Summary of findings
Description of the condition
Infection with Salmonella bacteria can cause typhoid fever in people if they are infected with Salmonella enterica enterica, serovar Typhi (S. Typhi) or S. enterica enterica, serovar Paratyphi (S. Paratyphi) A, B and C. Non-typhoidal Salmonella (NTS) disease is caused if the infectious agent is any of the NTS serovars, such as S. enterica enterica, serovar Enteritidis, or S. enterica enterica, serovar Typhimurium. This review focuses on NTS infection, which can present as either an invasive disease or as enterocolitis with diarrhoea. Another Cochrane Review studied treatments for typhoid fever (Effa 2011). This review examines the currently available body of evidence regarding antibiotic treatment of NTS infection. This review is an update of part of an earlier review which investigated the use of antibiotics for the treatment of both symptomatic and asymptomatic NTS infection (Sirinavin 2000).
NTS infection is an important cause of food poisoning in most areas of the world. The disease is often under-reported as affected people can sometimes be asymptomatic and hence do not go to the hospital for treatment (Rabsch 2001). In the USA, an estimated 1.4 million people suffer from the disease annually, of which about 80,000 to 160,000 seek medical attention, approximately 16,000 are hospitalized and about 600 people die from the disease (Mead 1999). Invasive disease due to Salmonella enterica enterica serovar Typhi as well as NTS is common in children younger than five years old in developing countries, particularly in many places in sub-Saharan Africa (Graham 2002).
Animals are a major reservoir of NTS infection. The infection is mainly acquired by eating contaminated food, such as poultry, beef and eggs. However, it can also be transmitted by handling farm animals, like chickens. Infection can be passed transovarially from chickens to their eggs. Furthermore, bacteria can be spread by pets, including snakes. There has been a report of fatal Salmonella sepsis following platelet transfusion from an asymptomatic donor who acquired the infection from his pet boa constrictor (Jafari 2002). Salmonella bacteria can also be transmitted from person to person by the faecal-oral route, by direct contact with a contaminated person or fomites, medicines and rarely by aerosols (Mason 2000).
Although the findings of a prospective study in Africa highlighted the importance of person-to-person transmission in Kenya (Kariuki 2006), animal-to-human transmission is still recognised as being more important in accounting for the current epidemiological patterns of NTS.
Certain host factors, such as gastric acidity, can give some protection from NTS infection and infection usually requires large bacterial inocula. However, in people whose host defence mechanisms have been compromised, for example those on acid-suppressing drugs, patients with pernicious anaemia and infants, there is a higher risk of NTS infection. Notably, liquids which pass through the stomach quickly, or milk and cheese that raise the pH, enable smaller inocula to be infective.
NTS infection can manifest in two distinct forms: either as an enterocolitis with diarrhoea or as an invasive disease, which can occur without diarrhoea. The latter form is particularly common in sub-Saharan Africa.
The syndrome of enterocolitis is more often present in developed countries and usually manifests with diarrhoea, abdominal pain and cramps, and sometimes fever. Symptoms usually start between six to 72 hours after exposure to the bacteria (but can sometimes be delayed for up a week) and tend to resolve within five to seven days (Hohmann 2001; MDH 2007). In general the incubation period depends on the immune system of the host and the bacterial inoculum size. Infection can cause acute severe diarrhoea or chronic and prolonged diarrhoea, which can result in the disturbance of fluid and electrolyte balances (Mason 2000). NTS infection can be severe, invasive and recurrent in patients with human immunodeficiency virus/acquired immune deficiency syndrome (HIV/AIDS), resulting in up to 47% mortality (Gordon 2002; Kankwatira 2004).
The infection can sometimes be invasive when it causes bacteraemia (bacteria in the bloodstream) or has extraintestinal manifestations (Chen 2007; Ispahani 2000). About 2% to 45% of people with diarrhoea may develop bacteraemia (bacteria in the bloodstream) with fever (Zapor 2005), while some may develop bacteraemia without diarrhoeal episodes (Boyle 2007). NTS can cause life-threatening infection in some individuals, especially those with HIV/AIDS. Recurrent Salmonella septicaemia is one of the conditions that defines the AIDS (Boyle 2007). Children with sickle cell anaemia are particularly at risk of NTS osteomyelitis.
Extraintestinal manifestations can result in complications, with various clinical focal syndromes affecting the meninges, bones, joints, adrenal gland, aorta, inner lining of the heart, the kidneys/urinary tract, and the lungs (Diez Dorado 2004; Zapor 2005). The risk factors for invasive disease in adults and children include immunosuppression of any cause, including HIV-positive status, malaria infection, severe anaemia, or malnutrition (Morpeth 2009). The development of extraintestinal focal infections is associated with higher mortality rates, more severe septic manifestations, longer hospital stays and a longer duration of antimicrobial therapy (Chen 2007).
In an attempt to further understand the molecular biology of the NTS strains responsible for invasive disease in sub-Saharan Africa, multilocus sequence analysis of certain strains of S. Typhimurium from patients in Kenya and Malawi was performed. A dominant genotype, ST 313, was identified which is responsible for many cases of the invasive disease. ST 313 isolates harbour genome signatures that differentiate them from S. Typhimurium causing gastroenteritis in other regions of the world. These include a novel repertoire of prophage elements and evidence of genome degradation (Kingsley 2009). In Africa, ST 313 infection presents as a separate clinical entity with generalized sepsis and focal infection due to its adaptability to the human host.
