^{1}

^{*}

^{2}

Conceived and designed the experiments: NdS. Performed the experiments: NdS. Analyzed the data: NdS AH. Contributed reagents/materials/analysis tools: NdS AH. Wrote the paper: NdS AH.

The authors have declared that no competing interests exist.

To assess if a probabilistic model could be used to estimate the combined prevalence of infection with any species of intestinal nematode worm when only the separate prevalence of each species is reported, and to estimate the extent to which simply taking the highest individual species prevalence underestimates the combined prevalence.

Data were extracted from community surveys that reported both the proportion infected with individual species and the combined proportion infected, for a minimum sample of 100 individuals. The predicted combined proportion infected was calculated based on the assumption that the probability of infection with one species was independent of infection with another species, so the probability of combined infections was multiplicative.

Thirty-three reports describing 63 data sets from surveys conducted in 20 countries were identified. A strong correlation was found between the observed and predicted combined proportion infected (r = 0.996,

A simple probabilistic model of combined infection with a small correction factor is proposed as a novel method to estimate the number of individuals that would benefit from mass deworming when data are reported only for separate species.

Mixed infections with roundworm, whipworm and hookworm are common, but survey reports often give only the separate prevalence of each type. However, the combined prevalence is important to estimate accurately the number of individuals who would benefit from control programmes and to make decisions about the frequency of treatment. Previous work suggests that mixed infections involving hookworm occur randomly, but that roundworm and whipworm infections are found together more frequently than would be expected by chance. We used 63 data sets from community surveys that reported both the proportions infected with individual types of worms and the combined proportion infected with any worm. We then calculated the proportion that would be infected with any type of worm if infections had occurred randomly and compared it with the observed combined proportion infected. We found a strong correlation between the observed and predicted combined proportions infected. A small downward correction of the predicted proportion infected by dividing by a factor of 1.06 brought it to a value that nearly equalled the observed proportion infected almost all the time. This simple model could be applied to published survey data to estimate accurately the number of individuals that would benefit from mass deworming.

The World Health Organization (WHO) estimates that intestinal nematode worms, also known as soil-transmitted helminths, are currently endemic in 130 countries in the world

As all these worms can be treated using a single dose of an inexpensive anthelmintic drug, the WHO recommends a strategy called “preventive chemotherapy”

With the resurgence of interest in controlling soil-transmitted helminth infections, much more field survey data are now available than when the probabilistic model was first proposed and tested

A database of 230 publications in peer-reviewed journals, grey and unpublished literature that had been compiled in 2003 to estimate the global prevalence of intestinal nematode worms (described in ref.

a = Proportion infected with

t = Proportion infected with

h = Proportion infected with hookworms

The proportions infected with each permutation of infection were then calculated as:_{ath}) is thus the sum of all seven equations above:_{at}) is the sum of the three equations above:

This can be simplified by cancellation to: _{ah}_{th}

Equation 1 was applied in an Excel spreadsheet to calculate the predicted combined proportion infected from the data from each survey and the values were plotted against the observed combined proportion infected in the same survey. When only two worms were identified in a survey if a value of zero is entered for the missing type then the spreadsheet calculates the correct proportion infected with either or both species and it is not necessary to apply Equations 2 to 4.

To investigate the degree to which the highest single species prevalence may underestimate the combined prevalence, the differences between the highest individual species value and the observed combined proportion infected were plotted against the observed combined proportion infected.

To investigate the degree to which individual species were associated, correlation coefficients (r) were calculated for data derived from all surveys for the proportions infected with

Thirty-three papers describing surveys conducted in 20 different countries were identified for this analysis: eight in Asia, six in Africa, five in Latin America and the Caribbean, and one in Oceania. Together they contained 63 sets of data: 30 from Asia, 23 from Africa, nine from Latin America & the Caribbean and one from Oceania (see

_{ath}) so that:

A plot of the observed combined proportion infected against this adjusted predicted combined proportion infected (not shown since it is almost identical to

To assess the magnitude of these underestimates in relation to the prevalence,

The data were also analysed for correlations between proportions infected with

This paper presents a simple equation (Eqn 1) to estimate the combined prevalence of infection with

The strong correlation reported here between the observed and predicted combined prevalences supports the hypothesis proposed by Booth & Bundy

However, the present analysis also shows that a small downward correction of the predicted combined proportion infected is enough to achieve a very high correlation between predicted values and values reported by field surveys. The apparent over-estimation of combined prevalence probably results from the association between

This analysis does not take into account potential errors in parasitological diagnosis, particularly false negatives leading to an underestimated prevalence. The sensitivity and specificity of diagnosis are likely to be related to the concentration of eggs in faeces, which is related to fecundity of worms, the dispersal and dilution of eggs in the faecal mass, and to the amount of faeces examined under a microscope

The difference between the prevalence of the single most common species of intestinal nematode, which is currently used by the WHO in the absence of data on combined prevalence, and the observed combined prevalence, seems to vary depending on the prevalence. At a combined prevalence of ≤40%, the difference is on average 7% or smaller, but when the combined prevalence is higher, the difference is about 12%. The difference is less also when the combined prevalence is very high (>90%). This has implications for mass treatment, especially at low prevalence rates. For example, if the proportion infected with

This analysis includes a modest number of data sets from all major geographical areas where intestinal nematode infections are endemic. It suggests that a simple probabilistic model with a small correction could be used to estimate the proportion of people infected with any intestinal nematode worm. This could help with the global mapping of disease and is likely to increase the estimated number of individuals that would benefit from mass deworming in the world today.

Data used for analysis.

(0.24 MB DOC)

Thanks to Veronica Tuffrey for help with the equations.