Cholera is an acute infection that causes sudden onset of profuse watery diarrhoea, and up to 40% of patients die if untreated. Cholera was a major cause of death in many countries in the past, although epidemics are now less common. Nevertheless, cholera remains an important cause of death in developing countries. In 2005, there were a total of 131,943 reported cases of cholera throughout the world, including 2272 deaths (WHO 2006a), and it is known that there were many more cases that were not reported. Ninety-five per cent of reported cases were in Africa. Cholera can lead to serious outbreaks: in 2005, the World Health Organization (WHO) confirmed 49 different outbreaks in 36 countries (WHO 2006a).
Cholera is caused by the Gram-negative bacillus Vibrio cholerae. There are over a hundred serological groups of V. cholerae, each with varying potential to cause disease. Until recently only one of these (V. cholerae 01) caused epidemic cholera. In 1992 to 1993, an epidemic of cholera originating in the Indian subcontinent was found to be caused by V. cholerae 0139, also called 0139 Bengal. Cholera strains are also classified by their biotype (Classical or El Tor), and within the biotype, the serotype (Ogawa or Inaba). Serotype differences are based on differences in structure of the lipopolysaccharide membrane. The various serological groups are important as each vaccine component tends to be specific to particular groups of V. cholerae.
Transmission of V. cholerae occurs predominantly when people ingest faecal contaminated water or food. The disease spreads rapidly where there is poverty, poor hygiene, and lack of sanitation. Waterborne spread can be responsible for devastating epidemics such as that which occurred due to El Tor cholera in the refugee camps of Goma, Zaire in July 1994. This resulted in 70,000 cases and 12,000 deaths (Sánchez 1997).
V. cholerae colonize the gut using small hair like structures ("pili") that attach to the small bowel. High stomach pH and blood group O appear to make colonization more likely. The attached bacteria then release a soluble toxin, which results in the symptoms of the disease. This toxin is composed of two subunits, A and B. The A subunit stimulates cellular mechanisms in the bowel cells that disrupt sodium transport. The net result is a high sodium chloride (salt) concentration in the gut lumen, which holds on to water by osmotic forces, leading to profuse watery diarrhoea, severe dehydration, and eventually death. The B subunit of cholera toxin does not cause toxic effects but does stimulate an immune response from the host. Colonization can be inhibited by specific antibodies which are generated after infection with V. cholerae.
Intravenous rehydration therapy can be very effective in treatment of cholera. However, health services in cholera endemic or epidemic areas often do not have sufficient capabilities for such treatment. Improving hygienic practices in areas of poverty and limited water supply can also be problematic. This has led to attempts to prevent cholera by vaccination. The first vaccine effectively used against cholera was probably that of Ferran, who in 1884 apparently successfully controlled an epidemic in Spain. A vaccine was also produced by the Pasteur Institute in the 1920s.
Widespread use of cholera vaccines began in the 1960s when there was a series of large trials in what was then known as East Pakistan (now Bangladesh), India, and the Philippines. Most of the vaccines used in these trials were composed of whole V. cholerae serogroup 01 cells, usually a mixture of biotypes and serotypes, which were killed by either formalin, phenol, or heat. There were also trials of cholera toxoid vaccines in the 1970s. The killed whole cell vaccines, which were subsequently licensed, are injected and usually given in one or two doses.
Injected (parenteral) whole cell vaccines grew in popularity until the 1970s when they went out of favour (Bhadra 1994) on the grounds that efficacy was thought to be low and short-lived, high titres of serum vibriocidal antibodies were thought not to provide sufficient intestinal immunity to prevent infection, and they were said to have a high rate of adverse effects. The advent of oral rehydration therapy, considered a highly effective treatment, was a major advance in reducing cholera morbidity, and led to a shift in interest away from injected vaccines. Even when injected cholera vaccines were in relatively widespread use in the early 1970s, it was never determined whether an individual's protection was likely to interrupt transmission to others in the community, or how important enteral or parenteral immunity is in bacterial shedding.
Recent cholera epidemics have shown that there still a requirement for an effective vaccine against this major disease (Sánchez 1997; Calain 2004; WHO 2006b). Oral vaccines have been under development since the 1980s, stimulated by the increasing recognition of the importance of stimulating local intestinal immunity in the prevention of the disease. Both killed and live oral vaccines are now licensed, but the injected vaccine is no longer used.
The original version of this review included both injected and oral cholera vaccines (Graves 2001), but this is now superseded and withdrawn. The current review assesses the results of trials with killed parenteral (injected) vaccines only. A separate Cochrane Review describes trials with oral cholera vaccines (Abba (in progress)).
To evaluate killed whole cell cholera vaccines and other inactive subunit vaccines (administered by injection) for preventing cases of cholera and preventing death, and to evaluate the adverse effects associated with the vaccination.