Some people may have infection caused by NTS without showing any symptoms of the disease while excreting the organism in their stool (asymptomatic carriers) (Jertborn 1990). In certain cases following recovery from symptoms of the disease, individuals continue to excrete NTS bacteria in their stools (convalescent carriers) (Buchwald 1984). These carrier states can be for a short period of time. However, excretion of the organism may be prolonged especially in children aged less than five years old and can persist for more than a year (chronic carriers) (Buchwald 1984; Mason 2000). A carrier is considered cured from the first day of three consecutive negative stool cultures (ie when the Salmonella is absent in the stool), and this may be difficult to achieve with antibiotic treatment (Carlstedt 1990).
Salmonellosis is diagnosed by isolating Salmonella bacteria from the stool, blood (if associated with bacteraemia and extraintestinal infection) (Kankwatira 2004), or, less commonly, urine (if there is a focal infection of the urinary tract) (Diez Dorado 2004; Vallenas 1985). Salmonella bacteria can also be isolated from bone marrow aspirates. The bacterial concentration in bone marrow can sometimes be as much as 10 times that in peripheral blood. In patients who have received antibiotic treatment, the bacteria may still be found in the bone marrow even when it may no longer be present in the blood when cultured.
Description of the intervention
Antimicrobial agents are either natural or synthetic substances which can kill or inhibit the growth of microbial organisms. They are generally described based on their mechanism of action which may determine if a particular antibiotic may be clinically useful for the treatment of an infection.
How the intervention might work
There are several reasons why clinicians have concerns regarding the use of antibiotics to treat NTS infection. Antibiotic use may not result in rapid control of symptoms or stop the excretion of the bacteria in stools. Instead it may lengthen the time period that bacteria are excreted in the stools, thereby increasing the risk of infecting other people (Lin 2003). Concerns about the development of antibiotic resistance have limited their use for NTS treatment (Hakanen 2006; Molbak 2002; Panhotra 2004; Rowe 1997) and infection with multiple-drug resistant strains of NTS has been noted to result in higher mortality and morbidity rates.
There have been several reports regarding the emergence of antibiotic resistant strains of NTS, particularly following the increased and more widespread use of antibiotics for treatment of NTS infection in livestock. The number of cases appears to be increasing (Frost 1996; Hakanen 2006; Molbak 2002).
Notably, some people continue to carry Salmonella bacteria even after the antibiotics have treated the symptoms. For example, the previous version of this Cochrane Review showed that antibiotic treatment may result in more negative stool cultures especially during the first week of treatment, with more positive stool cultures after the third week of treatment due to relapse of infection (Sirinavin 2000). Also, there are risks of adverse drug reactions to these antibiotics, such as skin rash with ampicillin, leucopenia with co-trimoxazole, and urticaria, severe headache, nausea, epigastric pain, and dizziness with fluoroquinolones (Reese 1991). There are also concerns regarding the use of quinolones in young children because of the risk of tendonitis (Yee 2002).
Why it is important to do this review
This review is an update of certain aspects of a previous Cochrane Review, which investigated the use of antibiotics for treating Salmonella gut infections in both symptomatic and asymptomatic people (Sirinavin 2000). Since the review, the use of new antibiotics, such as fluoroquinolones for adults and third-generation cephalosporins for children, has become more widespread and new trials have been conducted using these drug interventions. In this review, we have focused our investigation on NTS symptomatic patients only. We have updated the review methods to reflect recent methodological changes and we searched for new trials taking into account these changes.
To evaluate the efficacy and safety of antimicrobial agents for treating NTS infection.
Criteria for considering studies for this review
Types of studies
Randomized controlled trials (RCTs).
Types of participants
People with culture-proven NTS infection (excluding S. Typhi and S. Paratyphi A, B and C).
We also included studies that investigated diarrhoea in general and analysed patients with culture-proven NTS patients as a subgroup.
Studies evaluating only asymptomatic patients were excluded.
Types of interventions
Oral or parenteral antibiotic (at any dose and for any duration of treatment).
Placebo or no treatment.
Types of outcome measures
- Presence of diarrhoea at two to four days afer randomization.
- Duration of diarrhoea.
- Presence of diarrhoea at five to seven days.
- Clinical failure (defined as worsening or persistent symptoms at the end of the treatment regime).
- Presence of fever at two to four days (from commencement of treatment/randomization).
- Duration of fever (from randomization).
- Duration of illness.
- Presence of life-threatening extraintestinal focal infection (meningitis, septic arthritis, pneumonia, osteomyelitis, bacteraemia, pyelonephritis).
- All cause death.
- Microbiological failure (defined as culture-proven Salmonella infection at the end of the treatment regime).
- Faecal carriage of the same Salmonella serovar one month after the end of antibiotic treatment.
- Other adverse events.
Search methods for identification of studies
We attempted to identify all relevant trials regardless of language or publication status (published, unpublished, in press and in progress).
