A review and meta‐analysis of the impact of intestinal worms on child growth and nutrition

1.1 The gastrointestinal ecosystem

The human intestinal tract provides a protected habitat for several hundred species of viruses, bacte ria, yeasts, protozoa and worms. All the organisms that live in the intestinal ecosystem are parasites, because they are dependent for their existence on their host, and the basis of this dependence is usually nutritional (Hall 1985). The parasitic lifestyle is highly successful for worms in general mainly because, once established within a host, there are no predators and life is a sheltered steady state with a constant supply of nutrients that is sustained by the host's homeostatic mechanisms.

Because the gut is a cavity within the host, it is said to be immunologically privileged, as the organisms living there are not exposed to the full force of the human immune system. Nevertheless, intestinal worms do elicit an immune response, and hookworms and whipworms in particular come into contact with both the cellular and humoral immune systems to elicit a Th‐2 responses and cause a rise in the concentration of immunoglobulin E (Else 2005). But the fact that intestinal worms persist and are not expelled from the gut indicates that they are able to evade these immune responses, although the mechanisms by which they achieve this are unclear.

The only important non‐specific barrier to infection is hydrochloric acid, secreted as ions into the stomach lumen by the parietal cells in the gut wall. Although this acid can kill infectious stages of many potential pathogens, paradoxically it is also a necessary stimulus for the establishment of many parasites: exposure to acid is required for the excystment of Giardia duodenalis (Hautus et al. 1988; al‐Tukhi et al. 1991) and may be a necessary stimulus for the eggs of Ascaris lumbricoides to hatch, along with warmth and exposure to bile salts. Once in the intestine, the site where the worms come to maturity, all parasites of the gut have a body surface that is resistant to the action of the host's digestive enzymes, while some worms have developed specific antienzymes (Uglem & Just 1983), presumably for self‐protection. The role of antienzymes in causing malnutrition is putative rather than proven, and it seems most likely that they act locally to prevent damage to the worms’ surface by host enzymes, rather than being secreted to have a widespread effect in the gut and thus perhaps on human nutrition.

The infectious stages of parasites have an easy way to enter their host, usually through the mouth as a contaminant of food, water or fingers, while the next generation leaves the body in faeces through the anus in the form of spores, cysts, eggs or larvae. There are exceptions: a few parasites of the gut enter the body through the skin, notably the larvae of hookworms and Strongyloides stercoralis.

The major problem for parasitic worms is to get from one host to another, a journey that is facilitated in several different ways:

• • 

  by producing large numbers of infectious stages to increase the chances of infecting a new host; a fertilized female A. lumbricoides, for example, may produce up to 200 000 eggs a day (Sinniah 1982), therefore millions in a lifetime;

• • 

  by producing resistant infectious stages that can withstand adverse conditions; the eggs of species of Ascaris can survive for several months or years in warm, humid and sheltered conditions (Gaasenbeek & Borgsteede 1998), and are resistant even to 10% formalin (Sandars 1951) though not to exposure to ultra‐violet light or to desiccation (Crompton 1989);

• • 

  by infecting an intermediate host in which the parasite both multiplies and is dispersed, a feature of the life cycle of many trematodes, a group of flatworms whose species often reproduce in snails from which larval stages are released that are infectious to humans;

• • 

  by infecting or encysting on foods that are consumed by a new host (Fried et al. 2004);

• • 

  by the behaviour of infected people that puts others at risk of infection, such as defecating in the open, so that infectious stages are spread in the environment (Kilama 1989); and

• • 

  by taking advantage of the behaviours that put people at risk of infection, such as pica (Geissler et al. 1998); by using fresh human faeces (sometimes called ‘nightsoil’) as a fertilizer (Pan et al. 1954; Needham et al. 1998); and by poor personal hygiene.

As the gastrointestinal ecosystem offers such a rich habitat, it has been colonized by an enormous number of different species. The next section introduces the major species of worms that live in the human gastrointestinal tract.

1.2 Groups of intestinal parasitic worms that infect humans

The variety of general and specific names given to worms can be quite confusing to a novice (see Box 1). The terms ‘helminths’ and ‘worms’ are generic names for metazoan (multicellular) parasites that are classified by helminthologists into two main Phyla:

1.3 How worms may affect human nutrition and growth

There are several mechanisms by which intestinal nematodes could affect the nutritional status of their host:

• • 

  by feeding on the contents of the host's gut, including the host's secretions that make up the exoenteric circulation;

• • 

  by feeding on host tissues, including blood and serum, which leads to a loss of iron and protein;

• • 

  by causing maldigestion or malabsorption of nutrients;

• • 

  by inflammatory responses that lead to the production of substances that may affect appetite and food intake, or substances that modify the metabolism and storage of key nutrients such as iron; and

• • 

  through contingent responses to infection, such as fever, leading to an increased metabolic rate; by causing hypertrophy of muscles; and by immune responses to infection, all of which result in the diversion or use of nutrients and energy for purposes that would not have been necessary had worms not been present.

All intestinal parasites obtain their nutrients either from the food and intestinal secretions of their host, or from their host's tissues and body fluids. The nutritional needs of parasites are relatively small compared with a well‐nourished host, mainly because their relative biomass is small (see Box 2).

1.4 Design of studies to estimate the impact of worms

The main problem with studying the impact that worms have on child growth and nutrition has been touched upon in the previous section: the need for untreated controls. If worms impair growth, and if treating worms leads to extra or catch‐up growth, then it is necessary to measure the difference that treatment makes between treated and untreated subjects, not just the absolute amount of growth that occurs after treatment. This is because some amount of growth and weight gain should occur naturally in all children, unless they are severely undernourished or have a hormonal disease.

It could be argued that it is enough to express weight gain as a change in proportion to a reference value, such as a higher z‐score of weight‐for‐height, or a greater percentage of the median value. Such improvements in anthropometric status could occur as a result of secular changes in the food supply, or as a result of better health because of seasonality in the transmission of other diseases, such as malaria and diarrhoea. Concurrent and untreated controls are essential to the validity of the conclusions of any study of the impact of treating intestinal worms on child growth and nutritional status and are an important criterion for including any study in a meta‐analysis.

1.5 Aims

The aim of this review is:

• • 

  to describe the epidemiological factors that influence the impact that intestinal worms have on human nutritional status and growth;

• • 

  to describe the factors that affect the impact of anthelmintic treatment; and

• • 

  to undertake a meta‐analysis of the effects of intestinal worms on children's nutritional status and growth.

2.1 Species of intestinal worm

The most important species in terms of disease are A. lumbricoides, T. trichiura and the hookworms. These worms live in different parts of the intestine, differ in the route they take to reach their adult habitat and feed in different ways. This has been described in Section 1.2.

Although A. lumbricoides is undoubtedly the most common species worldwide, it is very hard to distinguish from A. suum (Crompton 1989). It is quite likely that both species occur together, especially in places where pigs are allowed to roam freely in their search for food in an environment inhabited by people (Kofie & Dipeolu 1983; Maruyama et al. 1997).

Although the two hookworm species are considered together because there is no easy way to tell them apart, there is evidence that A. duodenale is more pathogenic than N. americanus because it consumes more blood per worm (Roche & Layrisse 1966). As well as measurements of blood loss using radioactive isotopes (Roche & Layrisse 1966), there is epidemiological evidence from a study of school children in Pemba, a small Tanzanian island where both types of hookworms occur, that in schools where the prevalence of A. duodenale is high there may be more anaemia and iron deficiency than in schools where N. americanus is the predominant species (Albonico et al. 1998).

The conclusion is that different species will have different effects on the nutritional status and growth of children.

2.2 Prevalence of infection

The first important epidemiological parameter that describes the potential effect of worms on human health is the proportion infected, or prevalence. Infections are usually diagnosed by seeing the characteristic eggs of each species of worm in faeces examined under a microscope, which simply indicates that there is present in the gut at least one sexually mature female and one male worm. The exception is A. lumbricoides, because unfertilized female worms can produce infertile or ‘decorticated’ eggs; they can be identified because they are longer and narrower than fertilized eggs (Crompton 1989; WHO 1994a).

Infections with intestinal worms may therefore be missed if there are only female worms present, or only male worms, or only immature worms. Such infections are not clinically important, but they will lead to an underestimate of the prevalence.

Infections may also be missed if an insensitive method of diagnosis is used, such as a direct faecal smear, and if the concentration of eggs in faeces is low. A study in Bangladesh found that some 8% of infections with A. lumbricoides were missed when infections were diagnosed using a moderately sensitive ether sedimentation method (Hall 1981) and compared with a diagnosis made by expelling worms using an effective anthelmintic drug (Hall et al. 1999).

Table 2 shows the range in numbers of eggs estimated to be produced daily by a female worm of each species. They suggest that the sensitivity of diagnosing a light infection of a few worm varies between species, probably in the rank order A. lumbricoides, A. duodenale, T. trichiura and lastly N. americanus.

Infections with the three main types of intestinal worms number among the most common infections of children in the world today. Table 3 presents some recent estimates of the numbers of people infected with A. lumbricoides, T. trichiura and the hookworms, by age range. Table 4 presents recent estimates of the prevalence of these infections in young school‐age children, aged 5–9 years. This age group is particularly likely to have moderate to heavy infections and is vulnerable to their impact on nutritional status.

Figure 2 shows the typical relationship between age and the prevalence of infection with A. lumbricoides derived from a study of people living in an urban slum in Bangladesh; a similar relationship is commonly observed for T. trichiura. Figure 2 indicates that infections are acquired in the first 2 years of life and that around 80% of all age classes are infected, a high but typical proportion.

