disease. Although mainstream experts assert that objective findings are rare post-treatment [2], 73 % of Fallon’s patients had sensory abnormalities an average of seven years after the initial diagnosis and treatment of Lyme disease. The accompanying editorial suggests that long-term antibiotic therapy is ineffective. [3] The study suggests that 10 weeks of treatment with ceftriaxone may be insufficient to eradicate a pleomorphic bacterium like Borrelia burgdorferi, the Lyme spirochete. Cell-wall drugs promote transformation to the “cystic” form, which is unaffected by cell-wall agents, such as cephalosporins and penicillins, tetracyclines and macrolides. [4] Once the antibiotic threat passes, these cystic forms may revert to the motile state. The spirochete can invade a variety of human cell types [5] and intracellular pathogens require longer treatment courses utilizing suitable antibiotics. It is possible that ceftriaxone alone, regardless of treatment duration, may fail to eradicate a well established Lyme infection. That ceftriaxone therapy works in many patients underscores the need to better understand the variables of the host response in this illness. The patients entering this study were symptomatic despite a mean of 2.5 months of IV and 7.9 months of oral antibiotics. Additional ceftriaxone was able to induce clinical improvements as seen at week 12. However, as a single therapeutic agent, it was unable to sustain these gains when treatment was withdrawn. In this select group of patients, more of the same proved to be insufficient. Proponents of long-term treatment believe that antibiotic therapy must address the multiple mechanisms supporting survival of the Lyme spirochete. Enough is only enough when it is the “right stuff.” References 1. Fallon B, Keilp G, Corbera K, et al. A randomized, placebo-controlled trial of repeated IV antibiotic therapy for Lyme encephalopathy. Neurology 2008; 70: 992-1003. 2. Halperin J, Bever C, Belman A, et al. Practice Parameter: Treatment of nervous system Lyme disease (an evidence-based review). Neurology 2007;69:91-102. 3. Halperin J. Prolonged Lyme disease treatment: Enough is enough. Neurology 2008;70:986-987. 4. Kersten A, Poitschek C, Rauch S, Aberer E. Effects of Penicillin, Ceftriaxone, and Doxycycline on Morphology of Borrelia burgdorferi. Antimicrob Agents Chemother 1995;39:1127-1133. 5. Livengood JA, Gilmore RD Jr. Invasion of human neuronal and glial cells by an infectious strain of Borrelia burgdorferi. Microbes Infect 2006;8:2832-2840. Disclosure: The author reports no conflicts of interest.
Reply from the Editorialist 6 May 2008
John J. Halperin, Atlantic Neuroscience Institute Overlook Hospital, Summit NJ 07902 Send Correspondence to journal: Re: Reply from the Editorialist halperin{at}lineuro.com John J. Halperin	I am grateful to Dr. Maloney for the opportunity to address several points. First, she asserts that objective neurologic abnormalities are common after treatment, pointing out that 27 of the study’s patients had sensory abnormalities. Although this is 73% of the study subjects, the 37 patients in the study were selected from among 3,368 patients screened. During the study’s recruitment period, the CDC web site indicated there were approximately 85,000 new cases of Lyme disease. Observations in such an extraordinarily selected population are meaningless epidemiologically but certainly cannot be used to conclude that objective abnormalities are common. Second, the statement that “The accompanying editorial suggests that long-term antibiotic therapy is ineffective” misattributes this conclusion. It is the data in the study that demonstrate only a marginally statistically significant and unsustained difference between patients receiving ceftriaxone and those given placebo. Concluding that this supports the use of even longer courses of antibiotics is contrary to what we understand about the biology of organisms such as B burgdorferi, and ignores the significant morbidity associated with this approach. Allusions to ‘pleomorphic’ ‘cell wall free’ ‘cystic forms’ of this bacterium are frequently repeated. There is no evidence that this occurs in vivo or is of any clinical significance. The conclusion that B burgdorferi would be “unaffected by cell-wall agents, such as cephalosporins and penicillins, tetracyclines and macrolides” is unfortunately misguided since tetracyclines and macrolides act on intracellular protein synthesis, not on the cell wall. It is difficult for a patient-centered, evidence-based clinician to justify prolonged antibiotic therapy in patients with Lyme disease or post-Lyme disease symptoms given concordant observations [1,8,9] that this treatment poses substantial risks and provides minimal or no benefit. Disclosure: Dr. Halperin has served as a defense expert in cases alleging failure to diagnose or treat nervous system Lyme disease. Dr. Halperin holds equity in Abbott, Bristol Myers Squibb, Johnson & Johnson, Schering Plough and Vasogen.
