Summary of findings
Despite improving trends in mortality rates, diarrhoea still causes 15% of all deaths in children under five years and accounts for nearly 1.4 million child deaths in developing countries every year (Black 2010). It is estimated that on average a child under five years will have approximately 3.2 episodes of diarrhoea each year (Kosek 2003). Diarrhoea is also an important cause of malnutrition, particularly when it is prolonged (Brown 2003).
Treatment of diarrhoea with oral rehydration solution (ORS) reduces mortality due to dehydration. Zinc supplementation could help reduce the duration and the severity of diarrhoea, and therefore have an additional benefit over ORS in reducing children mortality (Bhutta 2008).
There are several different mechanism of action of zinc on acute diarrhoea (Berni Canani 2010). Zinc influences the activity of over 300 enzymes, some of which are responsible for DNA replication and transcription (IZiNCG 2004). Zinc promotes immunity, skin and mucosal resistance to infection, growth, and development of the nervous system (Hess 2009, MacDonald 2000, Prasad 2008). It is also an important anti-oxidant and preserves cellular membrane integrity (O' Dell 2000, Powell 2000). At the level of gastrointestinal system, zinc restores mucosal barrier integrity and enterocyte brush-border enzyme activity (Roy 1992; Shankar 1998), it promotes the production of antibodies and circulating lymphocytes against intestinal pathogens (Sazawal 1997b; Albert 2003; Raqib 2004), and has a direct effect on ion channels, acting as a potassium channel blocker of adenosine 3-5-cyclic monophosphate-mediated chlorine secretion (Hoque 2009, Hoque 2005).
Rationale for supplementation
Zinc deficiency is mainly due to inadequate dietary intake and is estimated to be common in many countries (IZiNCG 2004; Wagstaff 2004; Hess 2009). High levels of zinc are found in 'expensive foods' (eg meat and fish). Zinc is also present in nuts, seeds, legumes, and whole grain cereal, but the high phytate content of these foods interferes with its absorption. Zinc cannot be stored in the body, and nearly 50% of zinc excretion takes place through the gastrointestinal tract and is increased during episodes of diarrhoea. Young children who are regularly exposed to gastrointestinal pathogens and have diets low in animal products and high in phytate-rich foods are most at risk.
Factors that could influence the effects of supplementation
There are a number of factors that could influence the size of any effect when using zinc to treat diarrhoea, and these will be explored in this review.
Type of diarrhoea
Acute and persistent diarrhoea are very different conditions. Acute diarrhoea in children in developing countries is usually infectious, while persistent diarrhoea has a number of causes including malnutrition, parasitic infections, tuberculosis, human immunodeficiency virus (HIV), food intolerance, and malabsorption.
Zinc requirement varies with age and is highest in children due to their rapid rates of growth. Infants, however, have lower requirements (IZiNCG 2004) as healthy normal birthweight infants have adequate zinc levels at birth from maternal sources even if maternal stores are sub optimal (Iqbal 2001). Infants may be able to mobilize hepatic stores accumulated during gestation (Zlotkin 1988) and are less likely to have had a zinc-depleting illness. Breastfeeding will provide zinc supplementation and protective immune factors against infections (Krebs 1999).
The recommended daily allowance for zinc is markedly higher for malnourished children (2 to 4 mg/kg/day) than healthy children (3 to 5 mg/day for children under five years) (IZiNCG 2004). This is because zinc deficiency is considered more severe in malnourished children and thus the benefit of zinc supplementation may be greater.
Zinc supplementation may have different effects according to the level of zinc deficiency in the country. It is important to verify whether zinc supplementation is effective in countries with high, or even medium or low risk of zinc deficiency (IZiNCG 2004).
The World Health Organization (WHO) and United Nations Children's Fund (UNICEF) recommend 10 to 20 mg of zinc per day for children with diarrhoea, at least twice the recommended daily allowance (WHO/UNICEF 2004).
Types of zinc salt
Zinc is usually given as zinc sulphate, zinc acetate, or zinc gluconate, which are all water-soluble compounds (IZiNCG 2004).
Concomitant iron or copper supplementation
Iron and zinc deficiencies often co-exist. These two compounds may compete for the same absorptive pathways, and iron may interfere with zinc utilization (Gunshin 1997; Kordas 2004). A review of combined supplementation showed that giving zinc with iron resulted in a lower increase in iron levels compared to giving iron alone; iron supplementation alone had no effect on zinc status (Fischer Walker 2005). A trial that assessed combined supplementation on diarrhoea and malaria morbidity showed that zinc combined with iron reduced zinc's protective effect against diarrhoea (Richard 2006). Several trials have also reported a negative interaction of the combined supplementation on physical growth and development (Rosado 1997; Dijkhuizen 2001; Zlotkin 2003; Lind 2004; Bhandari 2007). Some protocols suggest supplementing malnourished children also with copper because these children are also prone to copper deficiency (Beshgetoor 1998).
Zinc effect may vary according to the study setting (hospital or community), due to differences in adherence rates, and other factors such as diet.
Zinc can cause vomiting because of its metallic taste (Fontaine 2001). In high doses, zinc can also cause epigastric pain, lethargy, and fatigue (IZiNCG 2004). One small study suggested a possible increase in mortality in malnourished children supplemented with 6 mg/kg/day of zinc compared to those supplemented with 1.5 mg/kg/day (Doherty 1998). Copper deficiency with zinc supplementation can occur although usually only when zinc is consumed in very high doses (100 to 300 mg/day for adults) over a long period of time (IZiNCG 2004), and malnourished children are at particularly high risk of this due to lower basal copper levels.