Hookworms tend to show a different pattern in which the prevalence typically increases with age, reaching a peak in late adolescence and adulthood (Bundy et al. 1992a). The reason for this difference is not clear, especially as children are less likely to wear shoes than adults and so could be considered to be more to be exposed to hookworm larvae on the soil. The differences in prevalence may reflect where worms are transmitted in what have been called ‘domains of infection’ (Cairncross et al. 1996). The transmission of both A. lumbricoides and T. trichiura is thought to occur within or around the household, in the domestic domain, while hookworms may be transmitted beyond the household, in the public domain which is frequented more by adults than children (Cairncross et al. 1996).

Although the prevalence of infection in children may provide some indication of the importance of each worm in terms of health and nutrition, there is no clear threshold prevalence associated with disease, largely because the relationship between prevalence and the intensity of infection is strikingly non‐linear. Figure 3 shows the relationship between the prevalence of infection with intestinal worms and the worm burden, derived from data collected in Bangladesh on A. lumbricoides (Hall et al. 1999). It can be seen that, below a prevalence of about 50%, the mean worm burden is relatively low, but rises almost exponentially above a prevalence of 60%. This means that even a prevalence of up to about 50% is associated with a low average worm burden, and that the prevalence of infection is a poor indicator of the probability of disease unless it is 70% or greater. This relationship helps to explain the use of a threshold prevalence of 50% for administering mass anthelmintic treatment for both soil‐transmitted helminths and schistosomiasis, a threshold that was endorsed by a World Health Organization (WHO) Expert Committee in 2001 (WHO 2002a). This threshold has subsequently been lowered by the WHO to 20% for mass treatment once a year in what they classify as ‘low‐risk’ communities, and the WHO now apply the 50% threshold to define a ‘high‐risk’ community where mass treatment twice a year is warranted (WHO 2006).

The relationship between prevalence and disease is further complicated by the mix of species: in many parts of the world it is common to find all three major types of worms together, so that some children have two or three infections. Although the prevalence of A. lumbricoides and T. trichiura is often correlated (Booth & Bundy 1992), it seems that an infection with one species does not predispose to the presence of another. Figure 4 shows a diagrammatic representation of the percentage of individuals who have multiple infections when the prevalence of A. lumbricoides is 60%, T. trichiura is 50% and the hookworms is 40%, all arbitrary but typical figures. If the probability of having one infection is independent of having another, then it can be estimated that 32% of individuals have at least two infections and 18% have all three infections (Fig. 4).

There is no method to assess the impact of multiple infections: they could be additive but might be multiplicative or even antagonistic if species occur in the same location in the gut, such as A. lumbricoides and the hookworms. As multiple infections may be as common as or more common than single infections, it is hard to estimate the relative benefit of treating each different species, especially as the drugs used to treat intestinal nematode worms are effective against all species, if to differing degrees (see Section 3.2).

In conclusion, the WHO threshold of 50% infection provides a reasonable basis for applying mass treatment. Below this threshold few people have moderate to heavy worm burdens that cause disease and the majority are uninfected; above this threshold the likelihood of moderate to heavy infections increases exponentially. This is not to say that worms do not cause disease below a prevalence of 50%, but the effect on a very small minority may be lost in the group average. A prevalence of 50% was taken as a minimum for studies of deworming to be included in the meta‐analysis reported below, and studies such as those of Garg et al. (2002) in which the prevalence of infection with any worm before treatment with an anthelmintic was only 11% were excluded. This threshold may seem somewhat arbitrary, but it is based on the relationship shown in Fig. 3 and it would be unusual to include any uninfected individuals in a trial of a drug to treat a bacterial diseases. A prevalence of 50–100% represents a typical range in many human communities in which mass anthelmintic treatment is given, and the effectiveness of treatment in such studies represent common epidemiological circumstances.

2.3 Number and distribution of worms

Each worm that becomes established in a host represents the successful hatching, migration, establishment and development of a single fertile egg. Nematode worms do not multiply within their host, except for S. stercoralis (see above) the eggs of which can hatch within the lower bowel to release infective filariform larvae that penetrate the gut wall in a process called autoinfection (Schad 1989). This explains the persistence of S. stercoralis in British soldiers who were prisoners of the Japanese in Asia during the second world war and who have developed strongyloidiasis in their old age (Gill & Bell 1979, 1987; Gill et al. 2004).

For each of the four major types of intestinal worms the probability of disease is related to the number of worms in the host, called the worm burden. Figure 5 shows the average number of A. lumbricoides recovered from males in 11 age classes in an urban slum in Bangladesh who were given a drug that paralysed their worms so that they were expelled by peristalsis, then recovered, washed and counted. It shows that the heaviest average infections were found in school‐age children from 5 to 15 years old, a characteristic typical of many helminth infections. The notable exception is the hookworms, for which the prevalence and mean worm burden tend to increase with age so that adolescents and adults tend to be most heavily infected (Bundy 1990; Bradley et al. 1992).

There are two theories to explain the shape of the distribution in Fig. 5. First, it could reflect differences in behaviour, because children are more exposed to worm eggs than adults, perhaps as a result of playing on faecally contaminated ground and poor personal hygiene.

Second, it could reflect the development of a partially effective immune response to infection, or perhaps a more effective immune response in some individuals than others as a result of repeated exposure to worm larvae. The study in Bangladesh however found a statistically significant difference in the mean worm burden of A. lumbricoides between adult males (who leave their community to work during the day) and adult females (who stay at home in the crowded, unsanitary environment), which suggests that exposure more than immunity influences this distribution (Hall et al. 1999).

When the distribution of worms among hosts is examined it is typical to find that it is highly skewed, so that most individuals have light infections while a minority have moderate to heavy infections. Figure 6 illustrates this distribution using data on the number of A. lumbricoides recovered from 1765 people who were treated with pyrantel pamoate to paralyse and expel their worms. The distribution shown in Fig. 6 is best described by the negative binomial, which is purely an empirical fit, and implies nothing about the biological reason why worms should be distributed in this way. This distribution, which is described as aggregated or overdispersed, is typical of most helminth infections and has been observed even in pigs each of which has been infected experimentally with the same number of fertile eggs of A. suum (Stephenson et al. 1980a; Boes et al. 1998). This suggests that the distribution derives from differences between hosts in the establishment of worms rather than in differences in exposure to eggs.

The equation that describes this distribution is defined by three parameters, the prevalence (P), the arithmetic mean worm burden (M), and k, a parameter that varies inversely with the degree of aggregation or clumping of worms in a few hosts, so that:

The parameter k captures the degree to which worms in a small proportion of hosts tend to be aggregated, clumped or overdispersed (terms commonly used in this context). Small values of k, typically less than 1, indicate a high degree of aggregation of worms, irrespective of age, sex or any other factor. The study in Bangladesh indicated that k varied with the mean worm burden so that, as the mean burden increased, the degree of aggregation decreased (Hall et al. 1999). The implication is that, when the average worm burden is large, more people are likely to have moderate to heavy infections and may be diseased as worms are less aggregated. Figure 7 shows the cumulative percentage of all worms recovered from 1765 people in urban Dhaka plotted against the cumulative percentage of hosts, for worm burdens between zero and 187 worms. It shows that a half of all subjects expelled only 10% of all worms, and that 80% of all subjects contained only 40% of all worms. The remaining 60% of worms were recovered from only 20% of individuals.

If it is typical to find that a small proportion of people harbour a large proportion of worms, what happens when they are treated? Evidence from studies of reinfection after treatment shows that heavily infected individuals tended to become heavily reinfected again while lightly infected people become lightly reinfected, leading to the theory that some individuals are predisposed to infection and others are not (Haswell‐Elkins et al. 1987; Bundy & Cooper 1988; Holland et al. 1989; Forrester et al. 1990; Chan et al. 1994; Kightlinger et al. 1995). If this is so, could efforts be concentrated on the individuals who were predisposed to worms? This was examined in a study of infection and reinfection with A. lumbricoides in 880 individuals in Bangladesh (Hall et al. 1992). It was found that, although there was evidence of a predisposition to moderate to heavy infections, over three rounds of treatment and two periods of reinfection of 6 months, about two‐thirds of all subjects were moderately or heavily infected at least once (Hall et al. 1992). This suggested that there was no benefit in identifying moderately to heavily infected people on the assumption that they were predisposed individuals, and mass treatment should be given rather than any form of selective treatment.

The biological implications of the aggregation of worms are that most individuals in a community at any one time have light infections, but a minority ranging from <1% to 40% will have moderate to heavy infections and are most likely to be diseased. If treatment is given periodically and reinfection occurs (which will happen because eggs can persist in the environment for many months, if not years), then over a period of 2–3 years a majority of individuals will be moderately to heavily infected at some point. This has implications for measuring the impact of mass treatment, as a minority will benefit more than the majority in the short term, but over a longer period of periodic treatment, an increasingly larger proportion will benefit.

A light infection probably has little effect on the nutritional status of a host, and the worm burden is the key indicator of the probability of disease. This requires that worms are expelled and counted, something that is difficult to do, and it destroys the worm burden at the same time, so that infected subjects can no longer be followed. So in most surveys of worms the concentration of eggs in faeces is used as an indicator of the intensity of infection. However, the perception of egg counts tends to be relative, so that what is seen to be a moderate egg count in one place could be considered to be high or even low in another. To provide some guidance on assessing egg counts a WHO Expert Committee in 1987 suggested threshold egg counts to classify light, moderate and heavy infections with A. lumbricoides and T. trichiura (WHO 1987). In 2002 a new Expert Committee added hookworm to the list although the previous committee had stated that egg counts for this worm could not be given because the ‘critical worm load differs locally depending on age, sex, iron intake and species of hookworm’ (WHO 2002a). These thresholds are shown in Table 5.