Criteria for considering studies for this review
Types of studies
Randomized or quasi-randomized controlled trials.
Exception: Phase 1 trials, reporting only adverse effects, for vaccines that never reached efficacy trials.
Types of participants
Well adults or children irrespective of immune status or special risk category.
Types of interventions
Killed whole cell cholera vaccines or other inactive subunit vaccines administered by injection
Placebo, control vaccines, or no intervention.
Types of outcome measures
- Cholera cases, as defined by each trial (usually diarrhoea more than three times in 24 hours with bacteriological confirmation of V. cholerae).
- All-cause deaths.
- Cholera deaths.
- Number and seriousness of adverse effects (classified as local and systemic).
- Systemic adverse effects include cases of malaise, nausea, fever, arthralgias, rash, headache and more generalized and serious signs.
- Local adverse effects include induration, soreness, and redness at the site of inoculation.
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 and strategy described in Appendix 1: Cochrane Infectious Diseases Group Specialized Register (1 September 2008); Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2008, Issue 3); MEDLINE (1966 to 1 September 2008); EMBASE (1974 to 1 September 2008); and LILACS (1982 to 1 September 2008).
Data collection and analysis
Selection of studies
Four authors (VD, TJ, PG, and JD) read all trials retrieved in the search and applied the inclusion criteria to determine eligibility.
Data extraction and management
PG and JD independently extracted and double-checked the following data: characteristics of participants (number, age, gender, ethnic group, risk category, and previous immunization status, if known); characteristics of interventions (vaccine type, placebo type, dose, immunization schedule, and length of follow up (in months); outcome measures; and trial date, location, sponsor, and publication status. All disagreements in the data extraction were resolved by discussion.
Adverse effect data were extracted individually for each adverse effect where possible. For trials where adverse effects were reported for more than one dose, the average of the number of people reporting each adverse effect for each dose was recorded. Where trials reported the occurrence of adverse effects over time following a single dose, the effects occurring in the first time period (typically 24 hours) were recorded if the total number of people reporting each effect in the complete follow-up period was not given.
We extracted incidence of cholera cases and death over particular time periods of follow up (eg first year following vaccination, second year etc) to determine the duration of protection.
Assessment of risk of bias in included studies
PG and JD independently assessed each trial's method of treatment allocation (random, quasi-random, sequential, not stated), blinding (double, single, or not blind), completeness (percentage of randomized participants completing the immunization schedule and the follow-up period), and the surveillance procedure used to detect cases.
The overall risk ratio (RR) was used to report the relative rates of cholera cases in vaccinated and placebo groups. This figure was converted to vaccine efficacy using the formula: % vaccine efficacy = (1-RR) x 100%.
Overall risk ratio was also used for adverse effect rates and other outcomes.
We anticipated between-trial variation in estimates of vaccine efficacy as there are several sources of heterogeneity which cannot be standardized. For example, the studies included in this review have been undertaken in a range of countries, each of which has a different pattern of exposure to the cholera pathogens. There are also major differences in the formulation of the vaccines.
To account for these differences in the analysis where significant heterogeneity (P < 0.1) was encountered between the study results, we have incorporated it into the analysis by reporting the results of the analysis using the random-effects model, presented in the results section as a letter R following a result. Elsewhere we have reported the results of analyses using the fixed-effect model.
It was defined a priori that subgroup analyses would be done for different age groups (under and over five years), and over time.
We split trials that included several active arms receiving separate vaccines into individual references (denoted as i, ii, iii, etc). As each active arm is compared to the same placebo group it is important that the analysis does not count the participants and cases in the placebo group more than once. This was prevented by dividing the placebo cases and participants as evenly as possible between the arms. The validity of this approach was confirmed in a second analysis in which the active arms within each trial were added together before the trials were pooled. This gave identical results in analyses using a fixed-effect model, and very similar, but slightly less conservative, results when using a random-effects model.
Description of studies
Sixteen trials fulfilled the inclusion criteria, although 26 comparisons are included in the review since some trials had more than one arm and we reported on these separately ('Characteristics of included studies'). To avoid counting the control group more than once in these trials, the control cases and participants were divided as evenly as possible between the arms. Fifteen trials were excluded ('Characteristics of excluded studies').
All the included trials tested injected vaccines of which there were two kinds: killed whole cell (KWC) or purified antigen. Various serotypes and formulations of KWC vaccines or purified antigen fractions were tested. All compared the vaccine with placebo (active or inactive).
The trials were conducted starting in 1963 and continuing until the late 1970s. Most were large and required massive programs to undertake the logistics of vaccination and surveillance. Several series of large trials (a total of several hundred thousand people in each site) were conducted in four sites in endemic areas: the Matlab study area of East Pakistan (later Bangladesh); Calcutta, India; Negros Occidental province, Philippines; and Surabaya, Indonesia. One smaller trial (998 participants) was conducted in the former USSR (Burgasov 1976).