We searched the following databases using the search terms detailed in Table 1: Cochrane Infectious Disease Group Specialized Register (up to August 2012); Cochrane Central Register of Controlled Trials (CENTRAL) published in The Cochrane Library (Issue 8 2012); MEDLINE (from 1966 to 6 August 2012); African Index Medicus (accessed on 14 August 2012), CINAHL (from 1981 to 6 August 2012), EMBASE (from 1980 to 6 August 2012); LILACS (from 1982 to 6 August 2012); and the Science Citation Index (from 1970 to 6 August 2012). We also searched the metaRegister of Controlled Trials (mRCT) on 6 August 2012 for both completed and ongoing trials ( Table 2) and the reference lists of relevant articles.
Searching other resources
Organizations and pharmaceutical companies
To help identify unpublished and ongoing trials, we contacted the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC) and the pharmaceutical companies Pfizer and GlaxoSmithKline. We also searched the WHO Clinical Trials platform for relevant ongoing trials.
We searched the following conference proceedings for relevant abstracts: the International Symposium on Typhoid Fever and Other Salmonellosis (from 2000 to 2010) and the International Symposium on Invasive Salmonelloses (from 2000 to 2008).
We checked the reference lists of all studies identified by the above methods
Data collection and analysis
Selection of studies
Two authors (IO, CO) independently screened the results (titles and abstracts) of the literature search for potentially relevant trials. We retrieved full text articles of the potentially relevant trials and independently determined whether they met the review inclusion criteria using a pre-tested eligibility form. For each step of the review, we resolved contentious issues through discussion. We consulted an editor from the Cochrane Infectious Disease Group where necessary. We also attempted to contact trial authors for further information where trial eligibility was unclear. We have listed all excluded studies along with the reason for exclusion (see Characteristics of excluded studies). We ensured that trials with multiple publications were included only once.
Data extraction and management
Two authors (IO, CO) independently extracted data using a pre-tested data extraction form. One author (CO) entered the data into Review Manager 5 while a second author (IO) cross-checked the data for completeness and accuracy. We extracted data concerning the number of participants randomized and the number of participants analyzed in each group for each reported outcome. For dichotomous outcomes we extracted data concerning the total number of participants randomized, the number of participants experiencing the events and number of participants in each treatment group. For continuous outcomes, we extracted the number of participants for each treatment arm, arithmetic means and standard deviations. Where we encountered data with skewed distribution, we extracted geometric means and standard deviations on the log scale where geometric means were reported, or medians and ranges if medians were reported. For rate and count outcomes (such as participants with outcomes that occurred more than once over the period of trial), we extracted the number of events or episodes experienced in each trial arm and person-time over which the events were experienced for each group. We extracted hazard ratios and standard deviations for time-to-event outcomes (such as the development of life-threatening extraintestinal focal infection). We extracted data on reported adverse events. We contacted the trial authors where the relevant details were not recorded or were unclear. We resolved any disagreements regarding data extraction through discussion and by asking the third review author to attempt data extraction. If necessary, we also sought assistance from an editor with the Cochrane Infectious Diseases Group.
Assessment of risk of bias in included studies
Two review authors (IO, CO) assessed the risk of bias independently according to the specifications of the latest edition of the Cochrane Handbook (Higgins 2011). We independently assessed the risk of bias within each included study in relation to the following five domains:sequence generation, allocation concealment, blinding, handling of incomplete outcome data and selective outcome reporting by using the ratings of 'Yes' (low risk of bias); 'No' (high risk of bias) and 'Unclear' (uncertain risk of bias).
Details of specific assessments are as outlined in the Cochrane Handbook (Higgins 2011).
Measures of treatment effect
We analysed continuous data if means and standard deviations were available. Where mean differences were provided, we extracted and utilized these values for the analysis irrespective of whether mean and standard deviation values were provided as we were interested in post-intervention values. We re-calculated the standard deviation values in instances where the standard error was reported. Also, we extracted data from studies that reported adequately on skewed continuous data as medians rather than means. Where appropriate, we have reported these data separately.
We analysed binary outcomes by calculating the risk ratio (RR) with 95% confidence interval (CI).
Dealing with missing data
When necessary, we attempted to contact the study author(s) to supply any unreported data (eg group means and standard deviations (SDs), details of dropouts, and details of interventions received by the control group). If a study reported outcomes for participants that completed the trial only or for participants who followed the protocol only, we contacted authors to provide additional information to facilitate intention-to-treat analyses. In instances where this was not possible we performed a complete case analysis.
Assessment of heterogeneity
We assessed statistical heterogeneity by examining the I
Assessment of reporting biases
We had planned that if there are more than 10 trials in a comparison, we will prepare funnel plots (estimated treatment effects against their standard error) to explore publication bias. Asymmetry could be due to publication bias, but can also be due to a relationship between trial size and effect size. As we did not identify at least 10 trials for any comparison, we did not prepare funnel plots to explore publication bias.
We conducted meta-analyses for trials with similar characteristics. We aimed to carry out an intention-to-treat analysis but we carried out a complete-case analysis where there was loss to follow-up. We used the fixed-effect model and presented all our results with 95% CI.
Subgroup analysis and investigation of heterogeneity
We planned to conduct subgroup analyses to assess the benefit of antibiotic treatment. Subgroups were as follows: participant age (infants < 1 year versus elderly > 60 years); route of drug administration (oral versus parenteral); hospitalization (hospitalized versus not hospitalized); and type of antibiotic (fluoroquinolone versus other antibiotics). Where there was sufficient data, we conducted subgroup analyses to investigate the effect of the antibiotic on the absence of diarrhoea at two days and at four days post treatment.