Since the Expert Meeting in 1987 it has been clearly shown for A. lumbricoides at least that the relationship between worm burden and egg count is both non‐linear and differs between worms in different countries (Hall & Holland 2000). For example, 20 worms was associated with around 1300 eggs per gram of faeces or 2300 egg g⁻¹ in two studies in Bangladesh, with 17 300 eggs g⁻¹ in Iran, with 22 000 eggs g⁻¹ in Nigeria and Madagascar, with 27 000 eggs g⁻¹ in Burma and with 44 500 eggs g⁻¹ in Mexico (Hall & Holland 2000). This indicates that there is no consistent relationship between egg counts and worm burdens that can be applied universally, for A. lumbricoides at least.

2.4 Duration of infection

One of the main factors besides the worm burden that is likely to contribute to the development of disease is the duration of infection, something that is not usually known. Table 6 shows the estimated life span of the major species of intestinal nematodes of humans.

But even if a worm such as A. lumbricoides can live for as long as 2 years and is then expelled from the gut when it dies, new worms are continually acquired so that people remain persistently infected with slowly changing numbers of worms. Figure 2 indicates that just over 60% of children aged 1–2 years in the study in Bangladesh were infected with A. lumbricoides and Fig. 5 indicates that they contained an average of about seven worms. As both the prevalence and mean worm burden are higher in older age groups, it suggests that most people are infected throughout their life.

2.5 Rate of reinfection

If the population of worms in a community of hosts is perturbed by giving mass treatment with an anthelmintic, reinfection can occur immediately and the number of worms will rebound to a similar number as before, a state called equilibrium. Without treatment there may be fluctuations in time in the numbers of worms in individuals as some die and others are gained, but the number of worms in the community seems to reach a relatively steady state, perhaps driven by factors such as sanitation, the contamination of the environment, behaviours that put people at risk and environmental conditions that favour the survival of infectious stages. Figure 8 shows the prevalence of infection with A. lumbricoides in 880 people at three rounds of treatment, 6 months apart and Fig. 9 shows the mean worm burden on the same occasions (Hall et al. 1992).

The prevalence, shown in Fig. 8, rebounded very quickly after treatments in the school‐age children, perhaps because the force of infection was greater in this age class, but was lower at each round in the four adult age classes. The mean worm burden shown in Fig. 9 was, however, lower at each round of treatment in all age classes except for the very youngest cohort, whose exposure to worm eggs probably increased as they became mobile and were increasingly exposed to worm eggs. 8, 9 show that periodic treatment may do little to perturb the prevalence of infection but, if given often enough, it can help to prevent reinfection with the same number of worms. In an environment such as an urban slum in Bangladesh, it may be necessary to give treatment for worms at least three times a year rather than twice annually, in order to sustain low worm burdens and reduce the probability of disease.

8, 9 also illustrate that the prevalence is a poor indicator of the effectiveness of worm control by periodic deworming, and that an indicator of the intensity of infection is better (Bundy et al. 1992a). Ideally both should be measured during a programme of mass deworming.

2.6 Summary

The probability that an infected child has disease or malnutrition caused by intestinal worms is related to:

• • 

  the species of worm;

• • 

  the mixture of species;

• • 

  the worm burden of each species; and

• • 

  the duration of infection before treatment.

The number of worms is difficult to assess without expelling them from the gut. The concentration of eggs in faeces may only represent a worm burden in a specific locality, because fecundity varies with the number of worms and perhaps between strains of worms in different locations around the world. There are no threshold numbers of worms to classify burdens as light, moderate or heavy. There are no methods to combine the probability of disease because of different numbers of worms of different species in the same individual.

3.1 Study design: controls and randomization

In order to estimate reliably the magnitude and statistical significance of the effect of a treatment on growth or nutritional outcome, it is necessary to have an untreated control group (Stephenson 1987). To ensure that naturally occurring differences between subjects are evenly distributed between groups before treatment is given, individuals also need to be randomly assigned to each study group. As there can be large differences between individuals in the intensity of infection because of the aggregated distribution of worms, if the sample size is small then subjects may need to be stratified first by egg count, before random allocation. If the sample size is large and the prevalence is high, then random allocation is likely to be sufficient.

If the treatments are allocated by clusters, such as villages or schools rather than to individuals, there must be a sufficient number of clusters to distribute any variation between clusters in the prevalence and intensity of infections evenly between study groups. Randomized cluster designs have the potential to provide the large sample sizes that are needed to be able to detect the effects of treatment on the minority of subjects who will benefit most from treatment.

Ideally all subjects in a control group in a simple randomized trial should be given an identical placebo. This is less easy to do in cluster randomized trials, especially if anthelmintics are given through the health system. Effectiveness trials, such as the one carried out as a part of Child Health Fairs in Uganda (Alderman et al. 2006), typically have untreated controls, but they usually receive nothing rather than a placebo. This opens the study to potential bias because of self‐treatment with anthelmintics among the control group.

3.2 Anthelmintic drugs and other treatments

The drugs used to treat infections with intestinal nematodes can be divided into two main types:

• • 

  drugs that act on the nervous system to paralyse worms so that they are expelled from the gut by normal peristalsis, such as piperazine, levamisole and pyrantel pamoate; and

• • 

  drugs that inhibit metabolic processes, such as the benzimidazole derivatives albendazole, mebendazole and tiabendazole, which block the uptake of glucose by microtubules in the mitochondria of worms (Horton 2000, 2002), and the relatively new treatment nitazoxanide, which acts in protozoa by inhibiting the enzyme pyruvate ferredoxin oxidoreductase (Sisson et al. 2002; White 2003; Esposito et al. 2005).

There are differences between these types of drug in the efficacy with which they treat each species of intestinal worm (de Silva et al. 1997). Efficacy is assessed in two ways:

• • 

  as the cure rate, which is the percentage of infected subjects in whose faeces worm eggs are no longer found after treatment (although some immature female or male worms could still remain, thereby overestimating the cure rate); and

• • 

  as the egg reduction rate, which is the percentage reduction in the arithmetic or geometric mean concentration of eggs in the faeces of infected individuals. The concentration of eggs reflects the number of female worms, and an even sex ratio is usually assumed.

The cure rate and egg reduction rate are typically estimated by collecting and examining under a microscope a faecal sample collected a few days before treatment and then again some 14–21 days afterwards. The gap after treatment is advisable because there is evidence that the drugs may temporarily inhibit egg production by worms that have survived treatment (Hall & Nahar 1994).

Because of their different modes of action the two types of anthelmintic drugs differ in their efficacy: drugs that paralyse worms tend to be less effective than those that inhibit metabolic processes, perhaps because the paralysis can wear off more quickly than the effects of a metabolic inhibitor, especially if worms are anchored to the gut wall. (If A. lumbricoides are expelled from the gut using a drug such as pyrantel or levamisole and placed in warm saline, they can recover their motility after a few hours, showing the transitory effect of the paralysis).

Table 7 gives a rough guide to the range in efficacy of each type of drug against the main species of intestinal nematode worms at the usual dosage given. Data are generally lacking for the drugs that paralyse worms, because they are older and less well studied than the highly effective benzimidazole drugs, mebendazole and albendazole, which superseded levamisole and pyrantel in 1975 and 1980 respectively (Horton 2003b). In general the benzimidazoles are less effective against T. trichiura than against A. lumbricoides and the hookworms, and are less effective against hookworm than against A. lumbricoides.

There also seem to be differences in the efficacy of the two benzimidazole anthelmintics, even though they are chemically similar. Table 8 shows data from a large study in Zanzibar of treating intestinal nematode infections with albendazole or mebendazole, which revealed statistically significant differences in the efficacy of treating hookworm (Albonico et al. 1994).

A reason for the lower efficacy of drugs to treat hookworm compared with A. lumbricoides may be that hookworms are usually physically attached to the gut and may be better able to resist the transient effects of anthelmintic drugs than A. lumbricoides which can only maintain their position in the gut by actively swimming against the intestinal flow.

The reason for the lower efficacy of T. trichiura compared with both A. lumbricoides and the hookworms may be that they live in the large intestine, where anthelmintic drugs may be more dilute than in the small intestine. Adult T. trichiura are also typically embedded in the gut wall by their whip‐like anterior end, which may provide an anchor so that they, like hookworms, can withstand any transitory effects of anthelmintic drugs. A study in Bangladesh showed that, in order to achieve a cure rate for T. trichiura of 80%, 400 mg of albendazole had to be given daily for three consecutive days (Hall & Nahar 1994).

There is also evidence that there may be a lower cure rate for T. trichiura in Asia than in Africa: an analysis of trials of albendazole showed a median cure rate of 33% in Asia compared with 61% in Africa (Bennett 2000).

There is obvious concern that repeated mass treatment could lead to the development of drug resistance, something that has been shown to develop as a result of repeated treatment (Albonico et al. 2003). There are substantial lessons to be learned from veterinary medicine in which the same anthelmintic drugs have been used for many years and far more intensively than in humans (Coles 1995). Strategies to avoid or delay the development of drug resistance should be applied (WHO 2002a), such as alternating drugs with different modes of action, and the efficacy of treatment should be assessed periodically at sentinel sites during mass treatment programmes, the interval to depend on the frequency of treatment (WHO 2002a).