Some of these trials (usually the first one in each series) investigated safety and immunogenicity. Most trials included clinical outcomes detected during massive population surveillance operations. All trials observed incidence of natural infection by cholera. The trial conducted in the former USSR investigated only safety and immunogenicity (Burgasov 1976). In terms of this review's outcome measures, 24 comparisons reported on vaccine efficacy (cholera cases and/or deaths) and 11 comparisons considered adverse effects. Nine reported on both types of outcome. Benenson 1968a and Burgasov 1976 provided data on adverse effects only.
Individual trial descriptions by location
East Pakistan (later Bangladesh)
Six quasi-randomized controlled trials were conducted in this region, but nine comparisons were included since two trials had several arms (denoted as i, ii, and iii), and we reported on these separately; see Benenson 1968b-i and Benenson 1968b-ii; and Mosley 1970-i, Mosley 1970-ii, and Mosley 1970-iii. One trial reported on adverse events only (Benenson 1968a). The comparisons differed in the participant age groups: four included all ages (Benenson 1968a; Oseasohn 1965; Benenson 1968b-i; Benenson 1968b-ii); four included children aged up to 14 years (McCormack 1969; Mosley 1970-i; Mosley 1970-ii; Mosley 1970-iii); and one included females of all ages and males up to age 15 years (Curlin 1975).
The trials compared types of cholera vaccine with active placebos or various schedules of vaccine against active placebos (shown in order of date started):
- Benenson 1968a: several types of injected KWC with one or two doses versus two active placebos (typhoid/paratyphoid A/paratyphoid B (TAB) and tetanus toxoid).
- Oseasohn 1965: injected, single-dose KWC versus active placebo (TAB).
- McCormack 1969: various schedules (one or two initial doses plus two annual boosters; two initial doses without boosters) of injected KWC versus two active placebos (tetanus and diphtheria toxoids).
- Mosley 1970-i, Mosley 1970-ii, and Mosley 1970-iii: three types of injected KWC vaccine (one initial dose, one booster dose at one year) versus two active placebos (tetanus and diphtheria toxoids). The three KWC vaccines were Classical Ogawa (Mosley 1970-i), Classical Inaba (Mosley 1970-ii), and El Tor (Mosley 1970-iii).
- Curlin 1975: two doses of lypohilized cholera toxoid (glutaraldehyde treated) versus active placebo (diptheria-tetanus toxoid).
Four randomized controlled trials were conducted in this region, but five comparisons are included. Two trials were reported in one publication (denoted as a and b), and one of these trials had two arms (denoted as i and ii): das Gupta 1965a; and das Gupta 1965b-i and das Gupta 1965b-ii. All age groups were included in the trials.
All four randomized controlled trials compared one dose of an injected KWC vaccine with an active placebo (shown in order of date started):
- Taneja 1965: one-dose injected KWC versus active placebo (TAB).
- das Gupta 1965a: one-dose injected KWC versus active placebo (TAB).
- Pal 1980: one-dose injected Classical KWC with alum adjuvant versus active placebo (tetanus toxoid).
One randomized controlled trial was conducted in this region, although it had two arms (Saroso 1978i; Saroso 1978ii). The trial compared one-dose injected KWC vaccine with an active placebo (tetanus toxoid) in all age groups. Saroso 1978i used a non-aluminium-hydroxide adsorbed KWC vaccine, while Saroso 1978ii used an aluminium hydroxide-adsorbed KWC vaccine.
Philippines (Negros Occidental province)
Four randomized controlled trials were conducted in this region, but nine comparisons are included since two trials had several arms (denoted as i, ii, etc) and we reported on these separately. The trials were conducted in all age groups. All trials compared a KWC vaccine with active placebo (shown in order of date started):
- PCC 1968: one or two doses (at three-week intervals) of El Tor KWC vaccines versus active placebo (typhoid vaccine).
- PCC 1973a-i, PCC 1973a-ii, PCC 1973a-iii, and PCC 1973a-iv: four different types of injected single-dose KWC vaccine versus active placebo (typhoid vaccine). The four KWC vaccines were El Tor Inaba (PCC 1973a-i), El Tor Ogawa (PCC 1973a-ii), Classical Ogawa (PCC 1973a-iii), and Classical Inaba (PCC 1973a-iv).
- PCC 1973b: single-dose Classical KWC injected subcutaneously or intradermally versus active placebo (typhoid vaccine).
One randomized controlled trial was conducted in this region (Burgasov 1976). This trial compared three types of one-dose injected Classical KWC and a partially purified cholera toxoid with inert placebo (sterile physiological solution). Only adults (both sexes) were included.
