Contemporary Clinical Trials Volume 33, Issue 6, November 2012, Pages
1132–1142 Antibiotic retreatment of Lyme disease in patients with
persistent symptoms: A biostatistical review of randomized,
placebo-controlled, clinical trials Allison K. DeLonga, , , Barbara
Blossomb, Elizabeth L. Maloneyc, Steven E. Phillipsd a Center for
Statistical Sciences, Department of Biostatistics, Brown
University, Providence, RI, USA b Department of Statistics, Colorado
State University, Fort Collins, CO, USA c Partnership for Healing and
Health, Ltd., Wyoming, MN, USA d Greenwich Hospital,
Greenwich, CT, USA Received 4 May 2012 Revised 16 July 2012 Accepted
14 August 2012 Available online 19 August 2012
http://dx.doi.org/10.1016/j.cct.2012.08.009, How to Cite or Link Using
DOI
Permissions & Reprints Abstract Introduction Lyme disease (Lyme
borreliosis) is caused by the tick-borne spirochete Borrelia
burgdorferi. Long-term persistent illness following antibiotic
treatment is not uncommon, particularly when treatment is delayed.
Current treatment guidelines for persistent disease primarily rely on
findings from four randomized, controlled trials (RCTs),
strongly advising against retreatment. Methods We performed a
biostatistical review of all published RCTs evaluating antibiotic
retreatment, focusing on trial design, analysis and conclusions.
Results Four RCTs met the inclusion criteria; all examined the
efficacy of intravenous ceftriaxone versus placebo at approximately 3 or
6 months. Design assumptions for the primary outcomes in the
two Klempner trials and two outcomes in the Krupp trial were
unrealistic and the trials were likely underpowered to detect clinically
meaningful treatment effects. The Klempner trials were analyzed
using inefficient statistical methods. The Krupp RCT was well-designed
and analyzed for fatigue, finding statistically significant and
clinically meaningful improvement. Fallon corroborated this
finding. Fallon also found improvement in cognitive functioning, a
primary outcome, at 12 weeks which was not sustained at 24 weeks;
improvements in physical functioning and pain were demonstrated
at week 24 as an interaction effect between treatment and baseline
symptom severity with the drug effect increasing with higher baseline
impairment.
Discussion This biostatistical review reveals
that retreatment can be beneficial. Primary outcomes originally
reported as statistically insignificant were likely underpowered. The
positive treatment effects of ceftriaxone are encouraging and
consistent with continued infection, a hypothesis deserving additional
study. Additional studies of persistent infection and antibiotic
treatment are warranted. Keywords Lyme disease;
Neuroborreliosis; Randomized controlled trial; Statistical power;
Minimum clinically important difference (MCID); Treatment guidelines
Figures and tables from this article: 
Fig. 1. Flow diagram of
the literature search for randomized, controlled trials evaluating
antibiotic retreatment in Lyme disease patients with persistent symptoms
following a standard course of treatment. Figure options
Fig. 1. Flow diagram of
the literature search for randomized, controlled trials evaluating
antibiotic retreatment in Lyme disease patients with persistent symptoms
following a standard course of treatment. Figure options
Table 1. Available measures of treatment effects for each trial and
outcome. NS effect not given, reported as not statistically
significant.“–”Within-arm effects were not reported for each trial.
Table options View in workspace
Table 2. SF-36 summary score changes
found to be clinically and statistically significant for chronic
diseases of similar severity to Lyme disease. ‡ Not determined;
§ Not significant. Table options View in workspace
Table 3. Estimated
differences in the proportion of patients expected to be classified as
improved using Klempner et al.'s categorization for
various mean treatment effects consistent with published MCIDs.
Klempner et al.'s results are provided and confirm clinically meaningful
mean differences of 2 to 5 points fall within Klempner et
al.'s 95% confidence intervals [16]. Table options View in workspace
Table 4. Mean response times of Lyme patients and controls on the
Alphabet Arithmetic test (Pollina et al., Table 3) [39] and
the differences in the two groups presented as the percentage faster
that healthy participants completed the task compared to the Lyme
patients. 1 In the STOP-LD study design, Krupp et al. [24]
assumed a 25% improvement as the MCID.2 Age- and education-matched
controls.
