Treatment Response in Enteric Fever in an Era of Increasing Antimicrobial Resistance: An Individual Patient Data Analysis of 2092 Participants Enrolled into 4 Randomized, Controlled Trials in Nepal

Summary This is the largest collection of enteric fever treatment data. The results, from trials conducted in Nepal since 2005, confirm that fluoroquinolones are failing for enteric fever treatment. The World Health Organization enteric fever treatment guidelines should be modified.

Enteric (typhoid) fever is a systemic infection caused by the Salmonella enterica serovars Typhi and Paratyphi A, B, and C. Enteric fever is a significant cause of morbidity and mortality in low-income regions [1] and was responsible for an estimated 12.2 million disability-adjusted life-years and >190 000 deaths globally in 2010 [2]. The fatality rate of enteric fever is low (<1%) but is higher when antimicrobial therapy is delayed or unavailable [3]. Therefore, antimicrobials are essential for the clinical management of enteric fever. Chloramphenicol, ampicillin, and cotrimoxazole were first-line treatments for enteric fever until the early 1990s when the increasing incidence of multidrug-resistant (MDR; defined as resistance to these 3 antimicrobial drugs) S. Typhi organisms led to the use of fluoroquinolones [4,5]. Yet, organisms with reduced susceptibility against fluoroquinolones became a problem in Asia soon after their introduction [6,7]. Recent phylogeographic analyses that document an ongoing epidemic of a global antimicrobial resistance (AMR) S. Typhi lineage suggest that the potential for regional or global dispersal of a lineage exhibits resistance to fluoroquinolones is now a real threat [8][9][10]. In the absence of effective and accessible vaccines and lack of sanitation improvements, development of tailored antimicrobial therapy recommendations is critical to reduce morbidity and prevent disease transmission.
In Kathmandu, Nepal, S. Typhi and S. Paratyphi A are the most commonly isolated organisms from the blood of febrile adults and children [11,12]. Over the last decade we conducted 4 randomized, controlled trials (RCTs) to evaluate enteric fever treatment in this endemic region [13][14][15][16]. Our aim in this study was to use the largest collection of individual patient data assembled to date from enteric fever treatment trials to evaluate the effect of treatment drug on differences in clinical outcome between S. Typhi and S. Paratyphi A infections and those with blood culture-negative enteric fever. We further sought to compare the antimicrobial susceptibility profiles over time between S. Typhi and S. Paratyphi A isolates and to investigate their impact on outcome. An in-depth understanding of trends and clinical implications of AMR enteric fever should guide policymakers and clinicians in decisions regarding treatment in an era of rapidly diminishing therapeutic options.

Ethical Approval
Written informed consent, which was required for participation in all trials, was provided by a parent or adult guardian if a patient was aged <18 years. The Nepal Health Research Council Ethics Committee and the Oxford Tropical Research Ethics Committee of the United Kingdom provided ethical approval for all 4 studies.

Patient Populations and Study Procedures
Individual patient data for this study were derived from 4 RCTs conducted at Patan Hospital in Kathmandu, Nepal, between 2005 and 2014, the methods and results of which have been described previously [13][14][15][16]. Patients who presented to the outpatient or emergency department with fever lasting longer than 3 days with a clinical diagnosis of enteric fever (undifferentiated fever >38°C with no focus of infection) were eligible. Patients were excluded if they were pregnant or lactating, were aged <2 years or weighed <10 kg, showed any signs of complications (jaundice, shock, gastrointestinal bleeding), showed hypersensitivity to the relevant trial drugs, or had been treated with a study drug in the week prior to going to the hospital. The study procedures between the 4 trials were comparable; however, there were several minor protocol differences between studies (outlined in Supplementary Table 1).
Patients were randomly assigned to 1 of 2 arms in each trial. Each trial was composed of a gatifloxacin arm (10 mg/kg/day, single dose orally for 7 days) and a comparator arm, which was cefixime (20 mg/kg/day, 2 doses orally for 7 days) [13], chloramphenicol (75 mg/kg/day, 4 divided oral doses for 14 days) [14], ofloxacin (20 mg/kg/day, 2 divided oral doses for 7 days) [15], or ceftriaxone (intravenous; 60 mg/kg for patients aged 2-13 years or 2 g for patients aged ≥14 years) [16]. Gatifloxacin was the constant comparator because it is inexpensive and given once daily.
Fever clearance time (FCT) was defined as the time from the first dose of a study drug until the temperature dropped to ≤37.5°C and remained below this temperature for at least 2 days. The composite endpoint treatment failure summarized unfavorable outcomes and was defined as the occurrence of at least 1 of the following: persistent fever (FCT of more than 7 days [trials 1 and 4] or more than 10 days [trials 2 and 3] after treatment initiation), the need for rescue treatment, microbiological failure (blood culture positive for Salmonella) on day 8, relapse or disease-related complications within 31 days of treatment initiation, or death. Blood was taken from all patients for microbiological culture on enrollment and on day 8 for culture-positive individuals or those with a potential relapse.
Microbiological investigations have been described previously [13][14][15][16]. Blood samples from adult patients were inoculated into media containing tryptone soya broth and sodium polyanethol sulfonate. For children, BacTEC Ped Plus/F bottles were used. Positive bottles were cultured onto MacConkey agar and presumptive Salmonella colonies were identified using biochemical tests and serotype-specific antisera. During all 4 trials, minimum inhibitory concentrations (MICs) were determined against the following antimicrobials unless otherwise noted: Augmentin, ampicillin, amoxicillin, azithromycin

