Summary of findings
Description of the condition
Shigellosis is a bacterial infection of the colon that causes diarrhoea and can lead to death. Dysentery (frequent mucoid or bloody stools) when caused by Shigella is called Shigella dysentery. Of the estimated 164.7 million Shigella diarrhoeal episodes occurring globally every year, most occur in developing countries (99%) and mainly in children (69%) (WHO 2006). Of the 1.1 million deaths due to Shigella, 69% are in children aged less than five years (Kotloff 1999; WHO 2006).
Microbiology and mode of spread
Shigella dysenteriae, S. flexneri, S. sonnei, and S. boydii are the four species of small, Gram-negative, non-motile bacilli that cause shigellosis, and all but S. sonnei have more than one genetically distinct subtype (serotype) (von Seidlein 2006). The species distribution varies globally; for example, S. flexneri was reported to be most prevalent in India (58%, Dutta 2002) and Rwanda (68%, Bogaerts 1983), while S. sonnei was the most frequently detected species in Thailand (85%, von Seidlein 2006), Israel (48.8%, Mates 2000), and the USA (75%, Gupta 2004; Shiferaw 2004).
Shigellae are transmitted by the faeco-oral route, via direct person-to-person contact, and via food, water, and inanimate objects. Only a small number of ingested bacteria are required to produce illness. The disease is communicable as long as an infected person excretes the organism in the stool, which can extend up to four weeks from the onset of illness. Secondary attack rates, the number of exposed persons developing the disease within one to four days following exposure to the primary case (Park 2005), can be as high as 40% among household contacts (Sur 2004).
Shigellosis occurs predominantly in developing countries and is most common where overcrowding and poor sanitation exist. It occurs in densely populated areas and institutions where populations are in close contact with each other, such as day-care centres, cruise ships, institutions for people with mental or psychological problems, and military barracks (Shane 2003; Gupta 2004).
The clinical manifestation of shigellosis ranges from an asymptomatic illness to bacteraemia and sepsis. Symptoms include fever, diarrhoea and/or dysentery with abdominal cramps and ineffectual and painful straining at stool or in urinating (Niyogi 2005). Shigellosis may be associated with mild to life-threatening complications, such as rectal prolapse, arthralgia (painful joints), arthritis, intestinal perforation, and toxic mega colon (extreme inflammation and distension of the colon), central nervous disorders, convulsions, enteropathy (protein-losing disease of the intestines), electrolyte imbalance of salts, and sepsis (Sur 2004; WHO 2005b). About 3% of those infected with S. flexneri and who are genetically predisposed can develop Reiter's syndrome (pains in their joints, irritation of the eyes, and painful urination) that can lead to a difficult to treat chronic arthritis (CDC 2005). Haemolytic uraemic syndrome (a complication resulting in kidney failure, bleeding, and anaemia) and leukemoid reaction (blood findings resembling leukaemia) complicate infection due to S. dysenteriae type 1 and may be fatal (Sinha 1987). S. dysenteriae type 1 is the only Shigella species with chromosomal genes encoding the protein known as Shiga toxin (Thorpe 2001).
The clinical features of fever with blood and/or mucous diarrhoea associated with abdominal pain suggest that the aetiology of diarrhoea is Shigella. Routine microscopy of fresh stool is a simple screening test that is cheap, rapid, and easy to perform; and visualization of numerous poly-morphonucleocytes suggests a bacterial aetiology. Definite diagnosis of shigellosis can only be made by stool culture (WHO 2005a). However, Shigella species die rapidly in unfavourable environments and stool culture should ideally be supplemented by attempts to identify Shigella DNA using polymerase chain reaction (PCR) (von Seidlein 2006).
Clinical relapse can occur. This manifests as an initial clinical improvement or apparent cure with the treatment, followed by the recurrence of diarrhoea after the course of drug treatment is completed. In some instances people have sought the continued presence of Shigella in cultures of stool after the treatment, irrespective of apparent clinical recovery and have documented these as bacteriological failures (Martin 2000), indicative of the potential for relapse. Relapse is an important indicator of treatment failure, though it is clinically difficult to differentiate a relapse of infection with the same species or serotype of Shigella without additional testing for Shigella DNA using PCR analysis (von Seidlein 2006).
The case-fatality rate is estimated to be less than 1% among those with mild illness (WHO 2005a), which is usually self-limiting (CDC 2005), and those affected are usually treated as out-patients. However, case fatality is as high as 15% among patients with S. dysenteriae type 1 who require hospitalization; this rate is increased by delayed arrival and treatment with ineffective antibiotics. Infants, non-breast fed children, children recovering from measles, malnourished children, and adults older than 50 years have a more severe illness and a greater risk of death (WHO 2005a).
Shigella and HIV infection
Human immunodeficiency virus (HIV) infection may be an important risk factor for Shigella infection. Particularly in HIV-positive people, shigellosis is associated with extensive illness, including Shigella septicaemia, and increased health-care expenditures. The diagnosis of shigellosis in an otherwise healthy adult without obvious exposure risk for Shigella should prompt consideration of the possibility of HIV infection (Huebner 1993; Baer 1999).