Previous systematic reviews
Previous meta-analysis indicated that zinc supplementation in diarrhoea is effective (Bhutta 2000b;Lukacik 2008 ;Patro 2008; Haider 2009). This Cochrane Review will have an up-to-date search for trials and will explore more outcome measures of interest and more possible sources of heterogeneity.
To evaluate oral zinc supplementation for treating children with acute or persistent diarrhoea.
Criteria for considering studies for this review
Types of studies
Randomized controlled trials.
Types of participants
Children aged between one month and five years with acute or persistent diarrhoea, including dysentery.
We excluded trials of infants below one month and studies that exclusively enrolled children with particular conditions such preterm or low birthweight infants and children with HIV.
Acute diarrhoea is usually defined as three or more loose stools in a 24-hour period. Persistent diarrhoea is defined as diarrhoea lasting more than 14 days. Dysentery is a diarrhoeal illness in which blood is observed in the stool. The final day of diarrhoea is usually defined as the last day meeting the above definition followed by 48 hours without diarrhoea.
Types of interventions
Oral zinc supplementation of any zinc salt at doses of 5 mg/day or more for any duration.
Concurrent supplementation of other minerals and vitamins are eligible only if administered to both intervention and control group.
ORS plus zinc and food fortification interventions (such as milk fortification) are excluded as the amount of ORS/food consumed, and hence the zinc intake, would be less certain.
Types of outcome measures
Measures of diarrhoea duration
- Diarrhoea duration.
- Diarrhoea at three, five, and seven days after starting intervention.
Measures of diarrhoea severity
- Stool frequency.
- Stool output.
- Death (from any cause and diarrhoea specific).
- Serious adverse events (life-threatening or requiring hospitalization).
- Any adverse event that results in the discontinuation of treatment.
- Other adverse events, such as vomiting and reduced copper levels.
Search methods for identification of studies
We will attempt to identify all relevant trials regardless of language or publication status (published, unpublished, in press, and in progress).
Databases of published trials
We searched the following databases using the search terms and strategy described in Table 1: Cochrane Infectious Diseases Group Specialized Register (20 February 2012); Cochrane Central Register of Controlled Trials (CENTRAL), published in The Cochrane Library (2012, Issue 2); MEDLINE (1966 to February 2012); EMBASE (1974 to February 2012); LILACS (1982 to 20 February 2012); CINAHL (1982 to 20 February 2012), the metaRegister of Current Controlled Trials (mRCT; February 2012), ClinicalTrials.gov (February 2012), and the WHO International Clinical Trials Registry Platform (ICTRP) (February 2012).
Researchers and organizations
For unpublished and ongoing trials, we contacted individual researchers working in the field, including researchers at the WHO.
We checked the reference lists of all studies identified by the above methods.
Data collection and analysis
Selection of studies
All trials identified by the search strategy were screened by both authors, and full articles were retrieved for all potentially relevant trials. Both authors independently applied the inclusion criteria to the full reports using a pilot-tested form and scrutinized publications to ensure each trial was included once. Trial authors were contacted for clarification if necessary, and any disagreements were resolved through discussion and consensus after referring to the protocol; their solutions were recorded and reported.
Data extraction and management
Both authors independently extracted data using a pilot-tested data extraction form and entered the data into Review Manager 5. When data were missing or unclear, we contacted the trial authors for clarification. For dichotomous outcomes, the number of participants experiencing the event and the number assessed in each group were recorded. For continuous outcomes, the arithmetic means, standard deviations, and number assessed in each group were extracted. If continuous data were reported using geometric means, the standard deviations on the log scale were extracted; medians and ranges were extracted and reported in a table.
Assessment of risk of bias in included studies
Both authors independently assessed the risk of bias for each trial using the 'The Cochrane Collaboration's tool for assessing the risk of bias' (Higgins 2008). We have categorized our judgments as low risk of bias, high risk of bias, or unclear risk of bias, and this information has been used to guide the interpretation of the results. Where our judgement for efficacy trials was unclear we attempted to contact the trial authors for clarification and any differences of opinion were resolved through discussion. If data were missing or unclear, we contacted the trial authors.
Data were analysed using Review Manager 5. All results are presented with 95% confidence intervals (CI).
For dichotomous data, outcome measures were reported using risk ratio (RR). Given the high variation in control event rates, we did not calculate the number needed to treat (NNT). For continuous data summarized by arithmetic means and standard deviations, we used the mean difference (MD) to combine the results in a meta-analysis. Continuous data summarized using other summary statistics that could not be combined in a meta-analysis were presented in a table. We calculated geometric mean ratios and transformed them in the log scale for analysis, and presented them on the natural scale.
Assessment of evidence quality
The quality of evidence has been assessed using the GRADE methodology (GRADE 2008). The GRADE system considers ‘quality’ to be a judgment of the extent to which we can be confident that the estimates of effect are correct. The level of ‘quality’ is judged on a 4-point scale. Evidence from randomized controlled studies is initially graded as high and downgraded by one, two or three levels after full consideration of: any limitations in the design of the studies, the directness (or applicability) of the evidence, the consistency and precision of the results, and the possibility of publication bias.