Risk of bias in included studies
The details for each trial are given under 'Method' in the 'Characteristics of included studies'. We assessed the efficacy and adverse effect trials separately.
Efficacy trials: 14 trials with 24 comparisons
The methodological quality of the efficacy trials was relatively high, considering their age.
Method of treatment allocation
Nine trials with 16 comparisons stated that the allocation method was randomization although only one trial mentioned a particular method (Latin Square (Azurin 1965i; Azurin 1965ii; Azurin 1965iii)). The other five trials (eight comparisons) used a sequential method such as alternate census number (Curlin 1975, all East Pakistan trials). These trials have therefore been classified as quasi-randomized controlled trials rather than randomized controlled trials, and allocation concealment is regarded as inadequate. All of the other trials mentioned some kind of coding system or identical preparation of placebo and are thus classified as adequate for allocation concealment.
All efficacy trials were stated to be double blind with the exception of Curlin 1975 (single, possibly double).
The major flaw in the reporting of the efficacy trials is the lack of information on the completeness (ie the percentage of randomized participants completing the immunization schedule and the follow-up period). In many trials there was a large difference between the number randomized and the number who actually participated. Some trials reported on the number completing the vaccination schedule (71% for Curlin 1975). Most trial reports provided little information on the percentage of participants who completed even the initial period of follow up. The India, Indonesia, and Philippines trials gave no information on this aspect of the trials. The little information on follow up that can be gleaned from some of the other trials suggests that dropout was not a serious problem – for example, dropout appears to have been less than 5% by two years in McCormack 1969, about 5% by one year in Mosley 1970-i, Mosley 1970-ii, and Mosley 1970-iii, and about 10% in Oseasohn 1965. It is easy to appreciate that keeping track of participants in trials with many thousands of people per arm would be a problem. However, since differential dropout is a serious potential source of bias in vaccine trials, and more information on this topic (perhaps by sampling a proportion of participants) would have been reassuring.
Surveillance procedure used to detect cases
Of the 14 efficacy trials, five used only active surveillance for cases, six used only passive, and three used both (details in Appendix 2). In all but one of the five East Pakistan/Bangladesh efficacy trials, it was claimed that daily or twice-weekly surveillance was carried out at home. In India, all four trials used passive surveillance, whereby cases were not detected unless they presented for treatment or sent a postal or telephone message. The Indonesia trial also used passive surveillance. In the Philippines, one trial used active surveillance only and the other three used a combination of passive surveillance and house to house visits, although the frequency and duration of this activity is not stated.
Adverse effect trials: 7 trials with 11 comparisons
Method of treatment allocation
Allocation method in the adverse effect trials was stated to be randomization in all trials except Benenson 1968a (sequential by census), which was classified as a quasi-randomized controlled trial. Allocation concealment was classed as inadequate in this trial; all other trials were classed as adequate for allocation concealment.
Blinding was only single (possibly double, but not clear) in one of the adverse effect trials (Burgasov 1976). All other trials were stated to be double blind.
In the adverse effect trials, follow up was usually short and dropouts minimal. Completion of follow up (ie the percentage of randomized participants completing the immunization schedule and the follow-up period) was 100% in two of the trials (Benenson 1968a; Burgasov 1976) but not stated in the others.
Surveillance procedure used to detect cases
The seven trials reporting on adverse effects were very poor in reporting the methods of surveillance (details in Appendix 2). However, since most reported on adverse events within 24 hours of vaccination, it is likely that individuals were actively assessed during that time period. In Burgasov 1976, home follow up continued for 30 days, although the frequency was not stated. In one trial in the Philippines where adverse effects were assessed (Azurin 1965iii), passive surveillance occurred in addition to active follow up because numerous participants reported to health facilities with adverse events.
Effects of interventions
Outcomes considered were cholera cases assessed after different lengths of follow up (up to seven months, up to one year, during year two, three, four, and five after follow up) for which data were available. During the first year of follow up, data from trials may appear in either 'up to seven months follow up' or 'up to one year follow up' depending on the duration of surveillance. Trial results appear in both categories only if the trials reported additional data for the second half of the first year of follow up. It should be noted that only three trials continued follow up for more than two years.
Injected cholera vaccine vs placebo (no booster)
The vaccination schedule for the trials considered here was either a single dose (all trials except McCormack 1969 and PCC 1968) or a 'short schedule'. In McCormack 1969, half of the participants had one dose and the other half had a short schedule of two doses given up to 35 days apart. For this analysis, participants who had booster doses at one year in McCormack 1969 were only included before one year of follow up. In PCC 1968 the data were combined from groups given either one dose or two doses at three-week intervals. Trials with booster doses at one year (Mosley 1970-i; Mosley 1970-ii) were excluded from this comparison after one year's follow up.
