Measuring morbidity in schistosomiasis mansoni: relationship between image pattern, portal vein diameter and portal branch thickness in large-scale surveys using new WHO coding guidelines for ultrasound in schistosomiasis


correspondence Dr Charles H. King, Division of Geographic Medicine, Case Western Reserve University and University Hospitals of Cleveland, Division of Geographic Medicine, W137, Cleveland, OH 44106-4983, USA. Tel: +1 216 368 4818; Fax: +1 216 368 4825; E-mail:


Objective  World Health Organization consensus meetings on ‘Ultrasound in Schistosomiasis’ in 1996 and 1997 anticipated further challenges in the global implementation of a standardized protocol for morbidity assessment in schistosomiasis mansoni. We evaluated the performance of the qualitative and quantitative components of the new Niamey criteria.

Method  Use of the Niamey protocol among 3954 subjects in two linked, cross-sectional ultrasound surveys of Schistosoma mansoni-endemic populations in Egypt and Kenya.

Results  There were significant differences between Egyptian and Kenyan sites in prevalence and age distribution of S. mansoni-related hepatic fibrosis (36%vs. 3%, P < 0.001). Protocol image pattern scoring could be performed quickly and was stable to interobserver variation. However, there were unintended but systematic differences between study sites in the measurement of portal vein diameter (PVD) and wall thickness. By Niamey criteria, a high prevalence of portal dilation was scored for normal Egyptian subjects, which reduced the predictive value of image pattern for portal hypertension. Using alternative height-indexing of PVD, image pattern plus PVD findings predicted 15% of Egyptians and 2.5% of Kenyans were at risk for variceal bleeding, whereas locally derived PVD norms estimated 25% of Egyptians and 12% of Kenyans to be at possible risk.

Conclusion  Niamey scoring criteria performed acceptably as a relative grading system for disease in schistosomiasis mansoni, but failed to account fully for site-to-site variation in test performance and morbidity prevalence. Consequently, standardized image pattern scoring appears to provide the most useful tool for detection and comparison of S. mansoni-associated morbidity in large-scale surveys.


Intestinal schistosomiasis caused by the blood fluke Schistosoma mansoni remains one of the most prevalent infections in the world (van der Werf & De Vlas 2001). Because health care resources are limited in many endemic countries, schistosomiasis control programmes have now put emphasis on prevention and control of morbidity rather than on the elimination of transmission (Bergquist 2001). In terms of risk assessment, although there are very good techniques for quantifying the presence of S. mansoni infection, these do not fully characterize an individual's risk for morbidity (Guyatt et al. 1994; Hatz 2001; Kariuki et al. 2001). In order to reach the goal of effective morbidity prevention, better techniques are needed for detecting and measuring S. mansoni-associated disease.

Portable ultrasound examination has the potential to provide both sensitive and precise measurement of S. mansoni-associated abdominal pathology. Using ultrasound examination, advanced degrees of S. mansoni-related liver fibrosis are easily distinguished from normal liver architecture (Homeida et al. 1988b; Abdel-Wahab et al. 1989; Doehring-Schwerdtfeger et al. 1992; Richter et al. 1992; Domingues et al. 1993). Difficulties have arisen, however, in establishing a common grading system for the ultrasound findings of disease (Richter et al. 2000; Hatz 2001). Differences have been noted in comparing severity scores between endemic sites and in using scoring to predict patient prognosis (Richter et al. 1992, 1998; Eltoum et al. 1994; Hatz 2001). In 1991, the WHO convened the first ‘Meeting on Ultrasonography in Schistosomiasis’ (Cairo) to establish a common grading system (WHO 1991). The essence of this system was strict quantitative measurement to assess periportal fibrosis, portal vein enlargement, and liver and spleen size. Disease was staged as levels I, II, or III to reflect increasing severity of pathological changes. This scoring system was found to be difficult to apply consistently, particularly for resolving low levels of morbidity. In addition, the prognostic value of specific findings remained unclear. As a result, the second (Niamey 1996) and third (Belo Horizonte 1997) ultrasound consensus meetings were convened. These resulted in the evolution of the standard ultrasound scoring protocol to include a qualitative assessment of liver parenchyma (according to reference patterns A to F) and an adjustment of organ measurements for height (Richter et al. 2000).