Reply from the author 6 May 2008
Brian A. Fallon, Columbia University baf1@columbia.edu Send Correspondence to journal: Re: Reply from the author baf1{at}columbia.edu Brian A. Fallon	Persistent symptoms after standard courses of antibiotic treatment for Lyme disease may occur for several reasons. These include persistence of infection, a post-infectious process, or incorrect diagnosis. If infection persists, then the cause may be due to either the immune evading survival strategies of the Borrelia burgdorferi (Bb) spirochete or the failure of antibiotics to eradicate the organism. [6] Dr. Maloney argues that ceftriaxone may not be sufficient to eliminate the Bb spirochete which at least in vitro has been shown to exist intracellularly and may not be effective against cyst-like forms. That viable Bb may persist despite treatment was recently demonstrated in a mouse study in which Bb were cultured after ceftriaxone therapy when the mice were subsequently given anti-tumor necrosis factor-alpha. [7] Persistence of symptoms, however, may also be explained by a post-infectious process, which includes autoimmune phenomena, neurotransmitter changes triggered by the initial infection, or damage. It is clear that the group of patients with post-treatment Lyme disease symptoms is heterogeneous. While some patients improve in response to antibiotic re-treatment, showing either short-term moderate improvement in cognition or longer term improvement in fatigue, pain, and physical dysfunction, [1,8] other patients do not. [8,9] In order to avoid exposing patients unnecessarily to the serious risks associated with antibiotic therapy, particularly when given intravenously, biomarkers of active infection need to be identified. These biomarkers will allow physicians to distinguish patients most likely to benefit from antibiotic re-treatment from those where other symptom-reduction strategies should be used. The care of patients with chronic persistent symptoms would be advanced by increased research into the heterogeneous etiology of symptom persistence and studies that seek to identify safe, effective, and specific treatments targeted to the mechanism of persistent symptoms in discrete patient groups. References 6. Embers ME, Ramamoorthy R, Philipp MT. Survival strategies of Borrelia burgdorferi, the etiologic agent of Lyme disease. Microbes and Infection 2004;6:312-318. 7. Yrjanainen H, Hytonen J, Song XY, Oksi J, Hartiala K, Viljanen MK. Anti-tumor necrosis factor-alpha treatment actives Borrelia burgdorferi spirochetes 4 weeks after ceftriaxone treatment in C3H/He mice. J Infect Dis 2007;195:1489-1496. 8. Krupp LB, Hyman LG, Grimson R, et al. Study and treatment of post Lyme disease (STOP-LD): A randomized double masked clinical trial. Neurology 2003;60:1923-1930. 9. Klempner M, Hu LT, Evans J et al. Two controlled trials of antibiotic treatment in patients with persistent symptoms and a history of Lyme disease. N Engl J Med 2001;345:85-92. Disclosure: The following disclosures relate to the study to which this Correspondence refers. This study was funded by a grant from NINDS to Dr. Fallon (R01- NS38636). Roche Pharmaceuticals supplied ceftriaxone free of charge for this study but were not involved in any other aspect of the study. Dr. Fallon has given expert testimony at hearings related to Lyme disease and its treatment. The other authors report no conflicts of interest.