Description of the intervention
The World Health Organization (WHO) recommends that all suspected cases of shigellosis based on clinical features be treated with effective antimicrobials (antibiotics). The choice of antimicrobial drug has changed over the years as resistance to antibiotics has occurred, with different patterns of resistance being reported around the world. The following antibiotics were used to treat Shigella dysentery:
- class: beta-lactams: ampicillin, amoxicillin, first and second generation cephalosporins (cefixime, ceftriaxone) and pivmecillinam;
- class: quinolones: nalidixic acid, ciprofloxacin, norfloxacin, ofloxacin;
- class: macrolides: azithromycin; others: sulphonamides, tetracycline, cotrimoxazole, and furazolidone.
The WHO now recommends that clinically diagnosed cases of Shigella dysentery be treated with ciprofloxacin as first line treatment, and pivmecillinam, ceftriaxone, or azithromycin as second line treatment and lists the others as ineffective (WHO 2005a). However, resistance to quinolones has also been observed since the late 1990s, and some authors have questioned the effectiveness of this class for Shigella (Datta 2003; Sarkar 2003; Sur 2003; Pazhani 2004; Talukder 2004).
Why it is important to do this review
When an effective antibiotic is given, clinical improvement is anticipated within 48 hours (WHO 2005a). This lessens the risk of serious complications and death, shortens the duration of symptoms, and hastens the elimination of Shigella and the subsequent spread of infection (WHO 2005a). Since the antibiotics used for treating shigellosis can have adverse effects ( Table 1; BNF 2007), some life-threatening, the clinician is faced with a dilemma in choosing an appropriate drug to treat shigellosis. This drug must be effective, locally available at affordable costs, be associated with minimum adverse effects and be sensitive to local Shigella species and strains. We undertook this review in the hope of identifying such a drug or group of drugs.
To evaluate the efficacy and safety of antibiotics for treating Shigella dysentery.
Criteria for considering studies for this review
Types of studies
Randomized controlled trials (RCTs).
Types of participants
Adults and children with clinical symptoms suggestive of Shigella dysentery. Both hospitalized and non-hospitalized participants were included.
Types of interventions
Antibiotics, irrespective of the dose or route of administration.
Other antibiotic of a different class (irrespective of the dose or route of administration), placebo, or no drug.
We included trials that used additional interventions if the interventions were used in all treatment arms.
Types of outcome measures
- Diarrhoea at follow up.
- Relapse, defined as the reappearance of diarrhoea associated with Shigella in the stool or dysentery during follow up.
- Fever at follow up: defined as body temperature above 37.0 ºC or 98.6 ºF.
- Time to cessation of fever.
- Time to cessation of diarrhoea.
- Time to cessation of blood in stools.
- Total number of stools per day.
- Bacteriological cure: defined as a negative stool culture at the end of a specified time period after treatment.
- Duration of hospital stay.
- Development of severe complications.
- Serious adverse events (i.e. those that are life-threatening or require hospitalization); those that lead to discontinuation of treatment; other types of adverse events.
Search methods for identification of studies
We identified all relevant trials regardless of language or publication status (published, unpublished, in press, and in progress).
We searched the following databases using the strategies and search terms set out in Table 2: the Cochrane Infectious Diseases Group Specialized Register; the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2008, issue 4); MEDLINE (1966 to June 2009); EMBASE (1974 to June 2009); and LILACS (1982 to June 2009). We also searched the metaRegister of Controlled Trials (mRCT) using 'shigell*' as the search term (June 2009).
Searching other resources
In Table 3 we list the conference proceedings searched for relevant abstracts, individual researchers working in this field contacted, organizations and pharmaceutical companies contacted to identify unpublished and ongoing trials, along with the dates when this was done. We also checked the reference lists of all studies identified by the above methods.
Data collection and analysis
Selection of studies
Two pairs of authors (PC and KVD, and SMJ and VS ) independently assessed the results of the literature search to determine whether the title or abstract of each trial cited was an RCT . We retrieved the full reports of all trials considered by one or both pairs of authors as potentially relevant as well as those that were unclear from scrutinizing the abstracts. Each pair used a standard eligibility form based on the inclusion and exclusion criteria to assess the trials. We resolved disagreements through discussion. If eligibility was uncertain due to unclear or inadequate information, we attempted to contact the trial authors for clarification. The reasons for excluding studies were noted in the 'Characteristics of excluded studies' table. Each trial report was scrutinized to ensure that multiple publications from the same trial are included only once, and all reports were linked to the original trial report in the reference list of included studies.
Data extraction and management
The pairs of authors independently extracted data from the trials using pre-tested data extraction forms. We extracted data on the inclusion and exclusion criteria for the participants, treatment/intervention given, total number randomized, number of participants in each group for all outcomes, drop-outs, and withdrawals and numbers experiencing each outcome. For every outcome, we extracted the number analysed and the number randomized in each treatment group to allow for the assessment of losses to follow up. Any disagreements about data extracted were resolved by referring to the trial report and by discussion. Where data were insufficient or missing, attempts were made to contact the trial authors.