Under the new protocol, weighted scores were assigned to each parenchymal pattern, then combined with height-indexed measurements of portal branch wall thickening and portal vein enlargement, along with signs of portal hypertension (ascites, collateral vein enlargement) in a scoring matrix to define the presence of possible, probable, definite, or advanced schistosomiasis-related periportal fibrosis and/or portal hypertension. However, no experimental basis was provided for the protocol's weighting system. There was also no consensus on the relationship between the qualitative and quantitative components of the standard. Adopting this new scoring system for common field conditions, with the expected constraints on time and available expertise, was expected to require simplification of exam techniques and of scoring (Hatz 2001). It also remained uncertain whether image pattern, portal thickness, portal hypertension, or a combination of these findings provided the best estimate of a patient's infection-related disability and risk for mortality. Our study, based on experience with 3954 ultrasound studies performed in cross-sectional surveys of two S. mansoni-endemic communities in Egypt and Kenya, sought to validate and reconcile the qualitative and quantitative aspects of the Niamey standard (Richter et al. 2000).

Materials and methods

Study recruitment

This study was conducted in accordance with the ethical guidelines for human research set forth in the Declaration of Helsinki, under a protocol approved by the Human Investigation Review Boards of University Hospitals of Cleveland, OH, USA, the Kenya Medical Research Institute (KEMRI), Nairobi, Kenya, and Cairo University, Egypt. All participants gave free and informed consent before participation in the study.

Populations and parasitology

In parallel studies performed during 1999–2000, one community from rural Kenya and one community from rural Egypt were surveyed for prevalence of S. mansoni-related morbidity. These communities were selected based on their relatively high prevalence of S. mansoni infection.

In Kenya, all individuals over 11 years of age from Katheka (Machakos District, in central Kenya) were examined. Demographic data, including gender, age and relationship to head of household were collected by trained local personnel in the local language. Schistosoma mansoni infection was detected and quantified by Kato-Katz testing for schistosome ova in two daily stool samples (Katz et al. 1972). Schistosoma haematobium infection was excluded by standard urine filtration (Peters et al. 1976). All subjects were measured for height to the nearest centimetre by means of a fixed tape measure. The presence of hepatic fibrosis and portal vein changes was determined by ultrasound (n = 2120) as described below.

In Egypt, approximately 40% of the population above age 11 in the village of Shamarka (Kafr el Sheikh governorate, Nile Delta) were interviewed, tested for schistosome ova in stools and urine, and examined for fibrosis by ultrasound (n = 2038), following the same research protocol.

Hepatic ultrasound

The portable ultrasound units employed in both communities were Shimadzu model SDU-350 A (Shimadzu Medical Systems, Torrance, California, USA) machines equipped with 3.5 MHz convex abdominal transducers. Images were obtained by physicians in Egypt and ultrasound technicians in Kenya. Operators were instructed in performing the ultrasound protocol by the authors using the draft Niamey document (Richter et al. 2000; Kariuki et al. 2001). Exam technique was standardized under the supervision of a senior radiologist (PM), who also reviewed all images from both sites for quality and final determination of pattern. Radiologists' agreement on image pattern was 84% (kappa statistic = 0.534).

Subjects were examined in the supine position with the transducer positioned to obtain the best image. In some cases, the positions described in the standard Niamey protocol (Richter et al. 2000) did not provide the best imaging views. Liver patterns were designated A, B, C, D, E, F, X, Y and U based on the extent of fibrosis or other parenchymal pattern (Table 1). Persons with X, Y or U patterns indicating liver pathology unrelated to schistosomiasis (such as cirrhosis and fatty liver) were scored as such and excluded from the current analysis.

Table 1.  Hepatic fibrosis by Niamey image pattern scores (Richter et al. 2000)
BDiffuse echogenic foci (‘starry sky’); minimal evidence of wall thickening around portal and subsegmental branches
CRing echoes around vessels in cross-section; pipe-stems parallel with portal vessels
DEchogenic ruff around portal bifurcation and mainstem; thickening of walls of main portal branches
EHyperechoic patches expanding into parenchyma
FEchogenic bands and streaks expanding from main portal vein and its bifurcation to liver surface, where they retract  organ surface
YFatty liver
UUnknown; indeterminate

For portal vein measurements, the external and luminal diameters of the largest first segmental portal branch were measured as close as possible to the branch point using magnification to increase accuracy. This segment was then re-imaged and replicate measurements obtained (Annex B; Richter et al. 2000). The mean was calculated from these repeated measurements, and the inner diameter subtracted from the outer diameter to yield a value for the two-wall thickness (portal branch wall thickness, PBWT). The inner diameter of the portal vein near its entry into the liver was likewise measured and the mean of the two measurements calculated (portal vein diameter, PVD). An index consisting of the mean measurements divided by height was also generated to provide correction for differences in subject size (Richter et al. 1998), and Annex C in Richter et al. (2000). Ascites, collateral formation, gallbladder thickening, and hepatic masses were recorded whenever present.