We planned to assess important clinical heterogeneity by comparing the distribution of important clinical (study participants, study setting, type of intervention and co-intervention) and methodological (randomization, allocation concealment, blinding of outcome assessment, losses to follow up) heterogeneity factors. However this was not possible due to insufficient data.
Also, we could not perform most of the planned subgroup analyses because of insufficient data.
We conducted sensitivity analyses to explore the effect of the methodological quality of the trials and to ascertain whether studies with a high risk of bias overestimated the effect of treatment.
Description of studies
Results of the search
Our search for this review ( Table 1) retrieved 70 potentially relevant records after duplicate records were removed. This search was last updated on the 6 August 2012 with no new relevant additions. We found one ongoing trial that met our eligibility criteria (Tsai 2012) .
Types of Studies
Twenty trials met our initial inclusion criteria but we excluded eight of these studies because patients with diarrhoea of different infectious aetiologies were randomized and the data for the Salmonella subgroup was not reported in a manner that would be of use to our review (Bessudo 1972; Dryden 1996; Lolekha 1988; Noguerado 1995; Pichler 1986; Pichler 1987; Robins-Browne 1983; Taylor 2006). We therefore included 12 studies in our review, which reported information regarding 767 patients with NTS.
Types of Patients
We included one study that involved both children (aged 12 years) and adults (Wistrom 1992). Five studies (n = 323) included adolescents and adults (Butler 1993; Goodman 1990; Neil 1991; Pitkajarvi 1996; Sanchez 1993) and five studies (n = 284) included infants aged over 6 weeks and children (Chiu 1999; Garcia de Olarte 1974; Kazemi 1973; Macdonald 1954; Nelson 1980). One study (n = 168) included all ages (Joint Project ASID 1970). Almost all studies excluded pregnant patients and those with underlying diseases, previous antibiotic treatment, severe illness and history of allergy to the group of study drug. One study included malnourished children (Garcia de Olarte 1974).
Eleven studies involved sporadic cases of patients presenting for treatment of either acute diarrhoea or travellers' diarrhoea. One study reported an outbreak in hospital personnel in the USA (Neil 1991). Randomization was conducted on diarrhoeal patients prior to culture results being available in eight studies (Butler 1993; Garcia de Olarte 1974; Goodman 1990; Kazemi 1973; Neil 1991; Nelson 1980; Sanchez 1993; Wistrom 1992). All studies included symptomatic patients, but two also included asymptomatic patients (Neil 1991; Pitkajarvi 1996). For assessment of microbiological failure, we used data that combined both symptomatic and asymptomatic patients as the distribution of asymptomatic patients was similar in both the treatment and control groups.
Duration of diarrhoea preceding entry to the study varied between the included studies, however the duration was similar between control and experimental groups in each study. In eleven studies, the history was short (< 7 days). One study had a range of between 1 to 34 days (Nelson 1980). One study did not specifically state the duration (Joint Project ASID 1970). Garcia de Olarte 1974 included patients who were less than and more than seven days with the diarrhoea. In one study, patients were randomized on the ninth day following onset of symptoms (Neil 1991).
Studies were from Europe and Scandinavia (four studies), North America (four studies), Australia (one study) Taiwan (one study). There were two international multicentre studies: one included Italy, Thailand, Indonesia, Ivory coast, Mexico and Israel and the second included Asia, South America and Italy. There was one study from Colombia.
Exact Salmonella serovars were not reported in all of the studies. The outbreak assessed in Neil 1991 was caused by S. java. About 90% of culture positive cases were caused by S. enteritidis in one study (Pitkajarvi 1996), and by S. typhimurium in another (Macdonald 1954). In two studies in infants and children (Kazemi 1973, Nelson 1980), S. typhimurium was the cause in 31% and 53% of the patients, respectively.
Types of Intervention
Ten different drugs were investigated including: norfloxacin (two studies, Pitkajarvi 1996; Wistrom 1992), cotrimoxazole (three studies Goodman 1990; Sanchez 1993; Kazemi 1973), ampicillin (three studies Garcia de Olarte 1974; Kazemi 1973; Nelson 1980), ciprofloxacin (three studies Goodman 1990; Neil 1991; Sanchez 1993 ), neomycin (one study Joint Project ASID 1970), chloramphenicol (one study Macdonald 1954), amoxycillin (one study Nelson 1980), azithromycin (one study Chiu 1999), cefixime (one study Chiu 1999 ) and fleroxacin (one study Butler 1993 ). Nine studies included a placebo comparison, and three studies included a comparison against no treatment. Dose schedules, route of administration and duration varied across trials (see Characteristics of included studies).
Duration of treatment varied between three to 14 days. One study included single dose treatment (fleroxacin) (Butler 1993), but all the rest of the studies included multiple dose treatment. Seven trials had treatment regimens that lasted for five days (Chiu 1999; Garcia de Olarte 1974; Goodman 1990; Joint Project ASID 1970; Nelson 1980; Sanchez 1993; Wistrom 1992). One trial lasted for three days (Butler 1993), three trials had regimens that lasted between 10 to 14 days (chloramphenicol, norfloxacin, or ciprofloxacin; Macdonald 1954; Neil 1991; Pitkajarvi 1996), and one trial lasted for seven days (Kazemi 1973).