The main limitation to alternating treatments has been the lack of effective alternatives to the albendazole and mebendazole, as single‐dose treatments. One possibility is a combination of pyrantel and oxantel which has been shown to be quite effective in comparison with mebendazole (Rim et al. 1975; Albonico et al. 2002), but this mixture is not widely available. A combination of albendazole and ivermectin has been tested as a treatment for lymphatic filariasis (1999a, 1999b; Horton et al. 2000; Belizario et al. 2003) and has effects on the major species of soil‐transmitted helminths as well. The advantage of giving two drugs in combination is that the chance of genes that confer resistance to both treatments at the same time is greatly reduced.

A relatively new drug called nitazoxanide has been shown to be an effective treatment for all major species of intestinal nematode worms as well as Hymenolepis spp. (Romero Cabello et al. 1997; Juan et al. 2002). The drug currently has to be given twice a day for 3 days whereas the others are single‐dose treatments, which ensures compliance in mass treatment campaigns. The major advantage of nitazoxanide is that it is an effective treatment for intestinal protozoa, including G. duodenalis and Cryptosporidium spp., as well as anaerobic bacteria such as Helicobacter pylori and Clostridium difficile (Megraud et al. 1998; Bobak 2006). Such a drug could have a significant impact on growth and nutritional status, as studies of giving metronidazole, another broad‐spectrum antibacterial and antiprotozoal drug have suggested (Gupta & Urrutia 1982; Khin Maung et al. 1990).

The differences in the efficacy of treatments for intestinal nematode worms will therefore depend on the drug, the dose of drug, the species of worm and the strain of the worm perhaps, which will all have consequences for the impact of treatment on nutritional status and growth. In some studies included in this review a single treatment was given, while in others repeated treatment was given using the same drug. In one study included in the review, two different drugs were given: pyrantel pamoate at the first treatment and then mebendazole for subsequent bimonthly treatments (Northrop‐Clewes et al. 2001).

The situation is further complicated if anthelmintic drugs to treat other types of worms are given, such as praziquantel or metrifonate to treat infections with Schistosoma species. This is a genus of blood flukes found in Africa, the Middle East, Asia and South America, often concurrently with infections with intestinal nematode worms. Schistosoma haematobium causes bleeding into the urinary bladder as a result of the passage of the sharp‐spined eggs through tissues, while the passage of the eggs of S. mansoni and S. japonicum through the gut wall also causes internal bleeding. Studies of the impact of treatment in areas in which schistosomes occur tend to give a drug to treat these infections, as well as a drug to treat intestinal nematode worms (Beasley et al. 1999; Friis et al. 2003). In addition to treating schistosomes, the drug metrifonate is also an effective treatment for hookworm (Kurz et al. 1986), which complicates studies and measuring outcomes resulting from treatment.

Finally, if any study is to assess the impact of treating intestinal worms alone, then additional treatments cannot be given unless a factorial design is used and an untreated control group is provided. One study of the effect of treating worms with mebendazole also gave all subjects metronidazole as well, a broad‐spectrum antibacterial drug that also treats infections with intestinal protozoa (Marinho et al. 1991). This meant that the effect of mebendazole could not be separated from the effect of metronidazole. In other studies supplements of micronutrients or food were also provided, which also may have had independent effects on nutritional status and growth so making it impossible to estimate the impact of deworming alone (Latham et al. 1990a; Majumdar et al. 2003; Lind et al. 2004).

3.3 Intervals between treatments

In an infected human population the death rate of worms tends to a reach a balance with the rate at which worms are acquired, so an equilibrium develops. The effect of mass treatment is to cure a proportion of people, which reduces the prevalence of infection, and treatment expels a proportion of the total worm population, which reduces the mean worm burden. As there is no fully protective immunity to worms, reinfection can occur immediately, especially as infective eggs can survive in the environment for months in suitable conditions. The rate of reinfection is strongly related to the prior prevalence and mean worm burden, as well as to local sanitation, which influences the number of infective stages in the environment. Studies of reinfection after treatment and mathematical models that simulate treatment and reinfection have shown that the prevalence of infection can rebound within a few months after treatment, but the worm burden takes considerably longer to reach the pre‐existing number (Hall et al. 1992). This means that the prevalence may show only a small difference between rounds of treatment, but the mean worm burden may decline.

Figure 10 shows data on the prevalence of infection with A. lumbricoides among 445 school‐age children in Bangladesh at the first treatment and after two more rounds of treatment, each 6 months apart. The dotted line shows the trend which, when extrapolated, suggests only a small decline over two future rounds of treatment. Figure 10 also shows the mean worm burden at each round of treatment, which declines by almost 60% between the first and third treatment. If the linear decline was extrapolated it would lead to a substantial reduction in mean worm burden after five rounds of treatment. The relatively slow decline in worm burden achieved in the study in Bangladesh suggests that treatment would actually be best given every 3–4 months or three to four times a year rather than twice a year.

Figure 10 makes the point that the interval between treatments will influence the average intensity of reinfection that occurs, assuming a high efficacy, especially in the absence of any measures to prevent exposure to infectious stages in the environment or if only a proportion of the population is treated. Ideally the intervals between treatment should prevent the prevalence rising above 50%, the threshold at which the probability of moderate to heavy infections begins to increase exponentially (see Fig. 3).

The conclusion is that the frequency of treatment is therefore likely to affect the impact of treatment, and the aim should be to prevent moderate to large worm burdens from being reacquired. The consequence for this review is that it is difficult to compare studies of different numbers of treatments given over different periods. Nevertheless it can be useful to standardize gains in weight or any other parameter per unit of time, such as per year (Alderman et al. 2006). But, if reinfection is rapid, then two or even three treatments a year may have a greater impact than one, assuming that worms are the only constraint on growth.

3.4 Duration of follow‐up

It follows from the discussion in the previous section that, as well as the number of treatments given, the duration of follow‐up after treatment will affect the magnitude of any difference between a treated and an untreated control group, and therefore its statistical significance. This is illustrated in Fig. 11 which shows hypothetical data on the weight of two cohorts of undernourished children, 300 treated and 300 untreated, over a period of 3 years during which the treated group were kept almost free of worms. The children were 5 years old at the start of the study and were underweight, with an average body weight that was 2 SD below the National Center for Health Statistics median. If the periodic treatment of worms led to an extra gain in weight of 0.5 kg a year, or 3.5% of the initial mean weight, then it can be seen from the confidence intervals around the means in Fig. 11 that it is not until the third year that a statistically significant difference between groups is achieved.

The duration of follow‐up is important because, as Section 2.4 has shown, a relatively small proportion of infected subjects will benefit from treatment so it is likely to take some time before the effects of repeated treatment are detectable in comparison with an untreated infected group. Some studies have attempted to measure a difference between groups in weight gain or growth over a very short period, such as 3 or 7 weeks after treatment (Hadju et al. 1996). Even though there was a statistically significant difference between study groups in such studies, it is too short a period to warrant inclusion in the review. Another study attempted to measure a change in haemoglobin and iron status over a period of 10 days after treatment (Karyadi et al. 1996), which is also a very short time in which to expect an improvement.

3.5 Outcomes measured and the need for controls

The primary outcomes usually measured during studies of the effect of anthelmintic treatment are changes in body weight, height, mid‐upper arm circumference and skinfold thickness. A change in haemoglobin concentration may also be expected if hookworm or T. trichiura occur. Measuring vitamin A status is hard to do as the concentration of retinol in serum is usually sustained from liver stores, so some sort of dose response test is usually used. Secondary outcome measures are anthropometric indices such as height‐for‐age and weight‐for‐age, which both require age to be known, ideally to within a month, and weight‐for‐height or body mass index. Indices of weight‐for‐height can only be calculated for girls younger than 10 years and boys younger than 11.5 years if the National Center for Health Statistics reference values are used (WHO 1983), which may limit their application in studies of school‐age children. Z‐scores of anthropometric indices are useful when controlling for the initial nutritional status of subjects based on the assumption that, when there is a large initial deficit, the impact of treatment is likely to be greater than if there is a small initial growth deficit. A change in z‐score can also help to give a perspective to the magnitude and nutritional significance of an improvement.

It cannot be assumed that a gain in body weight is due totally to an increase in lean body mass, because some increase could result from the deposition of fat in adipose tissue. Although there are now available weighing scales that can estimate the percentage body fat by applying standard equations to measurements of electrical impedance, they cannot usually be applied to young children, and may not apply to undernourished children in a developing country either.

Measurements of mid‐upper arm circumference and triceps skinfold thickness tend to be under‐used in studies of child growth after deworming. Their value lies in the ability to estimate the surface area of subcutaneous fat and muscle in the arm, assuming a standard area of bone (Gibson 1990) which could indicate both growth in lean tissue and the deposition of fat reserves.

A statistically significant weight gain can be achieved over a relatively short period, especially if food is available ad libitum, while a gain in height can take relatively longer. A period of about 12 weeks seems to be about the minimum period of follow‐up. A study of vitamin A and iron given to Tanzanian children after treatment with albendazole and praziquantel showed statistically significant extra gains in weight and height after 3 months of supplementation (Mwanri et al. 2000).

Studies of the effects of treating worms that cause blood loss require controls because the haemoglobin concentration can fluctuate naturally. Studies of giving iron supplements to school‐age children after deworming have shown statistical significance only because the mean value of the control group may go down in comparison with the treated children whose haemoglobin was sustained by the iron (Aguayo 2000; Hall et al. 2002; Roschnik et al. 2004). Natural fluctuations in haemoglobin may occur as a result of changes in the diet and diseases such as malaria, both or which may be seasonal.

3.6 Initial nutritional status

The impact of treatment on any indicator of nutritional status is likely to be related to the prior degree of undernutrition or deficiency, thus the margin for potential improvement. This was not the case in a study in South Africa of the effect of anthelmintic treatment and micronutrient‐fortified biscuits (Jinabhai et al. 2001b). One of the outcome measures was anthropometric status, but at the start of the study only 7.3% of children were stunted and 0.8% were underweight (Jinabhai et al. 2001b), not a large margin for improvement.