1. Cholera cases
Injected cholera vaccines were more effective than placebo at reducing risk of cholera cases for up to two years after immunization ( Analysis 1.1): up to seven months (RR 0.44, 95% CI 0.37 to 0.53; 2,098,146 participants, random-effects model); up to one year (RR 0.52, 95% CI 0.42 to 0.65, random-effects model; 1,512,573 participants); and year two (RR 0.58, 95% CI 0.45 to 0.75; 718,579 participants). Considering all age groups together, the vaccines were not significantly efficacious in years three (33,028 participants), four (18,969 participants), and five (18,969 participants); see Analysis 1.1. The corresponding estimates for vaccine efficacy for all vaccines combined are 56% (95% CI 47% to 63%) for up to seven months, 48% (95% CI 35% to 58%) for up to one year, and 41% (95% CI 24% to 55%) for year two.
Analysis 1.2 examines deaths (all-cause and cholera) in the first year of follow up. With the vaccine there was no reduction in all-cause deaths (RR 0.99, 95% CI 0.72 to 1.34; 26,743 participants), but there was a significant reduction in cholera deaths (RR 0.49, 95% CI 0.25 to 0.93; 837,442 participants). Note that this comparison included the Philippines trial arm of Azurin 1965iii that tested the cholera oil adjuvant vaccine, which had serious adverse effects.
Injected cholera vaccine vs placebo (stratified analyses)
3.1. Cholera cases by age group
This comparison includes only those trials that reported age-specific outcomes. A single dose or short schedule without booster was used by all trials except two trials that had one booster dose at one year after the first dose (Mosley 1970-i; Mosley 1970-ii). At each time point, we stratified the participants by those aged up to five years and those aged over five years: up to seven months' follow up ( Analysis 2.1); up to one year follow up ( Analysis 2.2); year two follow up ( Analysis 2.3); year three follow up ( Analysis 2.4); year four follow up ( Analysis 2.5); and year five follow up ( Analysis 2.6).
In the first year of follow up, there is little age-related difference in the reduction in risk of cholera cases between the vaccine and placebo when stratified by age group ( Analysis 2.2): up to five years (RR 0.45, 95% 0.35 to 0.59; 250,941 participants); and greater than five years (RR 0.51, 95% CI 0.42 to 0.63; 815,791 participants). These translate to efficacies of 55% (95% CI 41% to 65%) and 49% (95% CI 37% to 58%), respectively. There were two trials in which the efficacy at one year was notably better in the younger age group: Pal 1980 (89% versus 56%); and Saroso 1978ii (71% versus 43%). Both trials used alum-absorbed vaccine, suggesting that this adjuvant may increase efficacy in young children. This effect was not observed in Saroso 1978i, which used the same vaccine as Saroso 1978ii without alum (efficacy 43% and 44% in participants aged up to five years and over five years, respectively).
In the second year of follow up, the vaccines were not significantly efficacious in children aged up to five years (RR 0.83, 95% CI 0.52 to 1.31; 42,039 participants) whereas they were in older participants (RR 0.36, 95% CI 0.23 to 0.57; 241,578 participants); see Analysis 2.3. These translate to efficacies of 17% (95% CI -31% to 48%) and 64% (95% CI 476% to 7686%), respectively. This difference was similar also at year three when the vaccines had little effect in children aged up to five years (RR 0.64, 95% CI 0.39 to 1.09; 24,866 participants), but they were still protective in the older participants (RR 0.24, 95% CI 0.11 to 0.54; 41,852 participants); see Analysis 2.4. The equivalent efficacies are 36% (95% CI -10% to 63%) and 76% (95% CI 45% to 90%), respectively.
3.2. Cholera cases by vaccine schedule
This comparison specifically examined the question of whether booster doses of injected vaccines improve the duration of protection. Two trials (one with two sub-trials) included booster schedules: McCormack 1969 and Mosley 1970-i and Mosley 1970-ii. The McCormack 1969 trial included both non-booster and booster arms: the non-booster arm had two doses on a short schedule, while the booster arm had either one or two initial doses followed by additional doses after one and two years. Mosley 1970-i and Mosley 1970-ii tested Classical monovalent Ogawa (Mosley 1970-i) or Inaba (Mosley 1970-ii) vaccines with a single initial dose followed by one booster shot at one year. In both trials the non-booster and/or placebo groups received placebo shots instead of booster doses to maintain blinding. For up to one year of follow up before the booster was given, no difference would be expected between short schedule and booster subgroups: RR 0.55 (95% CI 0.45 to 0.68, random-effects model; 1,442,164 participants) for the single-dose schedule; and RR 0.35 (95% CI 0.13 to 0.98, random-effects model; 64,208 participants for the booster subgroups respectively; see Analysis 3.1. There was no difference between subgroups during year two either ( Analysis 3.2). In the third year, significant protection was observed in the booster schedule group: RR 0.34 (95% CI 0.15 to 0.77, random-effects model; 60,941 participants), but not in the other groups ( Analysis 3.3). Similar trends were seen in year four ( Analysis 3.4) and year five ( Analysis 3.5).