In all the included trials, most of the Salmonella strains were sensitive to the study drugs. One study reported on Salmonella strains resistant to ampicillin, which was the drug used in the study. This resistance was reported in three patients treated with ampicillin but not enough data was provided to enable comparison with the placebo group.
The period of follow-up varied between six months and five days. In two studies (Butler 1993; Garcia de Olarte 1974), follow-up was less than 14 days and was 14 days in two studies (Goodman 1990; Macdonald 1954). Length of follow-up was between five to eight weeks in five studies (Chiu 1999; Neil 1991; Nelson 1980; Sanchez 1993; Joint Project ASID 1970), three months in one study (Wistrom 1992) and six months in two studies (Kazemi 1973; Pitkajarvi 1996). However, in the longer periods of follow-up the number of evaluable patients dropped considerably.
Differences between the studies in the present review and the previous version
This review includes only RCTs that have investigated the use of antibiotics for the treatment of symptomatic NTS infection. We have excluded quasi-RCTs and trials that have investigated the use of antibiotics in the treatment of asymptomatic infection. These trials were included in the earlier version of the review (Sirinavin 2000).
Risk of bias in included studies
|Figure 1. Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.|
|Figure 2. Methodological quality summary: review authors' judgements about each methodological quality item for each included study.|
Generation of allocation sequence was reported and judged to be adequate in six studies (Butler 1993; Chiu 1999; Garcia de Olarte 1974; Macdonald 1954; Nelson 1980; Sanchez 1993). It was not reported and judged to be unclear in six studies(Goodman 1990; Joint Project ASID 1970; Kazemi 1973; Neil 1991; Pitkajarvi 1996; Wistrom 1992).
No studies explicitly reported concealment of allocation.
With regard to blinding, 10 studies were described as double blinded and also gave matching placebo to the control arm (Butler 1993; Garcia de Olarte 1974; Goodman 1990; Joint Project ASID 1970; Macdonald 1954; Neil 1991; Nelson 1980; Pitkajarvi 1996; Sanchez 1993; Wistrom 1992). Two studies were not blinded (Kazemi 1973; Chiu 1999) and we therefore judged these as having a high risk of bias.
Incomplete outcome data
Five studies (Garcia de Olarte 1974; Kazemi 1973; Macdonald 1954; Nelson 1980; Wistrom 1992) did account for incomplete outcome data. We judged Neil 1991 as unclear with regard to incomplete outcome because we were unable to assess how the trialists dealt with incomplete outcome data.
Regarding selective outcome reporting, we did not have the trial protocol for the included studies and could not determine whether the authors had reported them selectively or not. However, we had no reason to believe they were selectively reported and we have judged this factor as low risk in all studies.
Other potential sources of bias
Six studies (Garcia de Olarte 1974; Goodman 1990; Kazemi 1973; Neil 1991; Pitkajarvi 1996; Wistrom 1992) were funded by drug companies and the authors have not made any statements regarding the extent of involvement of these companies in the design, conduct, analysis and reporting of the trials. In four studies (Butler 1993; Joint Project ASID 1970; Macdonald 1954; Nelson 1980) we do not have information regarding the source of funding. One study (Chiu 1999) was funded by the National Health Research Council and one study (Sanchez 1993) was funded by the hospital's Department of Medicine. We judged as unclear those studies that did not provide information regarding the source of funding and we judged as low risk those studies that were funded by neutral bodies such as research institutes and hospital departments.
Effects of interventions
Presence of diarrhoea at two to four days
Duration of diarrhoea
This was reported in nine trials, but only four studies provided a measure of variance (Chiu 1999; Macdonald 1954; Nelson 1980; Sanchez 1993). A fixed-effect meta-analysis using the generic inverse variance method yielded a mean difference of -0.00 days (95% CI -0.54 to 0.54), thus not detecting any impact of antibiotics on duration of diarrhoea ( Analysis 1.2). However, two of the studies (Chiu 1999; Nelson 1980) in this meta-analysis had skewed data. This finding appears to be consistent across some of the other studies listed in Table 3.
Presence of diarrhoea at five to seven days
Clinical assessment at 5 to 7 days post treatment was reported in two trials with data that we could use (Garcia de Olarte 1974; Wistrom 1992); no effect of antibiotics was demonstrated (RR 0.83, 95% CI 0.62 to 1.12; Analysis 1.3).
We defined this as worsening or persistent symptoms at the end of treatment, and we were able to assess this in seven trials (Butler 1993; Chiu 1999;Garcia de Olarte 1974; Macdonald 1954; Nelson 1980; Sanchez 1993). No effect was detected overall (RR 0.88, 95 % CI 0.62 to 1.25; Analysis 1.4).
In the study by Pitkajarvi 1996, one of the patients was reported to have diarrhoea at day 10 when the treatment regimen had ended. We judged this patient to have encountered treatment failure but the trialists did not give specific information as to the group to which this patient belonged.
Presence of fever at two to four days
No studies defined and reported this outcome in a way that would enable us to extract it as a separate outcome for further analysis.
Duration of fever
Chiu 1999 and Sanchez 1993 reported on this outcome and we performed a meta-analysis to examine the overall effects of antibiotics on this outcome. There were differences in study characteristics in terms of their patient population and antibiotic intervention, which gave a mean difference of 0.27 days (95% CI -0.11 to 0.65; Analysis 1.5) and showed no difference that could be attributed to antibiotics . However, this result was generated with skewed data. This finding is consistent across the other studies that we could not combine. We have listed them in Table 4.