It may seem obvious, but it is also important that treatment and control groups should be similar before any treatment is given. A study of the impact of treatment with mebendazole given to young children as a part of an evaluation of the Integrated Management of Childhood Illness in western Kenya reported statistically significant differences in gains in weight, height and weight‐for‐age when infected and treated children were compared with infected and untreated controls (Garg et al. 2002). In this study the treated children were significantly more undernourished than the control group: the z‐score of weight‐for‐height of the group given the placebo was much greater than the treated children (treatment −0.89 versus control −0.19, P < 0.001) and there was a statistically significant difference in their initial weight‐for‐age as well (−1.43 versus −1.01, P < 0.001) (Garg et al. 2002). If the infected and treated children gained more weight it could easily have been because they started from a lower initial nutritional status and had a greater potential to achieve more catch‐up growth. Both the treatment and control groups need to be similar if the impact of treatment on growth or most nutritional variables is to be compared reliably.

3.7 Age of subjects

The age of subjects will influence the measurement of the impact of anthelmintic treatment. This is because growth is not linear during childhood and shows two spurts, the first during the first 2 years of life, and the second during adolescence. Although the prevalence and intensity of worm infections tends not to be as high during the first 2 years of life as during the school‐age years, the potential impact of worms early in life is greater, mainly because of the relative size of worms to their hosts, and the relative magnitude of any extra weight gain achieved after treatment. An extra weight gain of 500 g over a year by a 5‐year‐old girl whose weight is 2 SD below the mean is 3.6% of initial weight; for a 10‐year‐old it is 2.3% and for a 15‐year‐old it is 1.3%. Such an extra weight gain is of decreasing biological significance, unless it can be sustained each year, and will require a larger sample size to detect in the older age groups.

3.8 Remedial therapy after treatment

An implicit assumption of many randomized trials of anthelmintic treatment seems to be that removing worms will automatically lead to improvements in growth and nutritional status. This may happen, but only if the diet is adequate (Hall 2007). If worms have impaired growth, the haemoglobin concentration or micronutrient status, their effects are most likely to have been due to the mechanisms outlined in Section 1.3 which are worth repeating here: by feeding on the host's food, secretions, tissues and blood; by causing maldigestion or malabsorption; by effects on appetite and food intake; and by causing responses to infection that consume or divert resources unnecessarily. The losses or deficits caused by worms cannot be rectified or remedied simply by removing the worms alone, although halting any pathological processes is an important first step. After worms have been killed the need is for remedial treatment of the underlying nutritional deficits by providing energy, protein and micronutrients, so that catch‐up growth can be achieved. This is illustrated in Fig. 12 and discussed in some detail in Section 6.2.

Ideally, children should be protected from moderate to heavy infections throughout their childhood by repeated treatment to keep worm burdens low, by primary measures such as sanitation to prevent faeces containing worm eggs from getting into the environment, and by secondary measures to prevent exposure to worm eggs, such as personal hygiene.

3.9 Summary

There are a number of factors related to anthelmintic drugs and study design that need to be considered when either designing studies or evaluating papers that report studies.

• • 

  Different drugs give a different cure rate and egg reduction rate for each species of intestinal nematode.

• • 

  All anthelmintics are highly effective against A. lumbricoides.

• • 

  Albendazole 400 mg and mebendazole 500 mg are the best treatments available for hookworms and T. trichiura.

• • 

  At these dosages albendazole is more effective than mebendazole against hookworm infections, but both achieve egg reduction rates of 55–95%.

• • 

  Neither albendazole nor mebendazole given as a single dose cures many infections with T. trichiura, but such treatment can reduce egg counts by 50%.

• • 

  Albendazole and mebendazole may be a less effective treatment for T. trichiura in Asia than in Africa.

• • 

  Three daily doses of albendazole or mebendazole may be needed to achieve high cure rates for T. trichiura.

• • 

  Reinfection with worms can occur immediately after treatment, so treatment needs to be given periodically.

• • 

  The interval between treatments may influence the impact on growth and nutritional status.

• • 

  The subjects for study should be malnourished as well as infected.

• • 

  As a minority of children (ranging from 10% to 40%) have moderate to heavy infections, not all will benefit to the same degree from treatment, and the impact on the average of any outcome measure may be diluted.

• • 

  These factors should be taken into account so that the period of follow‐up after treatment should be sufficient to be able to detect a statistically significant effect.

• • 

  It cannot be expected that nutritional status will improve and show catch‐up if the diet of subjects is inadequate to meet the extra requirements.

4.1 Search terms

Table 9 shows the terms used in combination to search Medline. The search was limited to papers on humans in the following groups: infant, pre‐school, child and adolescent. Only papers in English were located. In addition, the reference list of each paper identified was scanned for other papers, so too was the Cochrane Review of 2000 (2000a, 2005) and the book by Stephenson (1987) Impact of Helminth Infections on Human Nutrition. One unpublished paper known to the principal author, who was a co‐investigator, was included (Partnership for Child Development 2001).

All papers identified were summarized under the following headings:

• • 

  Design. A summary of the study design, including whether placebo controlled and blinded or not.

• • 

  Follow‐up. The period children were studied after treatment, or the difference in time between the initial, interim and final measurements.

• • 

  Location. The region and country where the study was carried out.

• • 

  Age range.

• • 

  Infection prevalence at baseline. The prevalence of each species of worm recorded.

• • 

  Treatments. The drugs used and the doses given.

• • 

  Sample size. The number of subjects studied in each group.

• • 

  Outcomes. The principal outcome measures and any derived indices.

• • 

  Findings. A summary of the major findings with standard deviations and the statistical significance of any differences between groups.

• • 

  Notes. A summary of the prevalence of each intestinal nematode infection; a summary of the effect of treatment of worms; a summary of the degree of undernutrition estimated as z‐scores, the mean haemoglobin concentration or the prevalence of anaemia; and short notes on other major or different features of the study.

• • 

  Included in Cochrane review of 2000 (recorded as Yes or No):

• 12

  Dickson R, Awasthi S, Demellweek C, Williamson P. Anthelmintic drugs for treating worms in children: effects on growth and cognitive performance (Review). Cochrane Database of Systematic Reviews 2000, Issue 2. Art. No. CD000371. doi:10.1002/14651858.CD000371

• • 

  Included in meta‐analysis. Either Yes, or the reason the study was excluded is given.

4.2 Inclusion criteria

Papers were included in the analysis if the following criteria were met:

• • 

  Studies of treating children and adolescents (aged 1–19 years) for intestinal nematode worm infections with albendazole, levamisole, mebendazole, piperazine or pyrantel (pamoate or embonate).

• • 

  An experimental group given the anthelmintic alone.

• • 

  An untreated control group.

• • 

  Random allocation to treatment and control groups.

• • 

  An initial prevalence of infection in children in the study locality with any species of intestinal nematode worm of ≥50%, the threshold used by the WHO to classify a ‘high‐risk’ community (WHO 2006).

• • 

  A duration of follow‐up of at least 12 weeks for all outcomes except the modified relative dose response test to an oral dose of retinol (Tanumihardjo et al. 1996a).

• • 

  Outcomes measured included at least one primary outcome variable among weight, height, mid‐upper arm circumference, skinfold thickness, haemoglobin concentration, and a measure of serum retinol; or secondary outcomes based on weight‐for‐age, height‐for‐age or weight‐for‐height such as z‐scores or percentages of the median.

• • 

  Data given as a mean with standard deviation and sample size, or enough data to calculate the standard deviation.

Data from the study of Koroma et al. (1996) were entered separately for urban and rural children.

4.3 Exclusion criteria

Papers were excluded from the meta‐analysis for one or more of the following reasons (the references cited are examples, and the list is not comprehensive):

• • 

  Data on an outcome measurement not presented in a form that could be used, for example, change in percentage weight‐for‐height (Hadidjaja et al. 1998), difference in percentage improved (Kloetzel et al. 1982), change in percentage with anthropometric indices less than −2 z‐scores (Lai et al. 1995), weight gain presented as a graph (Sur et al. 2005), no data given for stated outcome measures for treated children (Olds et al. 1999); or data are unique so could not be combined with other studies (Tanumihardjo et al. 1996b).

• • 

  No standard deviations of outcome measures given (Willett et al. 1979; Greenberg et al. 1981; Michaelsen 1985; Rousham & Mascie‐Taylor 1994; 1999a, 1999b; Fox et al. 2005) so not amenable to meta‐analysis using Review Manager 4.2 software.

• • 

  The initial prevalence of worms was less than the ‘high risk’ threshold of 50% (Freij et al. 1979; Donnen et al. 1998; Awasthi et al. 2000; Awasthi & Pande 2001; Garg et al. 2002).

• • 

  Inadequate control group or method of allocation, for example, infected subjects treated and compared with subjects who were lightly infected or uninfected (Callender et al. 1994; Mahendra Raj 1998; Mahendra Raj & Naing 1998; Bhargava et al. 2003); a sample size of two, one village where children were treated and another control village where they were not (Fernando et al. 1983); or no untreated controls (Forrester et al. 1998).

• • 

  Treatment given for other conditions in addition to treatment for intestinal nematode worms, for example, metrifonate, which treats hookworm as well as Schistosoma spp. (1985, 1989a, 1989b); praziquantel to treat Schistosoma spp. combined with a drug to treat intestinal nematode worms (Kruger et al. 1996; Jalal et al. 1998; Beasley et al. 1999; Jinabhai et al. 2001a); or food or micronutrient supplements given in addition to anthelmintic treatment (Cerf et al. 1981; Marinho et al. 1991; Kruger et al. 1996; Palupi et al. 1997; Jinabhai et al. 2001b; Mwaniki et al. 2002).