The estimates of effect, and the GRADE assessments of our confidence in these estimates are displayed in 'summary of findings tables' for the main comparisons. Where we have downgraded the evidence our reasons for doing so are displayed in footnotes.
When making conclusions about the relative effects of the interventions we have used language which reflects the GRADE assessments and our confidence in the estimates, ie if the evidence is high quality we would say "zinc reduces..."; moderate quality "zinc probably reduces..."; low quality "zinc may reduce..."; and where the evidence is very low quality we have not drawn conclusions.
Subgroup analysis and investigation of heterogeneity
We assessed heterogeneity among trials by visually inspecting the forest plot, using the Chi
We stratified the analyses for acute diarrhoea or persistent diarrhoea as these are different conditions. We also stratified the results by age (children aged less than and greater than six months) because we observed clear a difference in zinc effect according to the age of children enrolled and significant heterogeneity if all the trials were pooled together. We explored the following potential sources of heterogeneity using subgroup analyses: nutritional status (malnourished children versus well-nourished plus moderate malnourished); geographical region (by continent and by high versus medium estimated risk of zinc deficiency as defined by the International Zinc Nutrition Consultative Group (IZiNCG 2004)); zinc dose (< versus > 20 mg/day); zinc salt (zinc sulphate versus zinc acetate versus zinc gluconate versus other type); concomitant copper or iron supplementation; and trial setting (hospital versus community trials). We also explored the effect of sex, although this was not specified in our original protocol.
We conducted a sensitivity analysis in which we limited the analyses to those trials with adequate allocation concealment, blinding (excluded those trials classified as unclear), and those that included an adequate number of randomized participants in the analysis (excluded those trials classified as inadequate or unclear). To take into account the participants for whom no outcome data were obtained we also conducted an intention-to-treat analysis for worst-case/best-case scenarios.
Description of studies
Twenty-four trials enrolling 9128 children met our inclusion criteria (see 'Characteristics of included studies'). The process of trials selection is reported in Table 2, and the reasons for excluding 127 studies are given in the 'Characteristics of excluded studies'.
Three of the included trials presented results divided in two or more subgroups: one trial presented two intervention groups of zinc 20 mg and zinc 5 mg, and one control group (Brooks 2005a); one trial presented data for three different study sites (Fischer Walker 2006); and one trial presented the results as children with low and normal zinc serum levels (Polat 2003). For these three trials there was no way to combine means and standard deviations, and thus we had to enter the data separately as Brooks 2005a (20 mg), Brooks 2005a (5 mg), Fischer Walker 2006 ETH, Fischer Walker 2006 IND, Fischer Walker 2006 PAK, Polat 2003 low Zn, and Polat 2003 normal Zn. Thus the number of total comparisons is 26.
Type of diarrhoea
Nineteen trials enrolled children with acute diarrhoea: 11 used this review's definition for acute diarrhoea (Faruque 1999; Dutta 2000; Strand 2002; Al-Sonboli 2003; Polat 2003; Bhatnagar 2004a; Brooks 2005a; Fischer Walker 2006; Dutta 2011); two trials respectively defined diarrhoea as presence of four (Sazawal 1995) or five (Bahl 2002) unformed stools in 24 hours; one shigellosis trial included patients with bloody mucoid diarrhoea (dysentery) or febrile diarrhoea less than five-days' duration (Roy 2008); one trial enrolled only children with diarrhoea and/or vomiting due to rotavirus infection (Dalgic 2011) and three trials did not report the definition of acute diarrhoea (Sachdev 1988; Roy 1997; Larson 2005). Five trials were on children with persistent diarrhoea (Sachdev 1990; Roy 1998; Bhutta 1999b; Penny 1999; Khatun 2001).
Two trials enrolled only children under six months (Brooks 2005a; Fischer Walker 2006). Thirteen trials only enrolled children over six months (Sachdev 1988; Sachdev 1990; Sazawal 1995; Bhutta 1999b; Faruque 1999; Penny 1999; Khatun 2001; Bahl 2002; Strand 2002; Roy 2008; Fajolu 2008; Patel 2009; Dutta 2011). Eight trials included children of different ages greater than one months (Roy 1997; Roy 1998; Dutta 2000; Al-Sonboli 2003; Polat 2003; Bhatnagar 2004a; Larson 2005; Patro 2010; Dalgic 2011).
Seven trials enrolled only malnourished children (Roy 1997; Roy 1998; Bhutta 1999b; Dutta 2000; Khatun 2001; Polat 2003; Roy 2008). One trial included only well-nourished children (Patro 2010), and one enrolled children regardless of nutritional status (Larson 2005). The remaining 14 trials enrolled children who were well nourished or with mild or moderate malnutrition. No trial included only severe malnourished children. There was some variability between trials in the definitions of malnutrition (most used 'weight/age'; only some used 'weight/height'); therefore we were unable to follow the definition of malnutrition proposed in the protocol.