Statistical analysis

Descriptive statistics for subject characteristics and outcomes were assessed using SPSS (Version 10, SPSS Inc., Chicago, IL, USA). Because of the non-normal distribution of infection intensity [reported as eggs per gram (epg) faeces], egg counts were normalized by using log (egg count + 1) for comparisons and modelling analysis. Multivariate logistic regression (PROC LOGISTIC, SAS v8.2, Cary, NC, USA) was used to determine age-, gender-, infection intensity- and site-adjusted relative odds for the presence of fibrosis, based on image pattern scores (Allison 1999). Because fibrosis risk varied with age in a non-linear fashion, age status was input into the model according to age group (by decade) using coded dummy variables. In further analysis of the risk for portal hypertension, associations between PVD and image pattern, PBWT, age, gender, height, infection intensity, and study site were determined by preliminary bivariate, and then multivariate regression analysis using PROC CORR, PROC REG and PROC GENMOD in SAS version 8.2 (Freund & Littell 2000).


Study participants

Table 2 summarizes the population characteristics of participants from the Kenyan and Egyptian villages involved in this study. Overall, 1862 Egyptians and 2092 Kenyans completed ultrasound examinations. Of these, 2055 participants were female and 1899 were male (52% and 48%, respectively). The median age of participants was 30 years in Egypt, and 24 years in Kenya. Kenyan participants had, on average, greater intensity of infection, with a geometric mean egg count of 16 epg faeces, as compared with 2.5 epg in Egyptian subjects.

Table 2.  Population characteristics of study subjects
CountryGendernMedian age
in years
Mean height ± SD
in cm
Geometric mean
egg count
EgyptFemale84430156 ± 122.0
 Male101830162 ± 212.5
 All186230159 ± 192.5
KenyaFemale121127153 ± 1015.8
 Male88119156 ± 1420.0
 All209224154 ± 1215.8
Total 395427157 ± 166.3

The age-specific gender distribution of participants is shown in Figure 1. These indicate variation in age- and gender-specific participation between the two study sites: there were comparatively lower levels of participation for older Kenyan males and for Egyptian male teenagers, whereas Egyptian women in their twenties were less well represented than Egyptian males of the same age group.

Figure 1.

Age and gender distribution of ultrasound study populations in Shamarka village in Kafr el Sheikh Governorate, Egypt (n = 1862, top panel) and in Katheka village, Machakos District, Kenya (n = 2092, bottom panel).

Risk for hepatic fibrosis

Figure 2a and b shows the distribution of ultrasound pattern scores according to age group for each study site. Remarkably, 36% of Egyptians had ultrasound findings indicative of definite periportal fibrosis (patterns C, D, E, or F), while fewer than 3% of Kenyans did despite their greater mean intensity of infection (χ2 = 738, P < 0.001). In Egypt, prevalence of these fibrosis-related ultrasound patterns increased progressively from 19% in the youngest age group studied (10–19 year) to 63% in the 50–59 year age group. In contrast, in Kenya the peak prevalence of fibrosis (4%) was noted in the 30–39 and 40–49-year age groups, with prevalence declining to less than 2% among older age groups.

Figure 2.

(a) Prevalence of ultrasound image pattern scores by age group in Shamarka, Egypt. (b) Age group distribution of pattern scores in Katheka, Kenya.