• • 

  Short period of follow‐up, for example, 3 and 7 weeks (Hadju et al. 1996), 10 days (Karyadi et al. 1996).

4.4 Meta‐analysis

Data from suitable papers were extracted and entered manually into RevMan version 4.2.8 for Windows. ²

Statistical comparisons were made between subjects treated with an anthelmintic drug and untreated controls for the outcomes listed in Table 13. For some studies the standard deviations had to be calculated from 95% confidence intervals and the sample size. For studies in which the mean and standard deviations of values before and after were given, the difference between mean values was used in the analysis with a pooled estimate of the standard deviation.

The weighted mean difference between treatment and control groups was calculated with 95% confidence intervals in RevMan 4.2 software (RevMan 2003) for each variable using a fixed effects model. Studies were disaggregated into subcategories by treatment except that studies in which mebendazole and albendazole were given were combined, as the two drugs are chemically similar and have the same mode of action.

5.1 Geographic origin of studies

Table 11 shows the countries represented in the papers found and used in this review. Of all 58 papers found, nearly 50% reported data from Africa. Of the 19 papers used in the analysis, a little more than 50% reported data from Africa.

5.2 Estimates of effects

Table 12 presents summary data on the impact of treatment on 12 variables for all studies in which intestinal worms were treated with any effective anthelmintic. The drugs used were albendazole and mebendazole, levamisole, piperazine or pyrantel pamoate. There were statistically significant differences for most variables related to growth; but the outcomes for micronutrients were not statistically significant.

Table 13 presents summary data on the impact of treatment on 12 variables for all studies in which intestinal worms were treated with albendazole or mebendazole only, as these drugs are now the most widely used and most broadly effective anthelmintics.

5.3 The figures and how to interpret them

13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 show the mean differences between treated and control groups for each study grouped vertically by the drugs used. The mean is shown as a box around the vertical line of zero difference, with 95% confidence intervals shown as bars. The diamonds represent the weighted mean difference between the treated and control groups for each treatment group and in total. The percentage weight given to each study, as well as the numerical value of the estimated difference with 95% confidence intervals, is shown. It should be noted that the scale in all figures except for 22, 25 and 25 is from −4 (favours control) to 4 (favours treatment).

5.4 Sources of error or bias

The decision to exclude studies in which the initial prevalence of infection is <50% is somewhat arbitrary, but is based on the WHO threshold which classifies them as ‘low‐risk’ communities (WHO 2006). Ideally all children taking part in studies of anthelmintic treatment would be infected with at least one species of worm, just as any drug trial would be conducted only on infected individuals. But most trials of the impact of anthelmintics on child growth tend to study groups of children rather than infected subjects only, so tend to be trials of effectiveness rather than efficacy. This type of study is driven also because of ethical problems with having infected but untreated children as controls, so studies tend to be carried out in communities in which the prevalence is known, and whether individual subjects are infected or not is usually unknown. Applying a threshold prevalence of 50% or more means that a half of all subjects or fewer will not benefit from anthelmintic treatment, and their inclusion in the study sample will to some degree dilute the impact of treatment on the infected majority. The threshold prevalence of 50% does not serve to maximize the impact of treatment, a prevalence of 100% is needed to do that, but a prevalence higher than 50% serves to reduce the effect of uninfected subjects. This means that the conclusions of the meta‐analysis can only be applied to circumstances in which the prevalence is more than 50% and excludes what the WHO calls ‘low‐risk’ communities, in which the prevalence of infection with any species of worm lies between 20% and 49% (WHO 2006).

About a half of all papers found and used in the analysis reported data from Africa. All of the papers reporting African data were from sub‐Saharan Africa, reflecting concern for the effects of worms in African children.

In many parts of sub‐Saharan Africa children are concurrently infected with species of Schistosoma, which may cause loss of appetite, internal bleeding, and tissue damage and inflammation resulting from reactions to the eggs of worms when they become lodged in tissues. Studies were excluded from this analysis in which subjects were also treated with praziquantel or metrifonate, the drugs most commonly used to treat Schistosoma spp., as this would overestimate the effect of treating intestinal nematodes. This could, however, have led to a bias against studies from sub‐Saharan Africa. But this did not seem to be the case, as a larger proportion of all papers from Africa were included in the final analysis than in Africa, Asia or the Indian subcontinent. Of the papers found and the papers used, the proportion from Africa was similar in both categories (28/59 or 47% and 11/24 or 46%).

6.2 Treatment alone is not enough

Whatever extra gains in nutritional or anthropometric outcomes are achieved after treatment with an anthelmintic drug, they do not occur solely as a result of that treatment. Catch‐up growth, for example, requires extra nutrients and energy, and if they are not available after treatment then growth rates are likely to remain unchanged. Alternatively a statistically significant difference between a treatment and control group could occur during a study because an untreated control group remain persistently diseased as a result of their infections, and thus gain less weight than the treated group.

The lack of extra food to meet the needs for growth could explain the lack of impact of deworming in many studies, such as the large cluster trial in Vietnam (Partnership for Child Development 2001). The expulsion of worms from the gut may remove a constraint that acts through effects on appetite and food intake, digestion and absorption, or because of the diversion of nutrients in response to infection. To achieve the maximum growth rate after treatment, energy, protein and micronutrients need to be provided to children, preferably ad libitum. Most studies did not give food supplements after treatment, but one that did showed an improved appetite measured as food intake as well as improved growth (Stephenson et al. 1993a).

Catch‐up growth is possible (Golden 1994) especially if the deficits occur early in childhood and are treated adequately. Studies of children treated for Trichuris dysentery syndrome, admittedly an acute disease, have shown an average gain of nearly 11 cm in height and 4 kg in weight a year, which was more than 2 SD above the expected gain in height and weight of British children of the same age (Cooper et al. 1995). Some of the 5‐year‐old children studied showed an annual growth velocity of nearly 20 cm a year and a weight gain of 10 kg.

To achieve such catch‐up growth children need to be fed ad libitum, not just with energy and protein, but also with micronutrients, so that no nutrient is limiting and the maximum possible growth is achieved (Hall 2007). If studies of deworming have failed to show an impact on growth or weight gain, it could be because children were not getting enough of the food and micronutrients they needed to achieve catch‐up growth. And the studies that have shown a statistically significant impact of treatment may have underestimated the potential effect if children did not have an ideal and balanced diet given ad libitum. No published studies have been found of randomized controlled trials of deworming with or without any form of remedial feeding given ad libitum; most studies have only given supplementary food or supplementary micronutrients.

A problem with giving food supplements is that there can be substitution, in which people eat less food at home after being fed at school or work (Wolgemuth et al. 1982). Or supplementary food may benefit the better off most: a study of a school nutrition programme in Vietnam found that better nourished children gained more weight than undernourished children (Hall et al. 2007), which is not what is hoped for in school feeding programmes.

The need for adequate remedial treatment is important when the outcome being measured is the haemoglobin concentration. A study in Tanzania of giving iron supplements to anaemic children after treating their infections with hookworm and S. haematobium showed no improvement in the mean haemoglobin concentration, but the serum ferritin concentration increased, an indicator of improved iron status (Beasley et al. 2000). This suggests that iron was not the micronutrient limiting the haemoglobin concentration.

This does not mean that improvements in growth and nutritional status cannot be achieved by deworming alone, especially if there are improvements in appetite and food is available to children. To have the maximum effect on growth and nutritional status, therapeutic nutrients are needed as well. A failure to have an impact could be because of this lack, or because the worms were not an important cause of any nutritional deficit in the first place.

The costs of only giving supplementary food can be quite high: the World Food Programme estimate that it costs $34 a year to feed a child in school (World Food Programme 2006). An alternative is to provide micronutrient supplements, a deficiency of which may also act to impair growth, and which have been shown to reduce the intensity of reinfection with schistosomes (Olsen et al. 2003). A multiple micronutrient supplement is a much less expensive intervention than food, and may cost only around one US cent per child per day if purchased in large quantities. But the key point is that if any nutrient is not provided in sufficient amounts to meet demands, and becomes rate limiting, then growth after deworming may be constrained.

The weighted average gains in parameters of growth shown in Table 12 therefore do not indicate what extra growth might be achieved after treatment or indicate the potential for growth, they only estimate what can be achieved after giving treatment. If a study included in the analysis showed no statistically significant difference between the treatment and control groups, the question is: did this happen because there was no effect, or because treated children were lacking sufficient energy or nutrients to show extra growth and weight gain? A lack of effect in such studies will serve to reduce the weighted average, and so underestimate the potential impact of treatment. It is also entirely possible that anthelmintic treatment may have had no effect even if children had been adequately nourished. But, as there are plausible biological mechanisms by which moderate to heavy burdens of worms can affect nutritional status, a beneficial impact on some (but not all) children seems likely.

This touches upon the aggregated distribution of worms and the uneven distribution of the impact of treatment. If disease and malnutrition are a consequence of moderate to heavy infections, then in any given sample of children a minority will benefit most from treatment, so the effect of treatment will not be evenly distributed between children. For example, if 25–49 A. lumbricoides is arbitrarily considered to be a moderate infection in children and 50 or more worms is a heavy infection, then of the 1069 children aged 1–14 years studied in an urban slum in Bangladesh, 18% were moderately heavily infected and 7% heavily infected (data from Hall et al. 1999). If only these children benefited from treatment then the impact on this 25% would in effect be diluted in the lack of change in the remaining 75%. The average treatment effect therefore includes a large proportion of children who would not be expected to benefit, and would underestimate the potential impact on moderately to heavily infected children. The proportion of lightly infected children will be very large when the prevalence is low, and when it is less than 50% it might be expected that very few children would benefit substantially from treatment. The general consequence is that the study average will not capture the larger benefit experienced by any moderately to heavily infected children.