For continuous outcomes, we extracted the arithmetic mean values, standard deviations, and the number of participants in whom the outcome was assessed in each of the two groups. We noted whether the numbers assessed in the trial were the number of participants that completed the trial or the number randomized. If medians were reported we extracted ranges, or interquartile ranges.
Assessment of risk of bias in included studies
The pairs of authors independently assessed the risk of bias in each included trial for the following six components: sequence generation, allocation concealment, blinding or masking, incomplete outcome data, selective outcome reporting, and other sources of bias. For each of these components, we assigned a judgment regarding the risk of bias as 'yes', 'no', or 'unclear' (Higgins 2008). We recorded follow up to be adequate if more than 90% of the randomized participants were included in the final analysis, inadequate if less than or equal to 90%, or unclear if this information was not available from the report or trial authors. We recorded these assessments in the standard table in RevMan 5 (Review Manager 2008), and summarized them in 'Risk of bias' tables and a graph (Figure 1; Figure 2). We used these assessments to perform a sensitivity analysis based on methodological quality when appropriate. We attempted to contact the trial authors for clarification when methodological details were unclear. We resolved differences by discussion and by contacting an Editor with the Cochrane Infectious Diseases Review Group.
|Figure 1. Methodological quality summary: review authors' judgments about each methodological quality item for each included study.|
|Figure 2. Methodological quality graph: review authors' judgments about each methodological quality item presented as percentages across all included studies.|
Measures of treatment effect
The measures of treatment effect used were risk ratio (RR) for dichotomous outcomes and mean difference for continuous outcomes with their 95% confidence intervals (CIs).
Dealing with missing data
Where possible, we extracted data to allow an intention-to-treat analysis in which all randomized participants were analysed in the groups to which they were originally assigned. If there was discrepancy in the number randomized and the numbers analysed in each treatment group, we calculated the percentage loss to follow up in each group and reported this information. For dichotomous outcomes, we recorded the number of participants experiencing the event and the number analysed in each treatment group. We assigned those lost to follow up the worse outcome, except for the outcome of death, since it would be unreasonable to assume that all those who were lost to follow up died.
Assessment of heterogeneity
We determined the presence of statistical heterogeneity among the same interventions by examining the forest plot and by performing the Chi
Assessment of reporting biases
All studies were assessed for adequacy of reporting of data for pre-stated outcomes and for selective reporting of outcomes. We noted judgements based on the risk of selective reporting in the 'Risk of bias' table for each study in the 'Characteristics of included studies' table.
Had there been sufficient trials we would have evaluated asymmetry in the funnel plot as an indication of publication bias.
The first two authors entered data into Review Manager 2008 using double-data entry. PC synthesized the data, which the co-authors checked. All results are presented with 95% CIs. The main comparisons were between any antibiotic drug and placebo, and any antibiotic drug and another antibiotic drug of a different class.
We synthesized dichotomous data using pooled and weighted RRs. Continuous data summarized by arithmetic means and standard deviations were combined using the weighted mean differences.
We used the fixed-effect model to synthesize data if heterogeneity was not substantial. When there was substantial heterogeneity and this could not be explained by subgroup analysis, we synthesized data using the random-effects model and recommended a cautious interpretation of the pooled result.
Subgroup analysis and investigation of heterogeneity
When there was significant statistical heterogeneity, we explored the possible sources using the following subgroup analyses: participant age (adults versus children) and percentage of participants with confirmed Shigella infection.
We performed sensitivity analyses for primary outcomes to assess the robustness of the meta-analysis among the same interventions by calculating the results using all trials and then excluding trials of a lower methodological quality (i.e. trials with inadequate generation of allocation sequence and allocation concealment, trials that were not double blind, and trials where less than or equal to 90% of randomized participants were analysed).
Description of studies
Results of the search
Out of 265 studies retrieved by the search, we obtained full texts of 123 studies. The rest were excluded as they were neither RCTs nor studies of antibiotic therapy for Shigella. Of the 123 studies, 16 parallel group, individually randomized trials met inclusion criteria (see 'Characteristics of included studies') and are summarized below. The reasons for excluding the other 106 trials are recorded in the 'Characteristics of excluded studies' table. One study awaits assessment (Carbo 1981).
Location, setting and length of follow up
Seven trials were conducted in Bangladesh, all at the International Centre for Diarrhoeal Disease Research (ICDDR,B). Two trials were from the United States of America (Haltalin 1973; Nelson 1976a) and one each from the following countries: India (Dutta 1995), Sri Lanka (Bibile 1961), Peru (Gotuzzo 1989), Israel (Leibovitz 2000), Guatemala (Prado 1993), Mexico (Rodriguez 1989), and Kenya (Shanks 1999). Twelve trials were carried out in hospitalized patients, three in out-patients and one did not mention the setting. The trials used different lengths of follow up: eight trials were for six days, three trials for five days, two trials for 14 days and one trial each for seven days, 10 days, and six months.