To adjust the fibrosis risk estimates for possible confounding effects related to differences in distribution of gender, age and burden of S. mansoni infection between the two study populations, we used multivariate logistic regression to determine the adjusted odds ratios for fibrosis (i.e. risk of having pattern C, D, E, or F score) at the two study sites. The results of this analysis (Table 3) indicate that the country effect (Egypt vs. Kenya) remained highly significant, with an estimated 26-fold increase in odds of fibrosis among Egyptians compared with Kenyans. Infectious burden (in terms of epg) at the time of ultrasound examination was an independent predictor of risk for fibrosis, although the effect was not strong {1.5-fold increase in odds for each log unit increase in egg count [95% confidence interval (CI) 1.3, 1.7]}. Men were significantly more likely than women to have ultrasound findings of fibrosis (adjusted OR = 2.6; 95% CI 2.1, 3.2). In addition, age was a significant factor – adults had a progressively increased risk of fibrosis from their teens up to their fifties – adults in their fifties had nearly a sixfold increase in risk compared with the youngest age group (teenagers), when adjusted for gender, country, and infection intensity (OR = 5.9, 95% CI 3.8, 9.0).

Table 3.  Adjusted odds ratio for detection of fibrosis pattern*on ultrasound examination of S. mansoni-infected subjects
EffectRatio estimates
Adjusted odds ratio95% confidence limits
  • *

     Image pattern C, D, E, or F.

  •  Not significantly different from 1.0.

Egypt25.617.8, 36.8
Male2.62.1, 3.2
 20–292.11.5, 2.7
 30–393.62.7, 4.8
 40–494.23.1, 5.7
 50–595.93.8, 9.0
 60–694.32.3, 8.0
 70+1.70.46, 6.2
Log eggs per gram1.51.3, 1.7

Prevalence of significant portal vascular changes

To assist in identifying patients with portal vein enlargement and thickening (and, by extension, risk for portal hypertension and consequent gastrointestinal bleeding) the Niamey protocol includes cut-off scores for abnormal main PVD and portal vein branch wall thickness (Richter et al. 2000). The size cut-offs provided are based on the standard deviations observed in organometry among a healthy, uninfected Senegalese population (Yazdanpanah et al. 1997); these are used to calculate a weighted score indicating relative severity of hepatic fibrosis and portal hypertension.

We examined the utility of this scoring system in our two study populations, and found that the numeric Niamey cut-offs seriously overestimated the risk of portal vein enlargement in our subjects. For Egyptian patients with normal hepatic ultrasounds (i.e. pattern A), 30% met Niamey height-indexed criteria for portal branch thickening, and 39% met criteria for portal vein enlargement. Similarly, among Kenyans, 18% of normal pattern A subjects were classed as having significant portal branch thickening, while 14% were scored as portal vein enlargement using the published criteria. These relatively high rates of ‘abnormal’ vessel measurements, detected among patients with minimal risk, indicated a need to reassess the population-specific predictors ofS. mansoni-related portal hypertension.

Regression analysis of portal vein enlargement

Figure 3 demonstrates percentile box plots of the distribution of PVD and PBWT measurements for each ultrasound pattern at the Kenyan and Egyptian sites studied. There was a significant, direct association between PVD and pattern score, as well as between PBWT and pattern score.

Figure 3.

Top panel: correlation between measured portal vein diameter (PVD) and ultrasound image pattern score. Data are represented as box plots indicating 5th, 25th, 50th, 75th and 95th percentile measurements for each site and pattern. Dashed lines indicate fitted linear regression lines for Egypt [PVD = 11.5 + (0.67 · pattern code), R = 0.479, P < 0.001] and Kenya [PVD = 8.2 + (0.70 · pattern code), R = 0.331, P < 0.001]. Bottom panel: correlation between measured portal branch wall thickness and image pattern score. As for top panel, box plots indicate percentiles observed for each pattern and study site. Best-fit linear regression lines were: for Egypt – thickness = 2.99 + (0.41 · pattern code) (R = 0.488, P < 0.001); for Kenya – thickness = 3.21 + (0.80 · pattern code) (R = 0.636, P < 0.001).

By bivariate regression, the correlation between PVD and ultrasound pattern (numerically coded 1–6 for patterns A–F) was R = 0.479 (P < 0.001), and the correlation between PBWT and pattern score was R = 0.483 (P < 0.001). In subsequent multivariate regression, adjusting for gender, location, age group and infection intensity, pattern score remained the strongest predictor of PVD (partial R2 = 0.233, F = 509, P < 0.001). The multivariate analysis indicated a significant and consistent site-specific 3.3 mm difference between the Kenyan and Egyptian ultrasonographers in portal vein measurements for all pattern categories (Figure 3, top panel). Further, a significant site/pattern interaction was noted, which tended to vary measurement of PBWT for patients with more advanced fibrosis depending on study site (Figure 3, bottom panel). The PBWT readings in Egypt tended to be progressively larger with more advanced fibrosis, as compared with PBWT readings for the same pattern in Kenya. Because of these country-specific differences, repeated multivariate analyses were performed according to study site. These indicated that pattern score remained the best adjusted predictor of PVD in Egypt (partial R2 = 0.261, F = 553, P < 0.001), whereas, in Kenya, the combination of age status between 30 and 49 years and the presence of an advanced ultrasound image pattern was more effective in predicting increased PVD (R2 = 0.244, P < 0.001, for the combined model).