Nineteen trials were conducted in Asia, two in South America, one in Europe (Patro 2010), one in Africa (Fajolu 2008), and one multicenter trial in Asia and Africa (Fischer Walker 2006) Thus, participants were from Bangladesh (Roy 1997; Roy 1998; Faruque 1999; Khatun 2001; Brooks 2005a; Larson 2005; Roy 2008), India (Sachdev 1988; Sachdev 1990; Sazawal 1995; Dutta 2000; Bahl 2002; Bhatnagar 2004a; Fischer Walker 2006 IND, Patel 2009), Pakistan (Bhutta 1999b; Fischer Walker 2006 PAK), Nepal (Strand 2002), Turkey (Polat 2003), Brazil (Al-Sonboli 2003), Peru (Penny 1999), Ethiopia (Fischer Walker 2006 ETH), Nigeria (Fajolu 2008) and Poland (Patro 2010).
Risk of zinc deficiency
All the trials were conducted in countries ranked as high risk for zinc deficiency (IZiNCG 2004), except for five trials conducted in countries at medium risk: Nepal (Strand 2002);Turkey (Polat 2003; Dalgic 2011); Brazil (Al-Sonboli 2003), and Nigeria (Fajolu 2008), and one recent trial from Poland, where zinc deficiency is considered rare (Patro 2010).
The zinc dose was 20 mg/day in eleven trials. Only two trials administered higher zinc doses: 40 mg/day (Dutta 2000); and 22 or 45 mg/day (Al-Sonboli 2003). Two trials, of which one in children aged less than six months only, gave zinc 10 mg/day (Fischer Walker 2006; Roy 2008), and another two trials used 10 mg for infants and 20 mg for children over six months (Fajolu 2008, Patro 2010). One trial used zinc 5 mg and 20 mg, but only in children aged less than six months (Brooks 2005a).
Four trials used different doses depending on age (zinc < 20 mg in infants and ≥ 20 mg in older children), but they did not report results separately for each treatment group (Faruque 1999; Bahl 2002; Strand 2002; Bhatnagar 2004a). We classified these trials as 'not assignable' and could not include them in the sensitivity analysis for zinc dose. Two trials reported a per kilo dose (2 mg/kg/day; 3 mg/kg/day;) (Bhutta 1999b, Patel 2009), and we were not able to include it in the subgroup analyses.
Type of zinc salt
Twelve trials used zinc sulphate, eight trials used zinc acetate (Roy 1997; Roy 1998; Faruque 1999, Khatun 2001; Strand 2002; Brooks 2005a; Roy 2008; Dalgic 2011), three used zinc gluconate (Sazawal 1995; Penny 1999; Bahl 2002), and one did not specify (Dutta 2011).
Concomitant copper or iron supplementation
One trial compared zinc alone versus zinc and copper versus placebo (Patel 2009).
Most trials were conducted in hospitals, with the exception of four community-based studies (Penny 1999; Bahl 2002; Strand 2002; Fischer Walker 2006), and one trial held both in hospital and community (Larson 2005).
Duration of treatment
Eleven trials administered zinc for two weeks. Of the remaining 12 trials, three gave zinc for seven days after recovery (Bahl 2002; Strand 2002; Polat 2003), two gave zinc until recovery (Al-Sonboli 2003; Brooks 2005a), one gave zinc for seven days (Khatun 2001), and one gave zinc for 10 days (Patro 2010). Five trials were unclear in respect of duration of zinc supplementation (Sachdev 1988; Sachdev 1990; Sazawal 1995; Dutta 2000; Dalgic 2011). One trial on adverse events administered only one dose of zinc (Larson 2005).
Zinc was administered as syrup in most trials; only four used powder (Sachdev 1988; Sachdev 1990; Penny 1999; Dalgic 2011), three used dispersible tablets (Al-Sonboli 2003; Larson 2005; Fischer Walker 2006), and one did not specify (Fajolu 2008).
The zinc dose was administered once a day in half of the trials, while the other half administered it twice a day (Sachdev 1988; Sachdev 1990; Khatun 2001; Roy 2008; Patro 2010) or three times a day (Roy 1997; Roy 1998; Dutta 2000; Polat 2003; Bhatnagar 2004a). One study administered zinc twice a day to infants, and a single dose to children over six months (Dalgic 2011). Two studies did not specify (Fajolu 2008; Patel 2009).
Twelve trials administered zinc alone; seven studies used zinc and multivitamin, which did not contain iron (Sazawal 1995; Roy 1997; Roy 1998; Bhutta 1999b; Khatun 2001; Bhatnagar 2004a; Roy 2008). One trial used zinc and vitamin A (Faruque 1999). One trial used concomitant copper (Patel 2009).
Twenty trials reported data on diarrhoea duration (Sachdev 1988; Sachdev 1990; Roy 1997; Roy 1998; Bhutta 1999b; Faruque 1999; Penny 1999; Dutta 2000; Khatun 2001; Bahl 2002; Al-Sonboli 2003; Polat 2003; Bhatnagar 2004a; Brooks 2005a; Fischer Walker 2006; Fajolu 2008; Patel 2009; Patro 2010; Dalgic 2011; Dutta 2011). Data were presented as means and standard deviations, or means and 95% CI. One trial reported data as medians and ranges (Roy 2008), and we could not compare these to the data from other trials.