The trials included a total of 1748 participants. All trials but one (Haltalin 1973) were randomized based on clinical symptoms of dysentery and prior to bacteriological confirmation. People with neither blood nor mucus in stools were excluded. Haltalin 1973 randomized participants after a presumptive confirmation of Shigella by immunofluorescence study of rectal swabs. Dutta 1995 did not seek microbiological confirmation for Shigella by culture of stool samples or rectal swabs. In the remaining trials, only the data from participants with microbiologically confirmed Shigella were reported and thus only those data were included in the analyses. Ten trials were carried out only in children, five in adults, and one included both. Among the 10 trials in children, only one (Dutta 1995) included malnourished children (11 of 72) but did not provide data on them separately. Two trials excluded children with malnutrition and the remaining seven trials did not provide such information. None of the trials reported the HIV status of participants. The other inclusion criteria were fairly similar across all trials.
Two trials (Kabir 1986; Rodriguez 1989) compared antibiotics and placebo or no drug. Both were three-armed trials. Rodriguez 1989 compared furazolidone, cotrimoxazole, and no drug. Kabir 1986 compared ceftriaxone, ampicillin, and a placebo. Six trials compared flouroquinolones and beta-lactams (Alam 1994, pivmecillinam and nalidixic acid; Bennish 1990, ciprofloxacin and ampicillin; Haltalin 1973, nalidixic acid and ampicillin; Leibovitz 2000, ciprofloxacin and ceftriaxone; Salam 1988, nalidixic acid and ampicillin; Salam 1998, ciprofloxacin and pivmecillinam). Two trials compared flouroquinolones and macrolides (Khan 1997a; Shanks 1999), both compared azithromycin and ciprofloxacin). Two trials compared cotrimoxazole and beta-lactams (Prado 1993, pivmecillinam and cotrimoxazole; Nelson 1976a, cotrimoxazole and ampicillin). Gotuzzo 1989 compared cotrimoxazole and flouroquinolones (norfloxacin). Dutta 1995 compared furazolidone and nalidixic acid. Islam 1994 compared oral gentamicin and nalidixic acid. Bibile 1961 was a four-armed trial: the first three had different types of sulphonamides: sulphamidine, sulphamethoxypyridazine, 'Streptotriad' and the fourth arm was tetracycline. Each tablet of Streptotriad contained streptomycin sulphate, sulphamerazine, sulphadiazine and sulphathiazole. This arm was not included in analysis (sulphonamide versus tetracycline) since it contained a non-sulphonamide drug, streptomycin.
This review had two primary efficacy outcomes. The first primary outcome, diarrhoea on follow up, was reported by all but three trials (Kabir 1986; Gotuzzo 1989; Islam 1994); the duration of follow up was five days in 10/13 trials. The second primary outcome, relapse, was reported by four trials (Haltalin 1973; Salam 1998; Shanks 1999; Leibovitz 2000); the duration of follow up for this outcome ranged from 10 to 20 days. Among the secondary outcomes, fever at follow up was reported by four trials, time to cessation of fever was reported by five trials, time to cessation of diarrhoea was reported by six trials, time to cessation of blood in stools was reported by three trials, bacteriological cure or failure was reported by 11 trials, and development of severe complications was reported by only one trial. Duration of hospital stay was not an outcome measured by any of the trials. One trial (Kabir 1986) reported the mean number of stools per day in a graph that did not permit extraction of data for analysis. Adverse events were reported by all but four trials (Haltalin 1973; Alam 1994; Islam 1994; Dutta 1995). Only Leibovitz 2000 reported serious adverse events related to antibiotic therapy leading to hospitalization. None of the trials reported any deaths.
We excluded 107 studies for the following reasons. Twenty-nine studies were not RCTs. In 59 studies the inclusion criteria for the participants was not dysentery. Eighteen studies compared antibiotics of the same class, which we decided should be the subject of a separate review. One trial was excluded as the interventions were not antibiotics (Raqib 2008). Carbo 1981 awaits assessment as it provided no data on the numbers allotted to interventions and we are awaiting a reply from the authors.
Risk of bias in included studies
See Figure 1 for a summary of the 'risk of bias' in each included study and Figure 2 for a summary graph of methodological quality expressed as percentages across included trials. The risk of bias for each study is summarized additionally in 'Characteristics of included studies'.
Among the included studies, 81% (13/16) had low risk of bias in the generation of the allocation sequence. Of these, four trials (Bibile 1961; Prado 1993; Salam 1998; Leibovitz 2000) used random number lists. The remaining trials (Nelson 1976a; Kabir 1986; Salam 1988; Gotuzzo 1989; Bennish 1990; Alam 1994; Islam 1994; Dutta 1995; Khan 1997a) used block randomization techniques. However, only 9/16 (56%) of the studies clearly reported adequate concealment of allocation (Salam 1988; Bennish 1990; Prado 1993; Alam 1994; Islam 1994; Dutta 1995; Khan 1997a; Salam 1998; Leibovitz 2000).