3.3. Cholera cases by vaccine type
This comparison examines the efficacy of different biotypes and serotypes of cholera vaccine (including purified antigens) up to one year after immunization. Overall, all types of vaccine except a toxoid (Curlin 1975) demonstrated protective efficacy which ranged from 38% (10% to 57%) (R) for El Tor 01 Ogawa plus Inaba KWC injected, to 86% (25% to 97%) (R) for Classical 01 Inaba KWC injected ( Analysis 4.1).
4. Adverse effects
4.1. Versus inert placebo
Only one trial used an inert placebo (Burgasov 1976). As shown in Analysis 5.1, the vaccine used in this trial caused malaise in 11% of recipients (RR 4.36, 95% CI 1.79 to 10.60), tenderness in 38% of recipients (RR 9.56, 95% CI 4.82 to 18.95), erythema in 28% of recipients (RR 2.82, 95% CI 1.83 to 4.34), and local infiltration in 14% of recipients (RR 14.04, 95% CI 3.50 to 56.33). All other adverse effects had no higher frequency in the vaccinated versus placebo groups.
4.2. Versus active placebo
When compared to active placebo ( Analysis 5.2), the vaccines caused vomiting in 1.5% of recipients (RR 10.43, 95% CI 1.34 to 81.22) and tenderness in 26% of recipients (RR 1.26, 95% CI 1.04 to 1.53). There were between-trial differences in the risk ratios for other adverse effects, with one trial (Pal 1980) showing large (RR > 1.5) increases in headache, fever, erythema, and swelling, although the averages across all trials were not statistically significant. In Benenson 1968a, the adverse effects were not classified beyond systemic or local;13% of participants had systemic effects (RR 2.30, 95% CI 1.10 to 4.80) and 40% had local effects (RR 3.48, 95% CI 2.14 to 5.63). One trial in the Philippines with three arms reported on trials with Classical vaccine (Azurin 1965i), El Tor vaccine (Azurin 1965ii), and Classical vaccine with oil adjuvant (Azurin 1965iii). There were serious adverse events observed in this trial when participants presented to health facilities (abscesses, ulcers, or hard masses at the site of vaccination), and it is reported that 96% of these occurred in the group who received the oil adjuvant vaccine (Azurin 1965iii). It was also reported that erythema, swelling, pain, induration, fever, and a feeling of weakness were experienced by participants. Overall the percentage of persons experiencing adverse events of any type was 0.8% in those receiving Classical vaccine (Azurin 1965i), 1.7% in those with El Tor vaccine (Azurin 1965ii), 96.1% in the Classical/oil adjuvant vaccine Azurin 1965iii, and 1.4% in the placebo group. However, no breakdown of specific symptom by vaccine group was given, except for the severe events mentioned above. Therefore we cannot include these results in the Analysis 5.2.
We included trials of injected cholera vaccines that had clinical outcomes (cholera cases, deaths, and adverse effects). We excluded trials with only immunological outcomes because our main questions were the efficacy and safety of injected cholera killed or subunit vaccines. The number of cases of cholera at different time periods after vaccination was our major outcome measure. We also assessed deaths (both all-cause and cholera specific) by year one of follow up, although this was investigated in few trials. Adverse effects were assessed by relatively few trials with a high variability in definition and measurement which made synthesis across different studies difficult.
The results of our review show that injected cholera vaccines are relatively efficacious in the first seven months. There was no evidence for a marked decline in efficacy in the second half of the first year or in year two, even without a booster dose.
Efficacy estimates stratified by age group (under or over five years) showed little difference in the first year. In year two, the vaccines were significantly less efficacious in children under five years than in older individuals. This difference persisted at year three when the vaccines had little effect in children aged less than five years (efficacy 20%, 95% CI -14% to 43%) but were still strongly protective in persons over five years (efficacy 57%, 95% CI 38 to 71%). By years four and five, neither age group was protected.
Both short vaccination schedules (single dose or two doses up to one month apart) and schedules with annual booster doses induced equivalent protection for two years in recipients. After year two, the booster schedule appeared to provide superior protection: efficacy was 9% for short schedules and 66% for booster schedules in year three and in year four the respective estimates were 7% and 39%. Our assessment of short versus booster schedules is based on a limited number of trials which included boosters (McCormack 1969; Mosley 1970-i; Mosley 1970-ii), only one of which continued beyond three years.