Predictions were not improved by use of the quasi-exponential coding (i.e. 0, 1, 2, 4, 6, 8) employed for ultrasound patterns in the Niamey scoring system. Inclusion of measured PBWT accounted for only 3% of the variance in PVD beyond the variance accounted for by pattern, gender and age predictors. Adjustment for subject height marginally improved prediction of PVD enlargement in Egypt (by 2%), but not in Kenya.

When analysed by country-specific norms, more than 47% of Egyptian subjects with hepatic fibrosis (patterns C–F) had PVD readings above the 95th percentile PVD measurement for Egyptian pattern A subjects. In Kenya, 26% of fibrosis subjects had PVD measured above the Kenyan pattern A 95th percentile cut-off. Compared with the test performance of the published Niamey cut-offs, selection of this local 95th percentile cut-off provided a strong negative predictive value (95%) for normal fibrosis pattern in terms of portal vein dilation, while providing balanced sensitivity and specificity (≥71% for each test characteristic) for detection of portal vein enlargement in subjects with more advanced fibrosis pattern scores.

Assessment of bleeding risk

In estimating the likelihood of severe morbidity (based on earlier longitudinal studies of bleeding risk associated with portal vein enlargement (Richter et al. 1992, 1998), our subsequent analysis addressed the distribution of a height-indexed PVD enlargement (>10 mm/m height) according to other study parameters. Figure 4 demonstrates the prevalence of indexed PVD enlargement according to age group and pattern score. At both Egyptian and Kenyan study sites, prevalence of height-indexed portal vein enlargement increased with advancing pattern score. In Egypt, risk of portal enlargement increased progressively with age through the sixth decade of life, whereas in Kenya, peak prevalence was noted for those aged 30–39 years. Overall, 10.4% of Egyptians and 1.9% of Kenyans had evidence of portal hypertension using the indexed PVD cut-off. Using the combined indexed-PVD/advanced image pattern risk score of Richter et al. (1998), we estimated that 15% of Egyptian subjects were at risk for oesophageal bleeding as compared with only 2.5% of Kenyans studied.

Figure 4.

Estimated prevalence of bleeding risk (based on portal vein diameter ≥10 mm/m of height according to criteria of Richter et al. (1998) in the Egyptian and the Kenyan study populations. Left panel: prevalence of risk according to ultrasound image pattern. Right panel: risk prevalence according to age group.

A repeat analysis, using PVD cut-offs derived from the present study, defined PVD enlargement as PVD measurements greater than the 95th percentile PVD for normal (pattern A) subjects in each country. By these criteria, a total of 24% of Egyptians and 11% of Kenyans had evidence of significant portal vein enlargement. Twenty-five per cent of Egyptians and 12% of Kenyans had either significant PVD enlargement or advanced portal fibrosis score (E or F) and, by extrapolation of the Richter scoring system, were considered at possible risk for variceal bleeding.


Our experience with the Niamey protocol in large-scale morbidity surveys indicated the following: (a) there was wide site-to-site variation in the prevalence and age-distribution of S. mansoni-related hepatic fibrosis; (b) despite the use of a standardized protocol, there were site-specific variations in measurement of PVD and PBWT; (c) this variation affected the utility of scoring systems used for classifying morbidity (Richter et al. 2000) and estimating risk for variceal bleeding (Richter et al. 1998); (d) advanced pattern scores were significantly associated with increased PVD and PBWT; and (e) based on locally derived percentile cut-offs, the image pattern score was an efficient means of indicating increased individual risk for enlarged PVD or increased PBWT, and by extension, risk for severe disease.