Eleven trials (69%) had low risk of bias for the component of blinding. Salam 1988, Khan 1997a, Salam 1998, Shanks 1999 and Leibovitz 2000 had blinded the participant, the provider, and the outcome assessor. Kabir 1986, Bennish 1990, Prado 1993, Alam 1994 and Islam 1994 had blinded the participant and the provider. Dutta 1995 had only the outcome assessor blinded. Bibile 1961, Haltalin 1973, Nelson 1976a, Gotuzzo 1989 and Rodriguez 1989 were open trials.
Incomplete outcome data
Only 25% (4/16) trials (Bibile 1961; Haltalin 1973; Nelson 1976a; Kabir 1986) were judged to have adequately addressed incomplete outcome data. The remaining 12 trials did not adequately address incomplete outcome data because they excluded participants from data analysis after randomization as their stool cultures were later negative for Shigella. This is a serious methodological flaw (see 'Potential biases in the review process').
All the studies were free of selective reporting.
Other potential sources of bias
More than 90% (15/16) of the studies had no other potential sources of bias. One study (Rodriguez 1989) had a significant baseline imbalance as the participants in one of the study arms had fewer days of diarrhoea than the other arms.
Effects of interventions
See: Summary of findings for the main comparison Antibiotic versus no drug or placebo for Shigella dysentery; Summary of findings 2 Fluoroquinolones versus beta-lactams for Shigella dysentery; Summary of findings 3 Fluoroquinolones versus macrolides for Shigella dysentery; Summary of findings 4 Cotrimoxazole versus beta-lactams for Shigella dysentery; Summary of findings 5 Cotrimoxazole versus fluoroquinolones (norfloxacin) for Shigella dysentery; Summary of findings 6 Cotrimoxazole versus furazolidone for Shigella dysentery; Summary of findings 7 Oral gentamicin versus nalidixic acid for Shigella dysentery; Summary of findings 8 Sulphonamides versus tetracycline for Shigella dysentery
We intended to prepare separate meta-analyses for trials of: (1) an antibiotic drug versus another antibiotic drug belonging to the same or different drug class; (2) antibiotic drugs grouped by drug class versus other antibiotic drugs belonging to a different drug class; and (3) monotherapy with any antibiotic drug versus combination drug therapy with two or more different drugs given together or sequentially. However, we were only able to synthesize data from trials comparing single antibiotics of different classes and of antibiotics grouped by class. Comparisons of antibiotics within the same class were deferred to a subsequent review and thus 17 potential trials of this comparison were excluded from this review and are listed as such in the 'Characteristics of excluded studies'. We did not identify trials of an antibiotic drug versus combination drug therapy with two or more different drugs given together or sequentially.
We present trial results grouped as eight sets of comparisons.
1. Versus no drug or placebo (two trials)
Diarrhoea on follow up (primary outcome):
Rodriguez 1989 compared both oral furazolidone and cotrimoxazole with no treatment. Fewer patients in the antibiotic group had diarrhoea at follow up (for furazolidone, RR 0.21, 95% CI 0.09 to 0.48, 73 participants; and for cotrimoxazole versus no treatment, RR 0.30, 95% CI 0.15 to 0.59; 76 participants, Analysis 1.1).
Kabir 1986 compared intravenous ceftriaxone (n=64) and intravenous ampicillin (n=60) with placebo (n=30). There was no difference detected in time to diarrhoea resolution ( Analysis 1.3), fever resolution ( Analysis 1.2), and time to resolution of blood in the stools ( Analysis 1.4), or adverse events ( Analysis 1.5).
2. Fluoroquinolones versus beta-lactams (six trials)
Diarrhoea on follow up (primary outcome):
Six trials measured this, and the comparative effects varied considerably between the trials, with no obvious trend (686 participants, six trials, Analysis 2.1; Haltalin 1973; Salam 1988; Bennish 1990; Alam 1994; Salam 1998; Leibovitz 2000). This variability was still present after exclusion of trials with a higher risk of bias (Haltalin 1973; Bennish 1990; Alam 1994; Salam 1988). Most of the trials were in children; one trial was in adults (Bennish 1990).
In trials where 90% or more of included patients were confirmed with Shigella, beta-lactams were more effective than fluoroquinolones (RR 4.68, 95% CI 1.74 to 12.59; 257 children, two trials, ( Analysis 2.1). (Haltalin 1973; Leibovitz 2000); in the four trials with less than 90% confirmed Shigella positive patients the results showed no obvious pattern ( Analysis 2.1). (Salam 1988; Bennish 1990; Alam 1994; Salam 1998).
Fever at follow up:
Heterogenous data from two trials (Alam 1994; Salam 1998) showed no difference between the groups (191 participants, Analysis 2.2). Subgroup analysis was not done as both trials were done in children and had less than 90% of participants with Shigella in stool culture.