Five trials reported on diarrhoea at day three (Penny 1999; Bahl 2002; Strand 2002; Polat 2003; Patel 2009), six trials on diarrhoea at day five (Penny 1999; Dutta 2000; Bahl 2002; Bhatnagar 2004a; Patel 2009; Dutta 2011), and 12 at day seven (Sazawal 1995; Faruque 1999; Penny 1999; Khatun 2001; Bahl 2002; Strand 2002; Polat 2003; Bhatnagar 2004a; Fischer Walker 2006; Roy 2008; Patel 2009; Patro 2010)
Stool frequency was reported in seven trials (Sachdev 1988; Sachdev 1990; Bahl 2002; Al-Sonboli 2003; Brooks 2005a; Fischer Walker 2006; Fajolu 2008). Three trials did not report cumulative data for the whole hospitalization period (Bhutta 1999b; Polat 2003; Patro 2010); instead they reported data on some given days (as number of stool on day two or four or other days), and these data could not be compared to the data from other trials.
Eight trials reported data on stool output (Roy 1997; Bhutta 1999b; Dutta 2000; Khatun 2001; Bhatnagar 2004a; Brooks 2005a; Patel 2009; Dutta 2011). Definitions and measurement units varied consistently between trials (see Table 3). Stool output was evaluated using pre-weighed disposable diapers with urine collected separately in three trials (Dutta 2000; Bhatnagar 2004a; Dutta 2011) and using pre-weighed containers with urine collected separately in one trial (Roy 1997). In one trial, stool weight was measured by using metabolic beds, and urine was collected separately using urine bags. The methods were not clearly stated in two trials (Bhutta 1999b; Khatun 2001).
One hospital trial (Dalgic 2011) and three community trials reported information on hospitalization (Penny 1999; Strand 2002; Fischer Walker 2006). The declared follow-up period for these trials was at least "until recovery from diarrhoea" in three trials (Strand 2002; Fischer Walker 2006; Dalgic 2011), and 15 days in another trial (Penny 1999).
Death was reported in seven trials. These trials had as follow-up times the duration of hospital stay (Roy 1998; Khatun 2001; Brooks 2005a), two weeks (Penny 1999, Patel 2009), until the diarrhoea episode was over (Fischer Walker 2006), and six months (Roy 2008).
Data on vomiting, as percentage of children who vomited, were available in 13 trials. Only one trial stated the case definition, with vomiting defined as a forceful emptying of stomach contents, and regurgitation as the unforced return of any amount of the swallowed syrup, liquids, or foods (Larson 2005). One trial on diarrhoea due to rotavirus reported the duration of vomiting (Dalgic 2011).
Risk of bias in included studies
|Figure 1. Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.|
|Figure 2. Methodological quality summary: review authors' judgements about each methodological quality item for each included study.|
Generation of allocation sequence
Sixteen trials reported methods that assured adequate allocation concealment. The remaining eight were unclear (Sachdev 1988; Sachdev 1990; Roy 1998; Khatun 2001; Al-Sonboli 2003; Brooks 2005a; Fajolu 2008; Dalgic 2011).
Inclusion of all randomized participants
Fifteen trials included more than 90% of the randomized participants in the analysis. Four included less than 90%, which we assessed as inadequate (Roy 1997; Bhutta 1999b; Roy 2008; Patro 2010), and the number included was unclear in the remaining five trials (Sachdev 1988; Sachdev 1990; Roy 1998; Dutta 2000; Fajolu 2008).
Effects of interventions
1. In children with acute diarrhoea
1.1. Diarrhoea duration
Overall, diarrhoea duration was reduced in children given zinc by about 12 hours (mean difference -12.63, 95% CI -21.05 to -4.21 hours, 4446 children, 15 trials, 19 comparisons, Figure 3), but there was significant heterogeneity between trials (I
|Figure 3. Zinc vs placebo for acute diarrhoea: diarrhoea duration (h)|
1.2. Diarrhoea on days three, five, and seven
Treatment with zinc resulted in fewer children continuing to have diarrhoea at day three (RR 0.77, 95% CI 0.67 to 0.89; 1568 children, three trials, four comparisons, Figure 4), at day five (RR 0.67, 95% CI 0.51 to 0.89;1730 children, five trials, four comparisons, Figure 5) and at day seven (RR 0.82, 95% CI 0.72 to 0.94; 5528 children, 10 trials, 13 comparisons, Figure 6). For all the three outcomes there was overall significant heterogeneity between trials. For diarrhoea at day three and five there were few trials, and it was difficult to explore heterogeneity. For diarrhoea at day seven heterogeneity was markedly reduced if results were stratified by age. No benefit of zinc was detected in children under six months (1074 children, three comparisons, Figure 6), while zinc had a benefit in children older than six months (RR 0.73, 95% CI 0.61 to 0.8; 3865 children, six trials, Figure 6) and in trials recruiting both age groups (RR 0.31, 95% CI 0.18 to 0.52; 589 children, three trials, four comparisons, Figure 6).
|Figure 4. Zinc vs placebo for acute diarrhoea: diarrhoea on day 3|
|Figure 5. Zinc vs placebo for acute diarrhoea: diarrhoea on day 5|
|Figure 6. Zinc vs placebo for acute diarrhoea: diarrhoea on day 7|
1.3. Stool frequency
There was no evidence of a benefit of zinc on stool frequency overall (2323 children, six trials, Analysis 1.5). Stratification by age reduced heterogeneity, and zinc was only shown to have a statistically significant benefit in the single trial which recruited both age groups (MD -5.90 stool/day, 95% CI -9.44 to -2.36, 74 children, one study, Analysis 1.5).
1.4. Stool output
Stool output was measured using different units at different time points, thus results could not be pooled together ( Table 3). Results are expressed as arithmetic mean difference (AMD) or geometric mean ratio (GMR).