Overall, the Niamey protocol provided a consistent improvement in the standardized grading of S. mansoni-associated liver morbidity. Despite this, although our experienced ultrasonographers followed a detailed written protocol with identical machines under the guidance of study co-ordinators, there were systematic differences in the performance of quantitative measures of the PVD and thickness. Such interobserver variation in quantitative measurement was reported previously using the earlier Cairo and Managil ultrasound classification systems (Thomas et al. 1997). Adjustment for body size (in terms of height) did not improve the correlation between image pattern and quantitative scores, and the PVD and PBWT scoring cut-offs provided with Niamey protocol proved unsuitable for our populations.

Since schistosomiasis mansoni was first described, its most striking pathological characteristic has been the presence of intense fibrosis, which is associated with connective tissue deposits tracking along the portal system, resulting in significantly narrowed portal vein branches. This ‘pipestem’ or ‘Symmers’ fibrosis, found either at autopsy or on biopsy, is pathognomonic for liver disease caused by S. mansoni. With the advent of ultrasonography, rapid non-invasive identification of periportal fibrosis became possible, first in hospital and then in field settings (Hatz 2001). In addition, accurate measurement of portal vein dilatation (indicating portal hypertension) could also be performed. Two studies have confirmed the relationship between ultrasound findings and liver biopsy in advanced schistosomiasis mansoni (Homeida et al. 1988a; Abdel-Wahab et al. 1989). Beyond the convenience and safety of a non-invasive procedure, abdominal ultrasound avoids the potential for sampling error that is inherent in hepatic biopsy procedures for detection of S. mansoni-related fibrosis.

Previous small-scale studies by Richter and colleagues (Richter et al. 1992, 1998) in Brazil indicate that both enlarged PVD and the presence of advanced fibrosis patterns (patterns E and F) on ultrasound examination are associated with increased (>70%) risk for oesophageal variceal bleeding in schistosomiasis. This advanced form of morbidity is a major cause of death in S. mansoni-endemic rural areas (Kheir et al. 1999). In contrast with the Brazil experience, studies by Eltoum et al. (1994) in Sudan found in multivariate analysis that PVD was a not an independent predictor of bleeding risk. However, this Sudan study included a direct grading of oesophageal varices detected on endoscopy. Because of the likely correlation of the graded score for varices with portal hypertension, it was not surprising that the significance of PVD score, beyond the contribution of varices, would be quite small. In our study, which did not include a direct assessment of varices, we estimated that at least 15–25% of Egyptians and 2.5–11% of Kenyans were at risk for significant illness in terms of oesophageal bleeding. Because we do not yet have longitudinal data, it is possible that our findings may not fully predict the natural history of S. mansoni-related morbidity. Specifically, it will be important to determine the prognostic utility of the Niamey scoring system vis a vis the alternative height-indexed scoring system and the local percentile scoring we have developed.

In the future, the most common protocol use of ultrasound testing is likely to be a screening tool for morbidity in schistosomiasis control programmes. Such screening should allow more focused intervention for patients at risk for advancing fibrosis or variceal bleeding. For this purpose, the finding of normal risk image pattern does not require further examination for portal vein measurements. Given our experience with the Niamey cut-offs, however, we would recommend that a subset of those with pattern A have quantitative measurements to develop local cut-off norms for the study population.

We agree that the current Niamey protocol for ultrasound scoring of morbidity in intestinal schistosomiasis will require further standardization and simplification. Streamlining is important, because the extent of studies will be limited by available funds and subjects are likely to develop study fatigue when procedures are prolonged. Overall, we feel that the standardized image pattern is likely to be the most sensitive and efficient component of the protocol for screening patients for risk for serious morbidity. For the patient, this requires only an estimated 1–2 min from entrance to exit.

At present, the prognostic features of individual ultrasound findings still remain to be confirmed in different endemic settings. However, for rapid morbidity screening in control programmes, it is likely that a score based on pattern and, where appropriate, a PVD normalized to local values, will be the most efficient tool (Richter et al. 1998). Research applications require a more detailed quantification of circulatory findings (Kariuki et al. 2001), so that adherence to the full Niamey protocol may be appropriate, provided local norms are used to determine abnormal values.


We thank the people of Shamarka and Katheka for their ready cooperation with this study. We are indebted to the Director of Medical Services, Ministry of Health, Kenya, and the Ministry of Health, Egypt for their assistance and support for this study. This work was supported in part by grants AI41680 and AI45473 from the National Institute of Allergy and Infectious Diseases of the US National Institutes of Health.