Pooled heterogenous data from five trials (Haltalin 1973; Salam 1988; Bennish 1990; Alam 1994; Salam 1998) showed no difference between the two groups for this outcome (450 participants, Analysis 2.4). However on subgroup analysis based on participant's age, the single study done on adults (Bennish 1990) showed that fluoroquinolones were better than beta-lactams in producing bacteriological cures (RR 0.28; 95% CI 0.08 to 0.95; 127 participants, Analysis 2.4). Even though the data from the children's subgroup (Haltalin 1973; Salam 1988; Alam 1994; Salam 1998) were homogenous, there was no difference between the two groups (223 participants, Analysis 2.4). The heterogeneity persisted on subgroup analysis based on number of participants with proven Shigella included in analysis.
Development of severe complications:
Data from two trials (Haltalin 1973; Salam 1988) showed no difference between two groups for this outcome (90 participants, Analysis 2.5). Though formal tests did not reveal significant heterogeneity, the differences in size and direction of treatment effect for the two trials is important to consider in interpreting this result.
For serious adverse events, Leibovitz 2000 showed no difference between the two groups ( Analysis 2.6, n=221); Bennish 1990 did not detect a difference in adverse events leading to discontinuation of treatment (127 participants, Analysis 2.7); for other adverse events, no difference was detected in four trials reporting this ( Analysis 2.8). ( Salam 1988; Bennish 1990; Salam 1998; Leibovitz 2000).
(See ' Summary of findings 2').
3. Fluoroquinolones versus macrolides (two trials)
Diarrhoea on follow up (primary outcomes):
Data from two trials (Khan 1997a; Shanks 1999) showed no difference between the two groups (189 participants, Analysis 3.1). Heterogeneity could not be assessed since the results from Shanks 1999 were not estimable (no patients had diarrhoea on follow up in both arms) and hence neither subgroup analysis nor sensitivity analysis was done.
Shanks 1999 reported on relapse but the results were not estimable as no patients had experienced relapse.
Fever at follow up:
Time to cessation of blood in stool:
(See ' Summary of findings 3').
4. Cotrimoxazole versus beta-lactams (two trials)
Diarrhoea on follow up (primary outcome):
Homogenous data from two trials (Nelson 1976a; Prado 1993) did not show any difference between the two groups (89 participants, Analysis 4.1). Exclusion of the poorer quality trial (Nelson 1976a) did not affect the results in sensitivity analysis.
Time to cessation of diarrhoea:
Time to cessation of fever:
Time to cessation of blood in stools:
(See 'Summary of findings table 4').
5. Cotrimoxazole versus fluoroquinolones (one trial)
(See ' Summary of findings 5').
6. Cotrimoxazole versus furazolidone (one trial)
Diarrhoea on follow up (primary outcome):
(See ' Summary of findings 6').
7. Oral gentamicin versus nalidixic acid (one trial)
Diarrhoea on follow up (primary outcome):
Fever at follow up:
Islam 1994 reported this outcome and found nalidixic acid more effective than oral gentamicin in reducing the number patients with fever on follow up (RR 2.37, 95% CI 1.11 to 5.07; 79 participants, Analysis 7.2). While both the antibiotics were effective against Shigella in vitro, nalidixic acid was more effective in vivo due to better absorption when taken orally.
(See 'Summary of findings table 7')
8. Sulphonamides versus tetracyclines (one trial)
Diarrhoea on follow up (primary outcome):
(See ' Summary of findings 8').
Summary of main results
This review identified 16 trials conducted over a span of four decades that randomized 1748 participants to evaluate the safety and efficacy of antibiotics in the treatment of Shigella dysentery. Most trials were at risk of bias due to limitations in reporting details of randomization or allocation concealment or blinding, but the most common source of bias occurred due to failure to report outcome details for participants who were randomized but in whom Shigella could not be isolated from stool culture.
In this review we focused on trials done with antibiotics belonging to different classes compared against placebo or no treatment or to each other. We found limited evidence to support the use of antibiotics in children and adults with Shigella dysentery compared to no treatment or placebo. One trial reported that antibiotics are effective in reducing the proportion of those with diarrhoea but it did not report on relapse. Another trial suggested that antibiotics were effective in reducing the duration of fever though they did not reduce the time to cessation of diarrhoea or bloody stool.
We did not find robust evidence to suggest that antibiotics of a particular class were better than those belonging to a different class. However, there were limited data from a subgroup of studies to suggest that a fluoroquinolone (ciprofloxacin) was more effective than a beta-lactam (ampicillin) in reducing diarrhoea among adults and that beta-lactams were more effective than fluoroquinolones in reducing diarrhoea among children with proven Shigella dysentery. Oral gentamicin was also reported to be inferior to nalidixic acid in achieving bacteriological cure and reducing fever in one small trial. The trials in this review report that at various periods of time different antibiotics have been effective against isolates of Shigella dysentery ( Table 4) in different parts of the world. They are: ampicillin, cotrimoxazole, nalidixic acid, fluoroquinolones like ciprofloxacin, pivmecillinam, ceftriaxone, and azithromycin. However oral gentamicin was relatively ineffective, due to poor absorption when given orally, compared to nalidixic acid and therefore is not recommended. There was insufficient evidence to comment on the use of tetracyclines, sulphonamides, and chloramphenicol.