One trial reported on children aged less than six months with no evidence of a difference (Brooks 2005a). Two trials reported on children aged more than six months with inconsistent results (Patel 2009, Dutta 2011). Three trials reported on children aged less than and greater than six months: two of these studies showed a reduction in stool output with zinc (Dutta 2000; Bhatnagar 2004a), while one showed no evidence of an effect (Roy 1997).
One hospital trial showed a significant reduction in the duration of hospitalization in children treated with zinc compared to those given placebo (MD 4.33 days ± 1.38 vs 5.81 ± 2.08 days). Of the two community trials reporting on hospitalization, one found no difference between groups (Strand 2002, 891 participants), and the second reported no hospitalizations in the zinc group and only one in the placebo group (Fischer Walker 2006, 1074 participants under 6 months of age).
Four trials specified the numbers of child deaths: two studies (316 children) did not observe any deaths (Brooks 2005a; Roy 2008); one trial (1032 children) reported one death in each treatment group (Fischer Walker 2006), and one trial (754 children) reported one death in the zinc group, no death in the zinc plus copper group, and two deaths in the placebo group (Patel 2009a (zinc)).
1.7. Adverse events
Eight trials reported vomiting, which was significantly more common in those given zinc across all age groups (RR 1.59, 95% CI 1.27 to 1.99 REM; 5189 children, 10 trials, 12 comparisons, Figure 7). There was significant heterogeneity among trials (P = 0.001, I
|Figure 7. Zinc vs placebo for acute diarrhoea: Adverse events (vomiting)|
One single trial on gastroenteritis due to rotavirus found no difference in time to resolution of vomiting between zinc and placebo (13.63 ± 10.33 hours vs 16.35 ± 11.34 hours, P = 0.1; Dalgic 2011)
Three trials reported on copper levels, with no significant differences between the zinc and placebo groups. Two studies reported the mean change in serum copper on the last day of supplementation (seven and 14 days after recovery): -1.1 ± 5.5 µmol/dL in the zinc group versus -1.5 ± 4.2 µmol/dL in the placebo group in one trial (Strand 2002), and -41.2 ± 418.8 µg/dL in the zinc group versus -79.4 ± 429.2 µg/dL in the placebo group in the second trial (Patel 2009). Mean serum copper after 14 days was 121 mg/L in zinc group versus 127 mg/L in the control in one trial (Bhatnagar 2004a),
1.8. Statistical heterogeneity
In the trials in children under six months no significant heterogeneity was detected in any outcome (Figure 3, Figure 6, Figure 7). One multicenter study further subgrouped by sex, breastfeeding practices, length for age, and age (under or above three months), and no significant differences were detected (Fischer Walker 2006).
We explored heterogeneity in the two groups of children > six months and children of all ages for the two outcomes mean diarrhoea duration and diarrhoea at day seven for which there were enough trials (Figure 8 and Figure 9). Each subgroup presented significant heterogeneity, and this indicates that no single feature could explain overall heterogeneity alone. In all the subgroups a significant effect of zinc over placebo was observed, with a few exceptions: 1) one trial in well-nourished children held in Poland, where there is a low risk of zinc deficiency, did not show any benefit of zinc (Patro 2010, 141 children, Figure 8, Figure 9); 2) one single study on zinc gluconate did not show a significant advantage for zinc over placebo (Bahl 2002, 805 children, Figure 8, Figure 9 ); 3) one recent factorial study (Patel 2009) reporting on zinc plus copper in India did not observe a benefit for the association of the two micronutrients but also for zinc alone (Figure 8 Figure 9).
|Figure 8. Forest plot of comparison: 2 Zinc vs placebo for mean acute diarrhoea duration: subgroup analysis excluding children < 6 months, outcome: 2.1 Diarrhoea duration (h).|
|Figure 9. Forest plot of comparison: 3 Zinc vs placebo for acute diarrhoea on day 7: subgroup analysis excluding children < 6 months, outcome: 3.1 Diarrhoea on day 7.|
We were unable to construct funnel plots to look for evidence of publication bias as none of the outcomes had sufficient numbers of trials to do this.
1.9. Sensitivity analysis
The sensitivity analysis against markers of methodological quality did not affect the direction of results. There was some loss of significance with diarrhoea duration, but overall the analysis did not change the point estimate of effects. The intention-to-treat analysis for worst-case/best-case scenarios did not altered the statistical significance of the results.
2. In children with persistent diarrhoea
All trials of persistent diarrhoea enrolled children aged over six months.
2.1. Diarrhoea duration
Zinc supplementation reduced the duration of persistent diarrhoea (MD -15.83 hours, 95% -25.43 to -6.24 hours; 529 children, five trials, Figure 10), with no evidence of heterogeneity.
|Figure 10. Zinc vs placebo for persistent diarrhoea: diarrhoea duration (h)|
2.2. Diarrhoea on days three, five, and seven
There was no evidence of a benefit with zinc in the one trial that reported on diarrhoea at days three ( Analysis 4.2) and five ( Analysis 4.3) (Penny 1999), and two trials that reported on diarrhoea at day seven ( Analysis 4.4) (Penny 1999; Khatun 2001).