Statistical Analyses
Data from the trials were combined and analyzed using Stata (v 13.1; College Station, Texas). Plots were drawn in R v3.1.1 (R Foundation, Vienna, Austria) using the ggplot2 package. Demographics and clinical variables were tabulated and compared between serovars. Comparisons of clinical parameters between patient populations were structured as logistic regressions with the patient population (either culture positive/negative or S. Typhi/S. Paratyphi A) as the main covariate and adjustment for age stratum (binary: <16 years/≥16 years). Multivariable models with random effects were fitted to adjust for study heterogeneity as follows: FCT was evaluated using Kaplan-Meier estimates and Cox proportional hazard models with treatment group and age as covariates; logistic regression was used to determine the odds of treatment failure between treatment arms, controlling for age; and linear regression was used to evaluate the relationship between FCT and log 2 MIC, also controlling for age. Generalized additive models (GAMs; identity link, cubic spline) were used to examine potential nonlinear trends of MIC over time.
There were several significant differences in clinical history between patient populations after controlling for age (Table 2).
Culture-negative patients were significantly more likely to report coughing (40%) and vomiting (22%) than culture-positive patients (31% and 17%, respectively). Culture-positive patients, however, reported diarrhea (24%) more often than culture-negative patients (17%) in addition to a higher temperature (median, 39.0°C and 38.7°C, respectively). Among the culture-positive patients, those with an S. Typhi infection were significantly more likely to report a history of anorexia (78%), coughing (33%), and diarrhea (28%) compared to the S. Paratyphi A patients (71%, 25%, and 15%, respectively) and presented with higher temperatures (median, 39.0°C vs 38.8°C). Salmonella Paratyphi A patients were significantly more likely to report a history of previous typhoid illness (23%) compared to S. Typhi patients (12%). Additionally, there were several significant differences in hematology parameters between the culture-negative/culture-positive patients and the S. Typhi/S. Paratyphi A patients (Table 1), despite the majority of the values falling within normal ranges. Asparate aminotransaminase and alanine aminotransaminase were significantly elevated in the culture-positive patients (median, 51 U/L and median, 38 U/L, respectively) compared to culture-negative patients (median, 42 U/L and median, 31 U/L, respectively).

Antimicrobial Susceptibility Trends
As shown in Figure 3, the MICs for S. Paratyphi A were significantly higher than those for S. Typhi with all antimicrobials (P < .001, Kruskal-Wallis), with the exception of cefixime (P = .375). Figure 4 shows the MIC time trends by serovar, which were significantly nonlinear over time for all antimicrobials in both serovars (GAM, P < .001 with the exception of S. Paratyphi A/ciprofloxacin [P = .052] and S. Paratyphi A/nalidixic acid [P = .003]). Most notably, the MICs against the fluoroquinolones rose significantly over time, and the MICs against azithromycin declined between 2007 and 2010. Last, all isolates were susceptible to ceftriaxone throughout the study period.

Impact of Antimicrobial Resistance on Clinical Outcomes
Increasing MICs against fluoroquinolones led to longer FCTs in S. Typhi patients. As shown in Figure 5, an increasing (log 2 ) MIC was associated with longer FCTs in patients treated with

DISCUSSION
Enteric fever remains the leading cause of febrile bacterial illness in Kathmandu [12]. With alarming AMR rates, a lack of immunization as a public health tool, and slow sanitation improvements, tailored antimicrobial therapies for the prevailing AMR profiles are required. Using systematic, longitudinal, individual patient data, we identified dynamic antimicrobial susceptibility profiles among S. Typhi and S. Paratyphi A isolates and a trend of increasing fluoroquinolone MICs correlating with poor outcome. This phenomenon was particularly apparent among S. Typhi patients. Although ceftriaxone was effective in treating culture-confirmed enteric fever patients, we documented poor clinical response in culture-negative patients. These data suggest that careful consideration is required for antimicrobial therapy of patients with enteric fever. In addition, fluoroquinolones should not be recommended for empirical treatment of this infection in South Asia [17]. By combining the largest number of enteric fever patients from a single location, we were able to identify several notable differences in both clinical presentation and clinical response between S. Typhi and S. Paratyphi A patients. Previous work conducted at the same center showed the 2 serovars to be clinically indistinguishable [18]. We found that, after controlling for age, S. Typhi patients were more likely to report anorexia, diarrhea, and coughing and presented with a higher temperature.  The precise mechanism driving the variability in MICs over time for both S. Typhi and S. Paratyphi A against several antimicrobials throughout 2005-2014 is unknown but may be determined by local prescribing practices. This hypothesis is consistent with notable declines in MDR organisms in both Nepal and India after fluoroquinolones became the first choice of treatment [12,19,20]. However, we predict a rapid rebound of MDR organisms with reversion to the prescribing of firstline antimicrobials due to the circulation of MDR plasmids in S. Typhi and other organisms [8,21].