One way to control for this would be to analyse the relationship between any outcome and an estimate of the intensity of infection, such as the concentration of eggs in faeces. This method was used by Stephenson et al. (1993a) in their studies of the effect of anthelmintic treatment on appetite. This can only be done if a prior individual diagnosis is made and, if it is, it could be considered to be unethical to leave children untreated whose infections have been diagnosed. One way to deal with this dilemma is to randomly allocate subjects to treatment and control groups but only diagnose infections in the subjects who will be treated. This was done in a study in Tanzania in which infections were diagnosed in the control group at the end of the study (Bhargava et al. 2003). The problem with doing this is that the intensity of infections can change in the control group as worms are lost or gained during a long study, so the treatment and control groups cannot be compared at any point during the study.

Two assumptions are commonly made about the effect of the intensity of infection. First, that treatment has an impact only above a threshold worm burden, which seems reasonable, although the threshold number is likely to be dependent on factors such as the size and current health of the host. Second, that above the threshold the impact of treatment on more heavily infected individuals is linear. This may not be the case, but is very difficult to assess without having heavily infected children as controls during a study, something that would be unethical. An alternative is to examine the nature of the relationship between the intensity of infection with a worm and an outcome that is easily measurable, such as hookworm and haemoglobin concentration. Cross‐sectional studies typically find a non‐linear relationship so that the subjects with the heaviest infections have a markedly lower haemoglobin concentration than lightly infected individuals (Roche & Layrisse 1966; Srinivasan et al. 1987; Pritchard et al. 1991; Lwambo et al. 1992).

This meta‐analysis found no statistically significant effect of treating intestinal nematode worms on the mean haemoglobin concentration of children (8, 9). Yet the potential impact of treating worms that cause internal bleeding, dysentery and the loss of iron is large. In Africa in particular, many children are infected with species of Schistosoma in addition to hookworms, which makes it difficult to assess the impact of treating the intestinal nematodes alone. An important study in Tanzania showed that supplements of vitamin A and iron are needed if rapid improvements in haemoglobin concentration are to be achieved after deworming (Mwanri et al. 2000). All children in the study were given albendazole and praziquantel, then were divided into four groups to receive on 3 school days a week either iron and vitamin A placebo, vitamin A and an iron placebo, iron plus vitamin A, or iron and vitamin A placebos (Mwanri et al. 2000). The haemoglobin concentration of children given both placebos rose by an average of only 3.6 g L⁻¹ in the 12 weeks after treatment which might have been a statistically significant increase, but there was no untreated control group to allow this to be assessed. The effects of the supplements of iron and vitamin A were partially additive and achieved an increase in haemoglobin concentration of 22 g L⁻¹ over the same period compared with 13.5 g L⁻¹ for vitamin A alone and 17.5 g L⁻¹ for iron alone. This study illustrates the importance of giving micronutrient supplements after treatment in order to achieve a rapid increase in haemoglobin concentration and may explain why no statistically significant effect of anthelmintic treatment alone was detected in the meta‐analysis. If the change of 3.6 g L⁻¹ over 12 weeks in the control group reflects the effect of anthelmintic treatment alone, then it might have taken six times as long to achieve an increase of 22 g L⁻¹, or nearly a year and a half. The inclusion of vitamin A in addition to iron makes the important point that not all anaemia is iron deficiency anaemia, and this could explain the results of a study in Tanzania that gave iron alone and failed to achieve a statistically significant increase in haemoglobin concentration (Beasley et al. 2000).

6.3 The Cochrane Collaboration Review

The Cochrane Collaboration supported a meta‐analysis in 2000 (2000a, 2000b). The reviewers noted heterogeneity between the results of the trials they included, and considered that they could be due to a number of factors. These were: the prevalence and intensity of infection, so that only heavily infected children benefited from treatment; the site of the study (school, community or health facility); the age of children; the prior nutritional status of children; and the manufacturer of the drug. The reviewers concluded that ‘there is some limited evidence that routine treatment of children in areas where helminths are common has effects on weight gain, but this is not consistent between trials’. The heterogeneity noted by the Cochrane Review really means a lack of effect in some trials which could be due, as explained here, to the fact that anthelmintic treatment alone is not sufficient to achieve improved nutritional outcomes. The review received considerable criticism concerning the methods of analysis and conclusions (Bhargava 2000; Bundy and Peto 2000; Cooper 2000; Michael 2000; Savioli et al. 2000).

This review was withdrawn in 2007 and replaced by a new analysis that examined effects on growth and school performance (Taylor‐Robinson et al. 2007). The review drew similar conclusions: deworming may lead to improved weight gain in some circumstances but not others, and no effect on cognition or school performance has been demonstrated (Taylor‐Robinson et al. 2007).

6.4 Characteristics of an ideal study

Any studies that attempt to estimate the effect of anthelmintic treatment plus any supplement should ideally have a factorial design. This is illustrated in Fig. 25 in which four groups of subjects are given either deworming or not with a nutritional intervention or not, although it could be applied to any treatment. The factorial design is powerful, but requires randomization of subjects to study groups and an untreated control group to allow the effect of secular change to be estimated and subtracted from the effect of treatment.

However, it is important to appreciate that it may be possible to bring about an improvement in a nutritional variable by a nutritional intervention alone, so that the effect of deworming will be masked. An analogy of a leaking bucket of water may be helpful to explain this concept. The level of water in a bucket, which is the measured outcome, can be sustained or even raised if the water flowing into the bucket is greater than the loss through any leaks. Simply plugging the leaks will not raise the water level, it is necessary to add more water to achieve this. Adding water alone could overcome or mask the effect of the leaks and lead to a rise in the water level. Once the bucket is full, the excess simply overflows but the leaks continue. This creates a paradox: once you start to fill the bucket the change in the level of water no longer accurately reflects the leakage that is occurring.

This analogy may apply better to anaemia than to body weight, as the concentration of haemoglobin does not rise unchecked, although some excess nutrients such as iron and vitamin A can be stored. When worms cause a loss of blood and iron and the haemoglobin concentration falls, it will do so when all stores of iron needed to make haemoglobin have been used up (Crompton & Whitehead 1993). Treating hookworms will not lead to an increase in the haemoglobin concentration if the diet does not contain enough nutrients to supply normal turnover, which is estimated to be about 0.8% of all red blood corpuscles daily, and to provide for increased erythropoiesis to replace corpuscles lost caused by worms. But the cumulative loss of iron and other nutrients could be rectified by giving supplementary food alone, provided that all other requirements are met. The rate of recovery from anaemia is likely to be very slow without nutrient supplements, and studies of haemoglobin concentration after anthelmintic treatment may need to last for a year or more to show the full effect of treatment alone. The same could be said for measuring the effect of anthelmintic treatment on growth.

It would be helpful for any future meta‐analysis of the effects of anthelmintic treatments if any new research studies of the impact of treatment on outcomes such as growth or haemoglobin concentration had certain key features.

• 1

  The study should be performed where worms and undernourished children are common. A prevalence of infections with intestinal nematode worms of more than 70% is suggested, and where >20% of children are underweight or >40% are anaemic.

• 2

  A randomized, controlled design should be used. If individuals are randomized to treatment and control groups then a placebo should be used; if clusters are randomized, then a placebo is ideal, but not essential. To overcome the problem of not giving a treatment to subjects in a cluster trial without a placebo, both groups could be given some other treatment, such as vitamin A.

• 3

  A sample size calculation should be done. As a rule of thumb there should be a minimum of 250 subjects in each group in a randomized trial. In a cluster trial there should be more than 25 clusters per group and a design effect of 2 should be applied.

• 4

  Anthelmintic treatment should be given at least every 6 months. Ideally initial egg counts should be reduced by >70% by the first treatment. This could mean giving repeated treatment initially to achieve a high efficacy, unless an effectiveness trial is being performed.

• 5

  If more than one anthelmintic is given, such as praziquantel with albendazole, or albendazole with micronutrient supplements, a 2 × 2 factorial design should be considered.

• 6

  The key outcomes to be measured are weight and height. It may be helpful to measure triceps skinfold thickness and mid‐upper arm circumference to distinguish between increased adiposity and growth in tissues. If infections with hookworms or T. trichiura are common, then the haemoglobin concentration should be measured as well with, ideally, a measure of irons status such as ferritin or transferrin receptor (WHO/CDC 2005).

• 7

  If growth is to be measured, the study should last a minimum of 1 year, and 2 years if possible. Specific criteria that state when a study should be stopped may need to be set.

• 8

  If the haemoglobin concentration is to be measured the study should last a minimum of 12 weeks, and should be continued to 52 weeks if possible if no micronutrient supplements are given after treatment.

• 9

  Drug efficacy should be assessed 21 days after each round of treatment as a matter of good practice, to assess the development of any anthelmintic resistance.

• 10

  The gains in any parameter related to growth should be standardized as mean gain per year. The mean and standard deviation of all outcome measures and any derived indices should be given in any report or paper for each group, with the sample sizes. If subjects are studied in clusters, then this should be taken into account during the analysis.