2.3. Stool frequency
2.4. Stool output
Stool output was measured using different units at different time points, thus results could not be pooled together ( Table 3). Results are expressed as the arithmetic mean difference (AMD) or geometric mean ratio (GMR). Two trials reported on children aged greater than six months, with five comparisons, and only one (Khatun 2001) reported a significant reduction in cumulative stool output at day seven in the zinc group (AMD -338 mg/kg bodyweight, 95% CI -413.6 to -262.4 mg/kg bodyweight; P ≤ 0.001).
The only community trial reporting on hospitalization did not observe any hospitalizations in the zinc or placebo group (Penny 1999, 275 participants).
One trial reported one death in the zinc group compared to five deaths in the placebo group, out of 95 participants in each group (Roy 1998). Two trials did not observe deaths in any participants, irrespective of their allocated group (Penny 1999; Khatun 2001).
2.7. Adverse events
Four trials that reported on vomiting (505 children) showed no difference between the zinc and placebo groups ( Analysis 4.6); three of the trials reported no incidences of vomiting in either group. One trial using zinc 3 mg/kg/day for 14 days in moderately malnourished and severely malnourished children reported a significantly lower plasma copper levels in the zinc-treated group by the end of the second week of therapy (56.2 ± 17.8 µg/dL versus 72.7 ± 18.3 µg/dL, P = 0.02; Bhutta 1999b, 87 children).
2.8. Statistical heterogeneity
There was heterogeneity between two trials for diarrhoea at day seven. This may be explained by differences in the geographical regions (India and Peru) or to other factors not explored in the review. Reporting of vomiting was heterogeneous between trials, and this may be due to difference in the population or in the definition of event, or to reporting bias.
2.9. Sensitivity analysis
The sensitivity analyses did not affect the direction of results. There was some loss of significance with diarrhoea duration, but no changes in the point estimate of effects. An intention-to-treat analysis for worst-case/best-case scenarios did not alter the point estimate or the significance of results.
We identified 24 randomized controlled trials that compared zinc with placebo in young children. Nineteen trials evaluated acute diarrhoea and the other five evaluated persistent diarrhoea. Overall, zinc was effective for acute and persistent diarrhoea in children aged over six months ( Summary of findings for the main comparison, Summary of findings 2). Two large trials were conducted in children aged less than six months with acute diarrhoea, and they showed no evidence of an effect on any of the outcomes ( Summary of findings 3).
Zinc reduced the duration of acute diarrhoea. The size of the effect was clinically important particularly for diarrhoea at day seven, which is an indicator for the risk of persistent diarrhoea. This benefit withstood extensive subgroup analysis for nutritional status, geographic region, background zinc deficiency, zinc type, and study setting, with the exception of well-nourished children in country with low risk of zinc deficiency. Evidence on diarrhoea severity was less clear, as fewer trials reported on this, and different units and time points were used.
Zinc also reduced the duration of persistent diarrhoea, but evidence was inconsistent regarding the severity of persistent diarrhoea.
No firm conclusions regarding zinc's impact on hospitalization or death can be drawn from this review as trials were not designed to look at these outcomes, and most were conducted in hospital where death rates were low. Large community trials would be needed to explore whether zinc treatment for diarrhoea reduces hospitalization rates.
Treatment with zinc was associated with an increase in vomiting, although the reduction in diarrhoea duration seems to outweigh this. This increase was consistent across trials in all age groups, including one large trial with adequate allocation concealment that was designed to look at safety. However, the trial reported that vomiting was limited to one episode in most children and mainly occurred within 10 minutes of administration (Larson 2005). Zinc has a metallic after-taste, and development of a more palatable formulation may minimize this. There was no clear evidence of copper deficiency resulting from zinc supplementation in the regimens used.
In general, the methodological quality of the trials included in this review was good. Most trials were conducted in countries with moderate to high risk of zinc deficiency, with the only study held in a country with a low risk of zinc deficiency showing no effect of zinc over placebo.
Applicability of these results to countries is likely to depend on local zinc deficiency and other population characteristics, such as the degree of malnutrition. Nearly all trials were conducted in hospital where participants were likely to adhere to the intervention, although one large community trial also showed a benefit with zinc.
Our results agree with those of other systematic reviews of zinc for treating diarrhoea (Bhutta 2000b, Lukacik 2008, Patro 2008, Haider 2009), except for the new finding of no effect of zinc in children aged less than six months. This review adds several new trials, includes a more extensive subgroup analysis, and reports on diarrhoea at different time points, diarrhoea severity, and adverse events.
The results of this review in children over six months support the current WHO/UNICEF policy to give zinc to children with diarrhoea (WHO/UNICEF 2004).
Summary of main results
Twenty-four trials, enrolling 9128 children, met our inclusion criteria. The majority of the data is from Asia, from countries at high risk of zinc deficiency, and may not be applicable elsewhere.
There is currently not enough evidence from well conducted randomized controlled trials to be able to say whether zinc supplementation during acute diarrhoea reduces death or hospitalization (very low quality evidence).
In children aged greater than six months with acute diarrhoea, zinc supplementation may shorten the duration of diarrhoea by around 10 hours (low quality evidence), and probably reduces the number of children whose diarrhoea persists until day seven (moderate quality evidence). In children with signs of moderate malnutrition the effect appears greater, reducing the duration of diarrhoea by around 27 hours (high quality evidence).