Table 4. Fever Clearance Time (in hours) for 4 Enteric Fever Patient Populations by Treatment
Our study period captured dynamic changes in MICs against fluoroquinolones, particularly among S. Typhi isolates in more recent years. Through whole genome sequencing, we have determined that this rise in MIC is associated with the emergence of an H58 variant with mutations in the DNA gyrase gene (gyrA) and the DNA topoisomerase IV gene (parC) [10,16]. Supporting these findings, we can conclusively show that FCTs and the rate of treatment failure increases with elevated MICs in S. Typhi patients treated with a fluoroquinolone, confirming results from small studies conducted elsewhere [7,22]. However, although S. Paratyphi A isolates had significantly higher MICs against all tested fluoroquinolones in comparison to S. Typhi, poor outcome was not significantly associated with increasing MIC. We suggest continued surveillance of S. Paratyphi A in the region to monitor for the emergence of high-level fluoroquinolone-resistant organisms similar to trends in the S. Typhi population.
As highlighted in our most recent RCT, patients with suspected enteric fever who were blood culture negative were treated effectively with gatifloxacin, yet fared less well when treated with ceftriaxone [16]. The present analysis shows that ofloxacin also performs well in treating those with culture-negative enteric fever. However, due to the low sensitivity of blood culture for the detection of S. Typhi and S. Paratyphi A [23], it is likely ofloxacin may have been effective against undetected enteric fever cases. We have documented that a reasonable proportion (22%, 21/96) of patients enrolled in the third trial included in the present analysis [14] who were blood culture negative were serologically positive for murine typhus [24]. Doxycycline is considered the drug of choice for rickettsial infections, although it seems that fluoroquinolones may also have clinical activity [24].
In 2003 the World Health Organization published guidelines that recommended azithromycin, ceftriaxone, or cefixime for quinolone-resistant S. Typhi and S. Paratyphi A infections [23].   Azithromycin is safe and efficacious for the treatment of uncomplicated typhoid [25,26]. Although there are no current clinical MIC breakpoints, the majority of isolates (88%) here were susceptible, using the previously suggested cutoff of <16 µg/ mL [27]. The low MICs against ceftriaxone and rapid FCTs throughout the study period indicate that this drug is likely to be effective for culture-confirmed enteric fever in Nepal. The cost and parenteral route of administration, however, make ceftriaxone less suitable for patient treatment in low-and middle-income countries, particularly as 60%-90% of enteric fever patients are treated as outpatients [3]. An alternative would be the oral, third-generation cephalosporin cefixime. However, our first trial, in which we compared gatifloxacin with cefixime, had to be stopped early by the Data Safety Monitoring Board because of the high failure rate in the cefixime arm (26/77) compared to the gatifloxacin arm (5/92; OR, approximately 9), despite all strains being cefixime susceptible [13]. Our analysis supports a recommendation for azithromycin or ceftriaxone for culture-confirmed enteric fever, and in the absence of rapid diagnostics for rickettsial infections [28], a combination of ceftriaxone and doxycycline in culture-negative febrile patients in this setting [16]. However, identification of extended-spectrum β-lactamase-producing S. Paratyphi A in India again suggests vigilance is required.
Our study has limitations. First, the poor diagnostic sensitivity of blood culture may lead to misclassification of a significant number of patients. Although a proportion of culture negatives are likely to be positive for Rickettsia spp., this was not directly assessed [24]. Furthermore, by combining patients from individual RCTs with some differing definitions, the data became nonrandomized; however, we included a random effect of study to account for heterogeneity between studies and controlled for age. Therefore, strong associations, such as odds of treatment failure between cefixime and gatifloxacin in culture-positive patients, may be reduced with the larger, nonrandomized data. Additionally, we were unable to access pharmacy records to evaluate the relationship of prescribing patterns for febrile patients and MICs against common antimicrobials. Notwithstanding these limitations, the results from this largest collection of trials with patient recruitment spanning a decade in an endemic location with a high burden of disease will help to inform therapy recommendations.
In conclusion, poor sanitation, low vaccine uptake, and the emergence of extensive ciprofloxacin-resistant S. Typhi in Kathmandu suggest that appropriate antimicrobial usage policies are required in order to limit morbidity, mortality, and transmission. In this large evaluation, we document shifting antimicrobial susceptibility profiles, an association between poor treatment outcome, and S. Typhi MICs in patients treated with a fluoroquinolone and again highlight the need for better diagnostics for febrile diseases in this setting. We reiterate that fluoroquinolones should not be recommended for the empirical treatment of enteric fever in South Asia [8,29] and advocate the use of azithromycin or ceftriaxone, in addition to surveillance for changes in AMR profiles.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.