Duration of illness
We were able to analyse the results for this outcome from two studies (Macdonald 1954; Sanchez 1993), which demonstrated no difference in duration of illness between the groups that could be accounted for by antibiotics (Mean difference -0.00 days, 95% CI -0.68 to 0.68; Analysis 1.6).
This finding was also found to be consistent with the other studies we could not perform meta-analysis on ( Table 5).
This outcome was reported by Joint Project ASID 1970 in the form of a graph and we attempted to extract this data from the graph. However we encountered two problems: firstly, the total percentage of patients reported in the graph did not add up to a 100% so it is possible that not all the patients were included in that analysis, and secondly, the standard deviation was not reported and could not be calculated.
Presence of life-threatening extraintestinal focal infection
There was no information on this outcome in any of the included studies. This may be partially because all existing studies excluded the types of patients in which this complication could have been more likely.
Garcia de Olarte 1974 reported 12 deaths among the patients in their trial, two of which occurred in patients culture proven for Salmonella. However, the trialists did not mention the group to which they were randomized.
Butler 1993 reported three deaths and these were not in the Salmonella subgroup of patients.
Salmonella cultures were conducted at varying periods in the included studies after the start of treatment. Many studies excluded from follow-up patients that had become culture negative (based on two to three consecutive negative cultures) so they could not detect patients who relapsed. Also, some studies had high dropout rates.
We defined this as culture-proven Salmonella infection at the end of therapy. We were able to extract data on this outcome from eight trials (Butler 1993; Garcia de Olarte 1974; Goodman 1990; Joint Project ASID 1970; Kazemi 1973; Macdonald 1954; Neil 1991; Pitkajarvi 1996) to enable meta-analysis to be performed. We conducted an a-priori subgroup analysis to investigate the differential impact of quinolone antibiotics versus placebo or no treatment compared to other antibiotics versus placebo or no treatment. Quinolone antibiotics appeared better at preventing microbiological failure when compared to placebo or no treatment (RR 0.33, 95 % CI 0.2 to 0.56; Analysis 1.7).There was no difference between antibiotic and no treatment with regard to microbiological failure with other antibiotics. We excluded the Wistrom 1992 study from the quinolone subgroup because outcome assessment in this study occurred much later after treatment and the results increased the inherent heterogeneity in the comparison. It showed no advantage over treatment with quinolones. The study with the largest weighting in the non-quinolone subgroup (Joint Project ASID 1970) used neomycin which is a non-absorbable antibiotic.
Other included studies in this review did not contribute data to this meta-analysis because they reported this outcome without stratifying according to which organism was cultured for the respective patients in their trials or reported only in qualitative terms.
Faecal carriage of the same Salmonella serovar one month after treatment
We were only able to extract data on this outcome from studies that assessed and reported this outcome for the subgroup of patients of interest in the review (Chiu 1999; Neil 1991; Wistrom 1992). In the study by Wistrom 1992, we extracted data from a graph that reported 39% of placebo patients versus 79% of the norfloxacin patients as being culture positive at 28 to 30 days post treatment. We translated this to 35 out of 45 patients in the antibiotic arm versus 14 out of 37 patients in the placebo arm.
We performed a meta-analysis on these three studies. This showed that antibiotic administration causes a higher incidence of carriage of the same Salmonella serovar at one month post treatment, with 41 of 62 events in the antibiotic group compared to 17 of 50 events in the placebo group (RR 1.96, 95 % CI 1.29 to 2.98; Analysis 1.8).
Nelson 1980 reported bacteriologic relapse in four patients in each of the two antibiotic arms and no relapse in the placebo arm. The relapses reported occurred between day 4 and day 52 post intervention. The difference between antibiotic arms and placebo arm was statistically significant (P = 0.003).
Other studies that assessed this outcome at different times also showed findings that appear consistent with the result of our meta-analysis above (Kazemi 1973; Nelson 1980; Pitkajarvi 1996; Sanchez 1993) and this is summarized in Table 6.
We defined serious adverse events as those leading to death, disability or prolonged hospitalization. We were also interested in adverse events that may have required stopping of treatment, for example gross derangements of biochemical markers of toxicity from baseline, and other adverse events that may have been noted during the course of the treatment trials. Not all of the studies reported adverse events. Some studies reported adverse events as overall events in all diarrhoeal patients randomized to comparative groups (Chiu 1999; Garcia de Olarte 1974; Goodman 1990; Joint Project ASID 1970; Macdonald 1954). Again, the individual studies looked at different antibiotic drug classes and different durations and routes of treatment. We decided not to perform a meta-analysis of the data from these trials regarding adverse events but instead perform a narrative synthesis of the reported events with respect to the antibiotic drug class in line with our a priori subgroup analysis.
Sanchez 1993 reported that 11 patients had slightly raised levels of liver transaminase and one had slight leukopenia. The differences in incidence were not statistically significant between drug and placebo groups. Neil 1991 reported an increase in diarrhoea in five of the eight patients randomized to ciprofloxacin as against one of the eight patients who received placebo. A case of vomiting and two cases of nausea were noted in the ciprofloxacin group. However, all were judged as minor events.