There is also insufficient evidence to indicate that any antibiotic class prevents relapse of Shigella dysentery.
None of the antibiotics studied in the trials were associated with major adverse events that were drug related.
Overall completeness and applicability of evidence
With respect to the review's objectives, this review found limited evidence that antibiotics reduce diarrhoea and the duration of fever compared to no antibiotic. However, we are unable to recommend an antibiotic or an antibiotic class for the treatment of Shigella dysentery. The studies identified could not sufficiently address relapse. All the antibiotics studied in this review were safe.
The studies addressed both adults and children. However, populations at risk for complicated Shigella dysentery, such as HIV infected populations and malnourished children, were not included (or adequately represented) in the trials we identified.
In current practice, antibiotics are recommended and used in the treatment of Shigella dysentery. The conclusions of this review confirm these recommendations and current practice. However this review is unable to recommend a specific antibiotic or antibiotic group as universally effective for the treatment of Shigella dysentery.
Even though mild forms of Shigella dysentery are said to be self-limiting, this review is unable to comment on the need for antibiotics in this group since the included trials did not grade patients with respect to the severity of illness.
This review did not include studies using drugs belonging to similar antibiotic classes. Another review is needed to study differences between antibiotics belonging to the same class and also between different antibiotic dosing schedules, and short-course versus longer-course therapy of the same antibiotic.
Quality of the evidence
The body of evidence identified does not allow a robust conclusion regarding the objectives of the review or strong recommendations regarding the choice of preferred antibiotics. Of the 16 trials (1748 participants) included in the review, most had methodological limitations including inadequate reporting of the generation of allocation sequence, inadequate allocation concealment, and lack of blinding. Many trials removed participants after randomization since they did not grow Shigella in their stool culture and had not reported their outcome. This is a serious methodological error. Most trials were thus graded of low or very low quality and further research may change the estimates of efficacy and our confidence in these estimates.
Potential biases in the review process
We selected trials that compared the efficacy and safety of antibiotics of different classes only and deferred inclusion of trials evaluating antibiotics of the same drug class to an update or a separate review. Seventeen trials were excluded on the basis of this. This might have biased the results and conclusions of this review. We also did not include comparisons of different doses, routes of administration, or duration of treatment of the same antibiotic in Shigella dysentery.
We selected trials which included participants with clinical evidence of dysentery. However, Shigella infection can also present as diarrhoea in up to three-quarters of infections, particularly in Asian countries (von Seidlein 2006). Excluding such patients in trials of antibiotics in Shigella and excluding trials using a broader definition than that used in this review could have biased the evidence presented. Many trials in this review also excluded participants randomized to receive antibiotics if their stool did not grow Shigella isolates. However, Shigella species and strains are highly sensitive to inhospitable environments and failure to grow Shigella in culture does not rule out Shigella infection (von Seidlein 2006). None of the included trials utilized alternative or additional, sensitive, diagnostic techniques such as identification of Shigella DNA using real-time PCR. Exclusion of data from such participants in these trials, and exclusion of more stringent inclusion criteria for the diagnosis of Shigella dysentery in this review is likely to have introduced reporting and selection biases, respectively.
Agreements and disagreements with other studies or reviews
The overall results of this review suggest that most of the antibiotics used were effective. However, only 10 of the 16 included trials reported the proportion of participants that were sensitive to the antibiotics used. The outcomes in these trials correlated with the sensitivity patterns of the antibiotics used.
The WHO recommended nalidixic acid as the first line treatment for Shigella dysentery until 2004 when complete resistance to nalidixic acid in large parts of China and Bangladesh led to recommendations by the WHO to avoid using nalidixic acid altogether in Shigella dysentery (Legros 2004; WHO 2005a). However, nalidixic acid continues to be a potential option in parts of the world where resistance to this drug is not, as yet, a widespread problem, such as the Dakar region of the Senegal, where resistance to ampicillin, chloramphenicol, tetracycline, and cotrimoxazole are common (Sire 2008). However, widespread use of nalidixic acid may increase resistance to ciprofloxacin due to cross-resistance of some strains of Shigella and thus has limited utility (WHO 2005a). The WHO recommends the use of ciprofloxacin as the first line antibiotic in suspected Shigella dysentery but also suggests that this choice should be based on sensitivity patterns of Shigella strains recently isolated in the area (WHO 2005a). Temporal and geographical shifts in Shigella strains are reported in parts of the world (von Seidlein 2006) and regular surveillance and ascertainment of antimicrobial sensitivity to local and regional strains is necessary to determine the choice of antibiotic to be used as first line in Shigella dysentery. Emerging drug resistance to ciprofloxacin and second line drugs such as pivmecillinam, ceftriaxone, and azithromycin is increasingly being reported in many parts of the world, as is multiple-drug resistance (Kosek 2008; Kuo 2008; Pazhani 2008). The results of this review provides systematically ascertained evidence that the most commonly used antibiotics are potentially effective against Shigella dysentery, provided the local species and strains of Shigella are susceptible. Regular, periodic antibiotic-susceptibility testing of isolates is required to guide local empiric therapy for Shigella dysentery.