Injected cholera vaccines reduced cholera deaths by half, but they were of marginal efficacy in preventing all-cause death.
Injected cholera vaccines appear to be reasonably safe and were relatively well tolerated. Injected cholera vaccines did not cause significant increase in most individual systemic adverse effects (fever, malaise, headache) compared to active placebo, although they did cause increased malaise compared to inert placebo, and increased vomiting and unspecified systemic reactions compared to active placebo.
Injected cholera vaccines caused an increased number of local adverse effects including erythema, tenderness, and infiltration compared to inert placebo, and unspecified local reactions when compared to active placebo.
Our decision not to include serological outcomes in this review is based partly on the uncertainty of the relationship between protection afforded by the vaccine and a rise in antibody titre following immunization. Previously, the mouse protection index was considered the best correlate measure (Joo 1974), but since then most studies have quantified immunogenicity in terms of serum anti-toxin antibodies and/or vibriocidal antibodies, often in assays which are serotype specific (Inaba and Ogawa). Anti-toxin antibodies may protect by neutralizing cholera toxin, while vibriocidal antibodies may protect against colonization. Currently the critical protective immunity is thought to be antibacterial (vibriocidal) rather than antitoxic (Davis 1995). However, elevated serum vibriocidal antibodies, which may exist in persons in endemic areas, are often not further boosted by either vaccination or exposure to cholera, so rates of seroconversion may not correlate well with vaccine efficacy in these areas. Moreover, studies with serological outcomes reported results in a variety of assays using different definitions of seroconversion, making it very difficult to sum-up in a homogeneous way. Trials with clinical outcomes are the definitive method of assessing protection.
Historically six criticisms have been levelled at injected cholera vaccines. Firstly the protection from these vaccines is frequently stated to be below 30% at four to six months after vaccination (Sánchez 1997), or "modest and short lived" (Clemens 1994), or 50% to 70% with a short duration of three to six months (Feeley 1978), or not exceeding 50% to 60% (Joo 1974). We concur with the estimates of the latter two authors of overall protective efficacy, since our estimate is 57% (50% to 64%) after seven months. But we do not confirm the short duration of efficacy, since our estimate is 51% (4%1 to 59%) efficacy in the first year and 47% (36% to 56%) in the second year.
Secondly, injected cholera vaccines were stated to give "incidence of significant local reactions in up to 30% of vaccinees" (Sánchez 1997) and "immunization is generally accompanied by mild fever, malaise and headache" (Joo 1974). We found some basis for the former statement, but not for the latter since only up to 13% of vaccinees had systemic adverse effects. In general, injected cholera vaccines were well tolerated and the nature of the relatively minor adverse effects must be weighed against the possible severity and catastrophic impact of cholera.
Thirdly, protective efficacy in children aged less than five years was stated to be "below 30%" (Sánchez 1997) or "poor" in the same age group (Joo 1974). Again the letter of these statements is not borne out by the results of our meta-analysis, as in the first year the level of protection is equivalent in under- and over-five year olds (51% and 55% in the two groups). However, protection certainly persists longer in persons aged over five years, in whom efficacy was 57% in the third year after immunization.
Fourthly, it was suggested that injected cholera vaccines do not reduce carriage of V. cholerae 01 (Clemens 1994; Sánchez 1997). We are not able to comment on this statement as the trials reviewed to date have not addressed this issue specifically. Moreover Clemens 1994 noted that the role of asymptomatic excretion of V. cholerae in epidemics is unclear, and consequently the public health importance of interrupting 'carrier' status is not known. Cvjetanović 1978a and Cvjetanović 1978b suggested that lifelong healthy carriers epidemiologically play a negligible role.
Fifthly, Sommer 1973b thought that injected KWC vaccines have no role in controlling an epidemic, assuming an efficacy of around 50%. Although the included trials have not addressed this issue, it certainly appears likely that seeking out and vaccinating household contacts of cases (as considered by Sommer 1973b) would be too late to prevent infection of such secondary cases, and this is supported by one excluded trial (Sommer 1973a). However we believe that this does not discount the potential indirect effect that vaccinating a community would have on controlling or preventing an epidemic (Clemens 1996). Vaccine trials to date have involved individual, rather than community, randomization. The efficacy (or rather effectiveness) of a vaccine is likely to be much greater if given to a whole community rather than to dispersed individuals, if the vaccines reduced excretion of bacteria or if herd immunity was attained. This would have to be tested in trials of different design than the ones reviewed here.
Finally, it is asserted that injected cholera vaccines necessitate more than one inoculation to be effective. We did not find this to be the case for parenterally administered vaccines. Most of trials in our analysis used only one dose. For example, eight of the 10 trials (16 of 18 subtrials) of injected cholera vaccine analysed at seven months' follow up used a single dose, with a summary estimate of 54% efficacy. At two years, our summary estimate for these injected cholera vaccines was 39% (21% to 52%) protective efficacy in the second year of follow up; this was derived from six trials, in five of which only one dose was given. Booster doses do not provide enhanced protection until years three and four.