Butler 1993 reported adverse events in 33 patients but there was no significant difference between the groups (antibiotic versus placebo). The most commonly reported symptoms were headache, dizziness, epigastric pain, stomach discomfort and anorexia. No changes were made to therapy.
Pitkajarvi 1996 reported that one patient in the norfloxacin group had nausea that led to discontinuation of treatment on day six of the trial. Wistrom 1992 reported adverse events in 16 and 13 patients in the norfloxacin and placebo groups, respectively. Headache or other central nervous system symptoms were the most common complaint in both groups, reported by 10 and eight patients in the norfloxacin and placebo groups, respectively. Three patients had a severe headache, one in the norfloxacin group and two in the placebo group. One patient reported a severe stomach pain in the placebo group. These adverse events caused a discontinuation of treatment.
Kazemi 1973 reported vomiting and generalized maculopapular rash in three and two patients respectively who received sulphamethoxazole trimethoprim. No patient had his drug discontinued.
Nelson 1980 reported candida skin rash in four infants and children after treatment with ampicillin and in one of the children treated with amoxicillin. There was a report of eosinophilia in two ampicillin treated patients and one each in the amoxicillin and placebo groups. Slight elevations of transaminase enzymes were also noted in two placebo patients and one amoxicillin patient. An elevation of blood urea nitrogen was also noted in one of the amoxicillin patients.
Summary of main results
The results of this systematic review suggest that antibiotics may not be of clinical benefit for the treatment of NTS diarrhoea. We were unable to demonstrate any statistically significant effect of antibiotic treatment on any of our clinical outcomes of intervention efficacy. Antibiotic administration appeared to increase the risk of microbiological relapse and fecal carriage at follow-up in patients who were treated compared to those who were either not treated or treated with placebo. Although no serious adverse events were reported among the patients in our included studies, a slightly higher number of other adverse events were associated with the use of antibiotics as compared with placebo or no treatment. Although most of the authors did not report statistically significant differences between the antibiotic and placebo groups, the observed adverse events are of enough clinical significance to discourage antibiotic use, particularly when its use is of questionable benefit.
These findings are of importance both from clinical and public health perspectives. Routine antibiotic administration for the treatment of NTS diarrhoea could potentially worsen disease transmission in the community as many of the treated patients could go on to excrete pathogens for longer periods than they normally would if they were not treated with antibiotics. This would have some impact on the incidence of acute bacterial diarrhoeal episodes. There is also the potential for the spread of resistant strains with the use of unnecessary antibiotics. The question regarding the appropriateness or otherwise of antibiotic administration with regard to NTS diarrhoea is one that has been controversial with respect to the findings that have been made in studies that have attempted to answer this question.
Notably, the actual number of patients in the included studies and subsequent meta-analysis were few and when taken alone, may not be enough to detect a statistically and clinically meaningful difference. However, the direction of effect was fairly consistent across the studies in the review.
Overall completeness and applicability of evidence
The studies we included in this review were mostly those that studied patients with acute bacterial diarrhoea, and we extracted data from the NTS subgroup. The studies did not assess the impact of antibiotics on severe diarrhoeal illness caused by Salmonella, as severe illness was an exclusion criteria in almost all of the studies. One of our included studies evaluated malnourished children, but no studies evaluated other immunocompromised people (people with AIDS, or other immunocompromising conditions). In one of the trials, immunocompromised people were specifically excluded. We therefore cannot answer the question as to the benefit or otherwise of antibiotic intervention in this group of patients. We could also not answer the question as to the benefit or otherwise of antibiotic administration in very young or very old people as most of the studies did not include these patients at all. In studies where they were included the numbers were very few and outcome assessments were not reported separately. We were therefore unable to perform a subgroup analysis on this category of patients.
One of the potential risks of intestinal salmonellosis in young infants is extraintestinal infection. No study reported on this outcome and this review is unable to provide information as regards the effects of antibiotics on this outcome in children. There was no study of the effect of a fluoroquinolone in infants and young children, partly because of safety concerns stemming from the observed effects on cartilage in animals.
A major concern regarding treatment with fluoroquinolones, and indeed all antibiotics, is the risk of emergence of resistance and outbreaks of infections due to resistant organisms, which could potentially cause serious extraintestinal infections in high risk groups.
Notably, we may have missed studies that assessed people with diarrhoea that included patients with Salmonella but did not refer to this group of patients in its title/abstract or MESH etc, but only in the full text or tables.
Quality of the evidence
Our review utilized evidence from RCTs. Some of the data that we have included in the meta-analysis are skewed and so our overall effect estimates must be interpreted with caution. Using the GRADE process to evaluate the quality of evidence from the trials, most of the evidence is very low quality.
Potential biases in the review process
We faced challenges in our data extraction, and assessment of the intervention effect on our pre-specified outcomes as a result of the generally poor quality of reporting of some of the outcomes in the trials, particularly with regard to the consistent reporting of continuous outcomes. Also, the included trials and some of the excluded trials could have contributed more to the review if the authors had performed subgroup analysis with respect to the isolated pathogens after stool culture. This would help to better elucidate the pattern of antimicrobial drug efficacy in the treatment of bacterial diarrhoea.
