Details of the vaccine trial, the CNS study, and the surveillance program have been described previously [9–11]; for background, we briefly summarize these three projects. First, the vaccine trial enrolled 1,696 children aged 6 weeks to 17 months from 2009 through 2013, and participants were followed for up to 53 months. Blood culture was performed for participants who were hospitalized or were evaluated for prolonged fever at outpatient clinics [9]. Second, the CNS study enrolled 320 participants aged ≥6 weeks who were hospitalized and met criteria for evaluation for acute central nervous system infection from November 2011 through December 2013 [10]. Blood and cerebrospinal fluid (CSF) culture were performed for all participants. Third, the surveillance program, which covers a population of about 25,000 persons, has conducted population-based infectious disease surveillance since 2005. Blood cultures are performed for outpatients meeting the case definitions for severe acute respiratory illness and acute febrile illness and for all inpatients admitted for conditions unrelated to injury or obstetric causes [11]. We analyzed surveillance program data from January 2009 through April 2014. The protocols for all three projects included routine culture of participant blood, and for all three projects, isolates were identified as NTS and serogrouped at the Kenya Medical Research Institute (KEMRI) laboratory in Kisumu using standard techniques [6, 9]. For all three projects, KEMRI assessed antimicrobial susceptibility using Kirby-Bauer disk diffusion methods and Clinical and Laboratory Standard Institute (CLSI) guidelines for determining resistance [12]. The detection of ceftriaxone-resistant serogroup O4 (formerly serogroup B) NTS isolates triggered the current investigation, and we conducted a secondary analysis of existing laboratory, epidemiology, and clinical data from the three projects described above.

We conducted a secondary analysis of existing data on cases of laboratory-confirmed invasive NTS infections identified through the vaccine trial, the CNS study, and the surveillance program, including all cases in which a blood culture yielded NTS. If NTS was isolated more than once from a person, we included the first isolate, except in cases in which ceftriaxone-susceptible and ceftriaxone-resistant isolates were both identified, when we included the first ceftriaxone-resistant isolate. Using existing epidemiologic, clinical, and antimicrobial susceptibility testing (AST) data from the three projects, ceftriaxone-resistant NTS infections were compared to ceftriaxone-susceptible infections; for some variables, information was available from only one or two of the projects; the data available from each project is detailed in Table 1. We conducted univariate analysis to calculate odds ratios (OR) with 95% confidence intervals (CI) for ceftriaxone-resistant infections, as compared with ceftriaxone-susceptible infections. We compared categorical data using the χ² test or Fisher’s exact test, as appropriate, and continuous data using the Mann-Whitney U-test. Statistical analyses were conducted using SAS 9.3 (SAS Institute, Cary, NC).

We conducted surveys (S1, S2 and S3 Appendices) at 3 hospitals (including 7 healthcare worker respondents), 23 pharmacies (including 23 pharmacy staff respondents), and 15 agrovets (shops that supply drugs, feeds, and other products for livestock; including 15 staff respondents). We included all registered facilities in Lwak and Siaya towns in Siaya county. The surveys covered questions about the availability and administration of antimicrobials, how patients or their families purchase and use antimicrobials, common selling practices, the type of livestock that were commonly administered antimicrobials, and the presence of antimicrobials in animal feeds and supplements that were available for sale on the day of the survey. Participation was voluntary and anonymous, and we received verbal permission from all the survey participants.

A convenience sample of 31 serogroup O4 NTS isolates (12 from the vaccine trial, 13 from the CNS study, and 6 from the surveillance program) was serotyped and tested for antimicrobial susceptibility using the broth microdilution method per CLSI guidelines in the Enteric Diseases Laboratory Branch (EDLB) at CDC [12, 13]. Whole genome sequencing was performed on 4 these isolates from the CNS study and the surveillance program using methods as described by Tagg K.A et.al. [14].

The Kenyan Ministry of Health and CDC determined that the activities we conducted for this public health response to the emergence of ceftriaxone-resistant Salmonella infections did not meet the definition of research. Therefore, formal review by research ethics boards was not required for the activities we report.