Conversely, In children aged less than six months, the available evidence suggests zinc supplementation may have no impact (low quality evidence), or even increase the proportion of children whose diarrhoea persists until day seven (moderate quality evidence).
No trials reported serious adverse events, but zinc supplementation during acute diarrhoea causes vomiting in both age groups (high quality evidence).
In children with persistent diarrhoea, zinc supplementation probably shortens the duration of diarrhoea by around 16 hours (moderate quality evidence),
Overall completeness and applicability of evidence
Most trials were conducted in hospital where death rates were low, and were consequently not powered to detect an effect on mortality. Large community trials are needed to explore whether zinc treatment for diarrhoea reduces hospitalization and death.
Most trials were conducted in Asian countries with moderate to high risk of zinc deficiency and high rates of mild and moderate malnutrition. The two studies available from Africa have shown no benefits of zinc over placebo, but this may be affected by other characteristics of the population or of the study (one study enrolled only children under six months, the other enrolled only 60 children). Only one study has been conducted in well-nourished children in countries at low risk of zinc deficiency, showing, as expect, no impact of zinc. The applicability of the results of this systematic review to countries is therefore likely to depend on the local prevalence of zinc deficiency or other population characteristics, such as the prevalence of malnutrition.
Nearly all trials were conducted in hospital where participants are more likely to adhere to the intervention; however one large community trial also showed a benefit with zinc.
The observed increase in vomiting was consistent across trials in all age groups. One large trial with adequate allocation concealment that was designed to look at safety reports that vomiting was limited to one episode in most children and mainly occurred within 10 minutes of administration (Larson 2005). Zinc has a metallic after-taste, and development of a more palatable formulation may minimize this adverse effect.
Quality of the evidence
The quality of evidence has been assessed using the GRADE methodology and is displayed in three summary of findings tables: Summary of findings for the main comparison; Summary of findings 2; Summary of findings 3.
The evidence for benefits on diarrhoea duration in children aged > 6 months is of low to moderate quality. This implies that we can have some confidence in the results but further research may alter the estimates of benefit and harm. The main reasons to downgrade were 'indirectness' and 'inconsistency' in the results. Zinc is currently recommended in Asia and Africa but almost all the evidence is from Asia. There is sufficient doubt that micronutrient deficiencies may be different across settings to consider this evidence 'indirect' for application to Africa. In addition, heterogeneity or 'inconsistency' between trials is very high, with many trials failing to show any evidence of benefit. This is perhaps not surprising given the variations in populations, settings, and interventions. However we were unable to completely explain this heterogeneity through subgroup analysis, and so our confidence that zinc supplementation can be broadly applied is decreased.
Agreements and disagreements with other studies or reviews
Our results agree with those of other systematic reviews of zinc for treating diarrhoea (Bhutta 2000b, Lukacik 2008, Patro 2008, Haider 2009), except for the new finding of no effect of zinc in children aged less than six months. This review adds several new trials, includes a more extensive subgroup analysis, and reports on diarrhoea at different time points, diarrhoea severity, and adverse events.
Implications for practice
In areas where diarrhoea is an important cause of child mortality, and the prevalence of zinc deficiency or mild/moderate malnutrition is high, zinc may be of benefit in children aged six months or more.
Implications for research
Causes of heterogeneity in the effect of zinc in children over six months should be further explored, and further research is necessary to justify continued supplementation in children less than 6 months.
The editorial base for the Cochrane Infectious Diseases Group is funded by the UK Department for International Development (DFID) for the benefit of developing countries. We thank Katharine Jones and David Sinclair for their help in reviewing the text.
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: 20 February 2012.
Protocol first published: Issue 3, 2005
Review first published: Issue 3, 2008
Contributions of authors
Both authors contributed equally to the preparation of the review.
Declarations of interest
Sources of support
- No sources of support supplied
- Department for International Development (DFID), UK.
Differences between protocol and review
We used the GRADE profiler, version 3.2.2 to create ’Summary of findings’ tables for the primary outcomes in the review.
2007, Issue 4 (first review version)
We made the following modifications while conducting the review.
- Changed inclusion criteria for participant age to "children over one month old" (rather than "two months") to avoid arbitrarily losing trials.
- Moved death to a secondary outcome measure following feedback from referees.
- Stratified the results by age categories since we observed significant heterogeneity when trials were pooled, and a clear difference in zinc effect was evident according to age.
- For subgroup analysis by nutritional status, it was not possible to refer to the definition of malnutrition given in the protocol (weight/height) as most included trials used another definition (weight/age), which is easier to measure. The difference between the two definition is that the first identifies children with acute weight loss or 'wasted', while the second includes both children with acute and chronic malnutrition ('wasted' and 'stunted').
- Two categories of 'zinc dose' were used (20 mg and > 20 mg) as most trials used zinc 20 mg/day, and only two trials used more than 20 mg/day.
- Gender was added since subgroup as it was recently identified as a possible effect modifier (Garenne 2005).
Medical Subject Headings (MeSH)
Acute Disease; Age Factors; Developing Countries; Diarrhea [*drug therapy; mortality]; Diarrhea, Infantile [drug therapy; mortality]; Randomized Controlled Trials as Topic; Time Factors; Trace Elements [adverse effects; deficiency; *therapeutic use]; Zinc [adverse effects; deficiency; *therapeutic use]
MeSH check words
Child, Preschool; Humans; Infant
* Indicates the major publication for the study