Agreements and disagreements with other studies or reviews
Two studies (Carlstedt 1990; Hatalin 1972), which were included in the previous version of the review (Sirinavin 2000), did not meet our inclusion criteria mainly because of methodological issues and because they did not include patients who were symptomatic for NTS gastroenteritis. The present review has utilized data from a new study that was not included in the earlier review (Chiu 1999).
This review is still affected by some of the methodological issues in the included trials in the earlier version of the review. However, the results of our review is in agreement with the previous review.
Only one new trial is now available that was not available at the time of the 1999 review. None of the identified trials have investigated invasive NTS disease. This highlights the need for more research into other aspects of NTS infection.
Implications for practice
We are unable to demonstrate a positive clinical effect of antibiotic therapy on the treatment of NTS diarrhoea in people with non-severe diarrhoea. Adverse drug reactions, although minimal, do occur with antibiotic treatment. Antibiotic administration, therefore, should not be routinely recommended. For patients with some underlying immunosuppressive disorder, or in patients who are very young or very old, current data are insufficient to make a conclusive statement as regards appropriate management.
Antibiotic therapy appears to result in early negative stool cultures, but higher rates of relapse afterwards.
Implications for research
We are unable to comment on the effects of antibiotic therapy on NTS intestinal infection in the high-risk groups for extraintestinal invasion (infants, elderly and immuno-compromised patients) and on severe diarrhoea. There is a need for further randomized, placebo controlled trials in these patients. These trials would have to be adequately powered to enable the detection of clinically meaningful effects and multicentre collaboration may be beneficial.
New antibiotics with potential for therapeutic usefulness in treatment of symptomatic Salmonella infection need to be investigated in the context of RCTs. One of the identified but excluded trials evaluated the use of rifaximin but could not be included in the review because the number of NTS patients was very small and included with other patients.
These trials can proceed to study all patients with bacterial diarrhoea but would perform subgroup analysis by the isolated pathogen. These trials need to include enough patients to be able to have statistical power and also need to study the patients long enough (for at least 8 to 10 weeks) to enable a clearer picture to be obtained as regards microbiological failure and detection of the same Salmonella serovar 1 to 2 months after treatment. Also these studies would need to continue to examine the stool cultures even after they become initially culture negative as this would enable it detect patients who relapse after treatment.
Ifeanyi Onwuezobe was awarded the Reviews for Africa Programme Fellowship by a grant from the Nuffield Commonwealth Programme through the Nuffield Foundation.
Chibuzo Odigwe was awarded the 2009 Aubrey Sheiham Public Health and Primary Care Scholarship of the UK Cochrane Centre.
We are grateful to Professor Michael Clarke, Drs Phil Wiffen, Sally Hopewell, Amy Drahota and Anne Eisinga of the UK Cochrane Centre for their help and advice. We also owe a great amount of gratitude to Professor Paul Garner, Reive Robb, Anne-Marie Stephani, Sarah Donegan, Alfred Musekiwa, Vittoria Lutje and Caroline Hercod of the Cochrane Infectious Diseases Group Editorial Base, Liverpool for help with searching and retrieval of some of the relevant papers and statistical advice. The editorial base of the Cochrane Infectious Diseases Group is funded by UKaid from the UK Government for the benefit of developing countries.
We also appreciate the very helpful guidance of Professor Martin Meremikwu, Drs Mical Paul and David Sinclair, Editors in the Cochrane Infectious Disease Group for their very constructive comments on the draft of the review. Dr Paul is the academic editor for this review.
We are also grateful to Dr Taryn Young of the South African Cochrane Center and Dr Emmanuel Effa of the Nigerian Branch of the South African Cochrane Centre for their help and advice. Special thanks also to Jini Hetherington, Patricia Atkinson and Sir Iain Chalmers for their support during the preparation of the review.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Last assessed as up-to-date: 6 August 2012.
Protocol first published: Issue 4, 2008
Review first published: Issue 3, 1998
Contributions of authors
Drs Ifeanyi Onwuezobe, Phillip Oshun and Chibuzo Odigwe wrote the protocol, applied inclusion criteria, assessed methodological quality, analysed the data and wrote the review.
Declarations of interest
We declare that we have no conflicts of interest.
Sources of support
- University of Uyo Teaching Hospital, Uyo, Akwa Ibom State, Nigeria.
- Institute of Tropical Disease Research & Prevention, University of Calabar Teaching Hospital, Calabar, Nigeria.
- Nuffield Commonwealth Foundation, UK.Reviews for Africa Programme Grant
- Department for International Development (DFID), UK.The Editorial Base of the Cochrane Infectious Disease Group is funded by the DfID.
- UK Cochrane Centre, Oxford, UK.Aubrey Sheiham Public Health & Primary Care Scholarship
Differences between protocol and review
The methods section has been revised as we allowed inclusion of studies that recruited patients with unspecified diarrhoea or gastroenteritis and extracted and analysed the data for the subgroup of patients with documented Salmonellosis.
Medical Subject Headings (MeSH)
Anti-Bacterial Agents [*therapeutic use]; Diarrhea [*drug therapy; microbiology]; Gastroenteritis [drug therapy]; Gastrointestinal Diseases [drug therapy]; Randomized Controlled Trials as Topic; Salmonella Infections [*drug therapy]; Salmonella paratyphi A; Salmonella typhi
MeSH check words
Adult; Child; Child, Preschool; Humans; Infant