We conclude that injected cholera vaccines are generally safe and relatively effective, with a combined estimate of 57% efficacy at one year and 47% at two years. Injected cholera vaccines achieve this level of efficacy after one injection or a short schedule of two doses; extending this level of protective efficacy for up to four years requires an annual booster. Vaccines were of equivalent efficacy in children under five years as in older age groups in the first year, but protection persisted longer (up to three years) in older children and adults.
These data provide the background information against which to compare the efficacy of oral cholera vaccines, which are the subject of a separate Cochrane Review (Abba (in progress)).
Disaggregation of study results for this review caused a considerable conceptual and logistic burden. Study reports frequently discussed more than one separate trial in the same published report, and there were multiple publications from single trials. Repetition of study results is to a certain extent inevitable in such large studies of long duration, which lend themselves to multiple publications of progress reports or partial reports of different outcomes. However, we feel that unnecessarily complicated trials, combined with multiple reporting, may have contributed to the underestimation of the extent and longevity of protection induced by injected cholera vaccines. Nevertheless it is not clear how the myth of requirement for six-monthly boosters for this type of vaccine originated, since no trial tested such a schedule, and the few trials comparing annual booster and non-booster schedules showed no advantage of booster until the third year of follow up.
Over one million people, including infants and children, have taken part in large, good quality efficacy trials of injected cholera vaccines over the last 35 years. The overwhelming majority of these participants have been poor residents of cholera-endemic areas. Also hundreds of researchers have devoted years of their careers to these trials. Their contributions deserve to be better recognized by thorough examination of the results of these trials. It appears that the adverse effects have been overestimated and relatively effective injected cholera vaccines have been underestimated.
Implications for practice
Injected cholera vaccines are not currently available and therefore not recommended for either residents of endemic areas or travellers. The accepted wisdom is that they provide weak, partial protection of very short duration and require multiple doses. However, this meta-analysis demonstrated significant protection for populations living in endemic areas for up to two years following a single dose, and for three to four years with annual booster. Risk of death from cholera was also reduced by 50% in the first year after vaccination.
Implications for research
All cholera vaccine trials to date have been individually rather than group randomized. Research is needed on whether cholera vaccines can control epidemics if given on a population basis.
Results of two trials showing that alum adjuvant KWC injected vaccines were more effective in young children than older persons suggest that modern adjuvants could possibly increase the efficacy of injected KWC vaccines in this age group.
This review provides a solid background of evidence for effects of cholera injected vaccines against which to compare the effects of oral vaccines.
The authors would like to thank Drs Daniela Rivetti, Franco Bottasso, Ron Behrens, and Alaistair MacMillan who applied quality criteria, and Professor Myron Levine and Mrs Carol Hobbs for assistance. We are very grateful to Mark Pratt who conducted literature searches and initial screening, and developed the trial register.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. Detailed search strategies
Appendix 2. Summary of trials, comparisons, outcomes, and surveillance methods
Total trials: 16 (efficacy 14, adverse effects 7); comparisons: 26 (efficacy 24, adverse effects 11).
Last assessed as up-to-date: 22 February 2009.
Protocol first published: Issue 1, 1998
Review first published: Issue 3, 1998
Contributions of authors
Vittorio Demicheli, Tom Jefferson, Patricia Graves and Jon Deeks read all trials or trial abstracts and determined eligibility. Patricia Graves and Jon Deeks extracted trial data and assessed quality with the assistance of persons named in the Acknowledgments. Patricia Graves. Jon Deeks and Tom Jefferson conducted analyses with input from all authors on the results. All authors commented on the draft review.
Declarations of interest
Sources of support
- Ministry of Defence, UK.
- Department for International Development, UK.
- European Commission (Directorate General XII), Belgium.
Differences between protocol and review
The following phrase was added to 'Types of Interventions – Intervention': "Exception: Phase 1 trials, reporting only adverse effects, for vaccines that never reached efficacy trials."
The following phrase was deleted from "Types of Interventions – Control": "(trials) comparing types, doses or schedules of injected cholera vaccines". No such comparisons were made in the review.
Medical Subject Headings (MeSH)
Cholera [immunology; *prevention & control]; Cholera Vaccines [*administration & dosage; adverse effects; immunology]; Injections; Randomized Controlled Trials as Topic; Vaccination [methods]; Vaccines, Inactivated [administration & dosage; adverse effects; immunology]
MeSH check words
Adult; Child; Child, Preschool; Humans; Infant
* Indicates the major publication for the study