6.5 Implications for programmes

Deworming is an easy and inexpensive thing to do, and the drugs are very safe. Treatment can be popular with parents (Brooker et al. 2001). Most anthelmintics can be given as a single dose so no adjustment for body weight is necessary, and a single 400 mg tablet of albendazole or 500 mg of mebendazole can be given to all children older than 1 year (WHO 2002b), although a syrup should be given to very young children. The United Nations Children Fund (UNICEF) in Ethiopia have issued warnings that young children should not be forced to take tablets because of the risk of choking. Anthelmintic drugs should not be given in the first trimester of pregnancy (WHO 1994b). Anthelmintic drugs are usually available over the counter and do not require a prescription, except in developed countries.

Over the last few years the costs of albendazole and mebendazole have been reduced to less than 5 US cents a tablet when bought in large quantities and, according to the WHO, these drugs can be given safely to all children once a year in areas where the prevalence of infection is more then 20% and twice a year where the prevalence is more than 50% (WHO 2006). The new recommendations are shown in Table 14. To promote mass deworming the World Health Assembly in 2001 passed resolution WHA54.19 asking countries to administer anthelmintic treatments annually to at least 75% of all school‐age children at risk of morbidity by 2010 (WHO 2002a).

There are three justifications for giving mass treatment at least once a year in areas where the prevalence of infection with all species of intestinal nematode worms is greater than 50%.

First, because the probability of disease increases exponentially above a prevalence of 50% (see Fig. 3).

Second, the cost of diagnosis is typically many times the cost of treatment. When the prevalence is greater than 50% diagnosis becomes a matter of identifying uninfected people who do not need treatment, and so the cost per case of finding them increases with the prevalence.

Third, most anthelmintic drugs, and the benzimidazoles in particular, are very safe, and there is no known harm in treating people who are not infected. Benzimidazoles have been taken daily for long periods to treat hydatid disease (El‐On 2003; Horton 2003a) and when used to treat intestinal nematodes the overall frequency of side effects is about 1% (Horton 2000). Although one of the potential studies identified for the meta‐analysis reported evidence of an adverse effect on the growth of children given 400 mg of albendazole given for 3 days, there was no untreated control group (so the study was excluded from the present analysis) and the children treated with albendazole were compared with children treated with another anthelmintic, pyrantel embonate (Forrester et al. 1998).

The use of a 20% threshold at which to apply annual mass treatment in school‐age children has cost–benefit implications. When the prevalence of infection is less than 50% the majority of children are uninfected and perhaps only a few per cent of children will actually benefit from treatment. Although the cost per child treated may be small in terms of the cost of each tablet, as the prevalence drops the costs per infected child treated will increase. This is shown in Fig. 26 for a drug costing 3 US cents per treatment. When the prevalence is 50% the cost per infected child treated is twice the cost per child treated, or 6 US cents per child; when the prevalence is only 20% the cost per infected child treated is five times the cost per child treated or 15 US cents per child; and if only 5% of all children actually benefit from treatment the cost per beneficiary is 20 times the cost of the treatment or 60 US cents per child.

Although these unit costs may seem small in absolute terms, the total cost to a government may be large. For example, if the Ethiopian government was considering treating the 11.5 million children enrolled in primary schools in 2005 with a drug costing 3 US cents per treatment, and if the prevalence of infection according to a recent national survey is 30% (SC/US 2007), then of the $345 000 it will cost to buy drugs, 70% or $241 500 will be spent on treating uninfected children.

These costs do not take into account the additional financial or opportunity costs of delivering and administering tablets.

The main issue for deworming programmes is how to deliver treatments at low cost to the three main groups who would benefit most from treatment: pre‐school children, school‐age children and women of reproductive age.

As there are many current programmes delivering treatments such as vaccinations and vitamin A to children less than 5 years, the additional cost of giving a tablet of albendazole or mebendazole is minimal. In recognition of this, the WHO and UNICEF now recommend deworming children as a part of all routine health programmes such as vitamin A supplementation (WHO/UNICEF 2004), and anthelmintic treatment is also recommend as a part of the Integrated Management of Childhood Illnesses after any acute illness has ceased.

A commonly unrecognized concern arises because of the lack of overlap between health programmes for pre‐school children, which typically stop at 5 years of age, and programmes for school children, which begin when children enrol in school. As the official age of enrolment in basic education in many countries is 6 or even 7 years of age, even if all children actually enrol in school at the correct age, there is a gap of 1–2 years when they may fall between programmes and miss out on treatment. As late‐enrolment in school is very common in sub‐Saharan Africa (Partnership for Child Development 1998b) and late enrolment is strongly associated with stunted growth and anaemia (Partnership for Child Development 1999b; Hall et al. 2001), the gap between programmes may be as much as 4 or 5 years, and such children may be in greater need of treatment than those who are actually in school.

In countries where enrolment rates are high, school health programmes offer an effective mechanism to reach a large proportion of school‐age children (Hall et al. 1996), an age group that harbours a large proportion of all worms in a community. The cost of delivering anthelmintics to school children has been shown to be a few US cents per child (Partnership for Child Development 1998a, 1999a). In addition to the potential nutritional benefits of treatment, periodically giving anthelmintics to school children alone has other benefits, as it can serve to bring down transmission to other, untreated groups (Asaolu et al. 1991; Butterworth et al. 1991; Bundy et al. 1992b). Such externalities, as they are called by economists, are not captured in a meta‐analysis such as this, which examines only one outcome at a time.

It is also possible that, in addition to affecting growth and micronutrient status, intestinal nematodes may affect children's cognitive function and educability, although the evidence for this is mixed and often comes from cross‐sectional associations, which are easily confounded (1992a, 1992b; Watkins & Pollitt 1997; Sakti et al. 1999; Dickson et al. 2000b; Jukes et al. 2002). The Cochrane Review of 2007 concluded that there was insufficient evidence that treating worms improves cognitive performance, which included educational outcomes including attendance and tests of educational achievements (Taylor‐Robinson et al. 2007). The controlled trial of 6‐monthly deworming for 2 years in Vietnam found no statistically significant difference between treatment and controls in terms of tests of mathematics and Vietnamese (Partnership for Child Development 2001).

The last group that would benefit considerably from periodic anthelmintic treatment are women of reproductive age, although treatment should only be given before pregnancy or after the first trimester. Nevertheless accidental treatment does not seem to be dangerous: a study of some 400 Sri Lankan women who had taken mebendazole during their first trimester of pregnancy did not find a greater rate of congenital defects than in pregnant women who had not taken the drug (de Silva et al. 1999). The sample size was too small to detect a greater risk of rare congenital defects, so the recommendation of the study was still to avoid treatment with anthelmintics (de Silva et al. 1999).

The main effect of anthelmintic treatment on pregnant women is likely to be on haemoglobin concentration as a result of treating infections with hookworm or Schistosoma spp. A randomized controlled trial in Sierra Leone of giving albendazole to pregnant women after their first trimester prevented a fall in haemoglobin concentration of 6.6 g L⁻¹ (Torlesse & Hodges 2000, 2001). When albendazole was given with iron, both treatments prevented a fall in haemoglobin concentration of 13.7 g L⁻¹ (Torlesse & Hodges 2000, 2001).

As a general rule for all programmes in which anthelmintic treatment for worms is given, drugs should not be administered to individuals who are sick for any reason. This is not because the drug may be ineffective, although it may be if there is diarrhoea. The reason for not giving treatment is because of the possibility that the subject may die of a pre‐existing illness, and it could be concluded that the anthelmintic had caused the death. In any programme in which very large numbers of children are being treated on the same day, such an event becomes increasingly statistically probable, so every effort should be made to ensure that only otherwise healthy children are treated. A possible example of this was the >30 deaths reported after the administration of vitamin A to some 3 million young children in the state of Assam in India in 2001 (West & Sommer 2002), although there was some concern that children may have been overdosed (Kapil 2002, 2004). Reports of deaths of school children during mass deworming have not been documented.

The other concern during mass treatment campaigns is for adolescent girls who may be pregnant and either not know it, or be afraid to report it. Teratogenic effects of drugs are most likely to occur during the first few weeks of pregnancy, a period when a woman may not yet be aware that she is pregnant. It cannot be assumed that school‐age girls are not sexually active. A study of 9000 school children in grades 4 to 6 in Tanzania found that 20% of girls reported having had sex, but only 39% of 114 girls with biological markers of sexual activity such as an infection acknowledged having had sex, indicating that such activity was greatly under‐reported (Todd et al. 2004). In an analysis of official education statistics, also in Tanzania, pregnancy was reported to be a cause of school dropout for six or seven girls per 1000 in grades 6 and 7 respectively (Partnership for Child Development 1998b).

These data should provide a warning to programmes that give mass treatment with anthelmintic drugs to school‐age children as well for those considering adding mega‐dose supplements of vitamin A to treat vitamin A‐deficient or anaemic adolescent girls. It cannot be assumed that adolescent girls are not pregnant.

6.6 Conclusions

Treating intestinal nematode worms can lead to significantly better weight gain and growth when treated children are compared with untreated controls.

The magnitude of the effect of treatment is specific to local epidemiological circumstances, and not all infected children will benefit to the same degree, if at all.

If no significant effect of anthelmintic treatment is detected in a study, it cannot be assumed that worms did not contribute to undernutrition; it is likely that removing worms is insufficient to lead to improved growth without extra food and nutrients.

The maximum size of any effect of treating worms cannot be estimated; studies can only indicate what difference is achievable. To show a significant effect it is necessary to have an untreated control group and a sufficient period of follow‐up. This is likely to become less tenable given the measured benefits of treating worms that have been shown. Nevertheless it would be useful to have more studies of the effectiveness of anthelmintics delivered through health programmes, especially those that focus on pre‐school and school‐age children, and perhaps given with multiple micronutrients to make sure that no vitamin or mineral is a factor limiting catch‐up growth.