Implications for practice
We recommend the use of antibiotics for moderate to severe Shigella dysentery. The choice of antibiotic to use as first line against Shigella dysentery should be governed by periodically updated local antibiotic sensitivity patterns of Shigella isolates. Other supportive and preventive measures recommended by the WHO (WHO 2005a; WHO 2005b) should also be instituted along with antibiotics (eg health education and handwashing).
Implications for research
Randomized controlled trials which adhere to the CONSORT guidelines (CONSORT 2008) are required to address many of the issues such as the need for antibiotics in mild Shigella dysentery, the class or classes of antibiotics best suited against Shigella in populations at risk of high case-fatality such as malnourished children, older adults, patients presenting with serious complications due to shigellosis, and HIV infected individuals.
Trials should stratify participants according to severity of clinical presentation and report the effects of antibiotics separately for each group. Trials must report outcomes for all randomized participants including those with confirmed Shigella and those with negative culture. Antibiotic sensitivity patterns should also be studied and reported. Data regarding outcomes presented in graphs and pictures also need to be expressed in numbers. See Table 5 for the suggested features of a future trial.
We acknowledge the support of Prathap Tharyan, Katherine Abba, Paul Garner, Sara Bhattacharji, Gagandeep Kang, and Thambu David Sudarsanam, at various stages of this review. This protocol and review are the product of workshops conducted by the South Asian Cochrane Network & Centre that were partly funded by the Effective Health Care Research Programme Consortium (with funds from the Department for International Development (DFID), UK). The views expressed are not necessarily those of DFID.
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: 17 July 2009.
Protocol first published: Issue 4, 2007
Review first published: Issue 4, 2009
Contributions of authors
PC conceived the review and drafted the protocol. KVD, SMJ, and SV helped develop the protocol. Two teams of authors (PC and KVD & SMJ and SV) independently selected trials, assessed quality, extracted and entered data. All authors analysed and interpreted results and wrote the final review.
Declarations of interest
Sources of support
- Low Cost Effective Care Unit, Christian Medical College, Vellore, India.
- South Asian Cochrane Network & Centre, Vellore, India.
- Department for International Development (DFID), UK.
- Indian Council of Medical Research, India.For support and funding for the Prof. BV Moses and ICMR Advanced Centre for Research & Training in Evidence-Informed Healthcare
Differences between protocol and review
We intended to analyse combinations of an antibiotic drug versus another antibiotic drug of the same class or different drug classes. Comparisons of antibiotics within the same class were deferred to a subsequent review and thus 17 potential trials of this comparison were excluded from this review. The protocol was developed using Review Manager (RevMan) 4.2 and the review using RevMan 5 (Review Manager 2008). We intended to assess methodological quality of included studies using the methods described in Juni 2001. However, since the introduction of RevMan 5 (Review Manager 2008), a more detailed assessment of the risk of bias in included trials was undertaken, reported in 'Risk of bias' tables for each trial and graphically summarized in Figure 1 and Figure 2. We also used the GRADE profiler, version 3.2 (GRADE 2004) to create 'Summary of findings' tables for the primary outcomes in the review.
Had continuous data been summarised using geometric means, we would have combined them on the log scale using the generic inverse variance method and reported them on the natural scale.
Had outcomes been reported both at baseline and at a follow up or at trial endpoints, we would have extracted both the mean change from baseline and the standard deviation of this mean for each treatment group. We would also have extracted the means and standard deviation at baseline and follow up in each treatment group. If the data had been reported using geometric means, we would have recorded this information and extracted a standard deviation on the log scale.
Had count data been reported in trials, we intended to extract the total number of events in each group and the total amount of person-time at risk in each group. We also intended to record the total number of participants in each group. If this information was not available, we would have extracted alternative summary statistics such as rate ratios and confidence intervals if available. Again, if count data were presented as dichotomous outcomes, we would have extracted the number of participants in each intervention group and the number of participants in each intervention group who experienced at least one event. If count data were presented as continuous outcomes or as a time-to-event outcomes, we would have extracted the same information as outlined for continuous and time-to-event outcomes. Count data would have been compared using rate ratios when the total number of events in each group and the total amount of person-time at risk in each group are available, or by relative risks or weighted mean difference had the data been presented in dichotomous or continuous forms respectively. Hazard ratios from survival data would have been combined on the log scale using the inverse variance method and presented on the natural scale.
Had time-to-event outcomes been reported, we would have extracted the estimates of the log hazard ratio and its standard error. If standard errors were not available we would have extracted alternative statistics such as CIs or P values.
Medical Subject Headings (MeSH)
*Shigella; Anti-Bacterial Agents [*therapeutic use]; Diarrhea [drug therapy]; Dysentery, Bacillary [*drug therapy]; Furazolidone [therapeutic use]; Randomized Controlled Trials as Topic; Trimethoprim-Sulfamethoxazole Combination [therapeutic use]
MeSH check words