In total, 286 invasive NTS infections were identified in the three projects; 43 NTS isolates were ceftriaxone-resistant (Table 2). The absolute prevalence of ceftriaxone resistance varied between these methodologically diverse projects, with 18% (16/90) of NTS isolates resistant to ceftriaxone in the vaccine trial, 89% (16/18) in the CNS study, and 6% (11/178) in the surveillance program (Table 2). The percentage of the NTS that was ceftriaxone-resistant increased over time in all three projects; from 8% (1/12) in 2009 to 64% (7/11) in 2013 for the vaccine trial, from 80% (8/10) in 2012 to 100% (8/8) in 2013 for the CNS study, and from 2% (1/41) in 2009 to 25% (1/4) in 2014 for the surveillance program (Fig 1). The NTS isolates that were resistant to ceftriaxone were also co-resistant to other antibiotics tested; all isolates were co-resistant to trimethoprim-sulfamethoxazole and sulfisoxazole (43/43 and 40/40, respectively), and almost all were co-resistant to ampicillin (98%, 42/43), gentamicin (95%, 41/43), tetracycline (95%, 38/40) and chloramphenicol (93%, 40/43) (Table 3).

Although not statistically significant, presentation with diarrhea and vomiting was more common in patients with ceftriaxone-susceptible invasive NTS infections than ceftriaxone-resistant NTS infections. The percentage of surveillance program participants who were hospitalized and the duration of their hospital stay did not differ between patients with ceftriaxone-resistant and ceftriaxone-susceptible invasive NTS infections. No deaths were reported among the patients with ceftriaxone-resistant infection.

Kenyan informants reported that the third-generation cephalosporins first became available as early as 2005 in western Kenya. According to hospital respondents, the third-generation cephalosporins were “always” prescribed for children admitted with suspected meningitis (5 of 7 responders) and for those with suspected sepsis (3/7), and “often” prescribed for patients with acute febrile illnesses (2/7) (S1 Table). Interviews with clinicians revealed several issues related to prescription practices and adherence to prescribed dosages, including drug supply shortages at hospitals requiring inpatients to purchase drugs from outside pharmacies, antimicrobial purchase without a prescription, self-medication, and using antimicrobials in sub-therapeutic doses. In addition to ceftriaxone, clinicians commonly prescribed other antimicrobials, such as amoxicillin, trimethoprim-sulfamethoxazole and gentamicin, to which the ceftriaxone-resistant NTS was found to be co-resistant. Carbapenem antimicrobials were not available at any of the hospitals on the day of interview.

All of the 23 interviewed pharmacist dispensed antimicrobials, and 20 of them sold third-generation cephalosporins (S2 Table). Other beta-lactam antibiotics were available at 22 pharmacies, trimethoprim-sulfamethoxazole at 21, gentamicin at 19, and ciprofloxacin at 19 on the day of interview. Two of the 23 pharmacies had carbapenems available on the day of interview. Of the 23 pharmacies, respondents for 13 reported that they would sell a partial prescription to a customer. During the interviews, we observed that people could purchase antibiotics without a prescription at some pharmacies.

The practice of dispensing antimicrobials for veterinary use was similar to that for human use; antibiotic purchase without a prescription, treatment without veterinarian’s consult and subtherapeutic dosing were common. Agrovets also sold a wide range of antimicrobials for animal use to which ceftriaxone-resistant NTS in humans were co-resistant; all or most of the 15 interviewed agrovets sold sulfonamides (15), beta-lactams (14), tetracyclines (13) and macrolides (13) (S3 Table). In addition to antimicrobials, more than half of commonly sold brands of poultry nutritional supplements had tetracycline or oxytetracycline additives (15 of 27 observed supplements) that could contribute to co-selection of ceftriaxone-resistance.

All 31 ceftriaxone-resistant serogroup O4 NTS isolates referred from the KEMRI laboratory to CDC for further characterization were serotype Typhimurium (S4 Table). Whole genome sequencing of 4 ceftriaxone-resistant NTS isolates from the CNS study and the surveillance program identified several plasmid types and multidrug resistance genes, including bla_CTX-M-15 that confers resistance to broad spectrum cephalosporins and several folate pathway inhibitor genes (dfrA1, dfrA14, sul1, sul2) that confer resistance to trimethoprim-sulfamethoxazole. All four isolates were a single multilocus sequence type, ST313.