which is again the difference between the mean intensity in females (7f) and that in males (7m) corrected for sample size (here denoting the

Fig 2.6 Frequency distribution of differences in parasite prevalence from host-parasite systems involving fish, bird and mammal hosts. Arrows indicate the arithmetic mean difference. Values to the left of the broken line denote higher prevalence in males, while values to the right indicate higher prevalence in female hosts. From Poulin (1996), with permission of the University of Chicago Press

Fig 2.6 Frequency distribution of differences in parasite prevalence from host-parasite systems involving fish, bird and mammal hosts. Arrows indicate the arithmetic mean difference. Values to the left of the broken line denote higher prevalence in males, while values to the right indicate higher prevalence in female hosts. From Poulin (1996), with permission of the University of Chicago Press numbers of infected individuals). Differences in intensity were expressed as a proportion of the intensity in females to standardize for the variability in the mean intensities recorded, which ranged from a few parasites to several thousand parasites per host. If there is no sex bias in levels of infection, differences in prevalence and infection are expected to be normally distributed around a mean of zero. Also, the number of positive differences (higher levels of infection in females) should equal the number of negative ones (higher infection in males).

Fig 2.7 Frequency distribution of differences in infection intensity from host-parasite systems involving fish, bird and mammal hosts. Arrows indicate the arithmetic mean difference. Values to the left of the broken line represent greater intensity in males, while values to the right indicate greater intensity in females. From Poulin (1996), with permission of the University of Chicago Press

Fig 2.7 Frequency distribution of differences in infection intensity from host-parasite systems involving fish, bird and mammal hosts. Arrows indicate the arithmetic mean difference. Values to the left of the broken line represent greater intensity in males, while values to the right indicate greater intensity in females. From Poulin (1996), with permission of the University of Chicago Press

The results of the study indicated a tendency for infection prevalence to be higher in males in many types of host-parasite associations, particularly for nematode infections in birds and mammals. The male bias in prevalence was also apparent in birds and mammals when all parasite species comparisons were pooled (birds X = —5.43, df=9o, t (two-tailed to compare estimated value against expected mean of zero)=3.97o, ^<o.oo1; mammals, X=-3.78, df=1o9, t=2.993,^<o.oo5). There were also more negative (male-biased) than positive (female-biased) differences in prevalence among bird and mammal hosts (birds: 63 vs. 25, X2=16.41,p<o.oo1;mammals: 64vs. 41, x2=5.o4,

^<0.025; see Figure 2.6). By contrast, intensity of infection showed no clear sex bias except for nematodes parasitizing mammals, differences in infection intensity once again being significantly male-biased. When all comparisons were poled by host type, a male bias was again observed only in mammals X =-0.69, df=58, t=4.086, ^<0.001). The frequency of male-biased differences were also more common than female-biased ones (37 vs. 21, x2=4-41, ^<0.05; Figure 2.7).

The Association between Microfilaremia and Chronic Disease in Lymphatic Filariasis

A long-held tenet in the epidemiology of lymphatic filariasis, the major mosquito-borne helminth infection of humans, is that patent infection (microfilaraemia) is negatively related to chronic disease. In conjunction with immunological findings (Ottesen, 1992), this perception had led to the conventional explanation that chronic pathology patients (i.e. those with lymphoedema and hydrocele) are negative for patent infection because of re-expression of antiparasite immunity.

Michael and colleagues (1994) employed meta-analysis techniques to examine the empirical evidence for the relationship between an individual's microfilarial and disease status using published data from field studies carried out in a variety of bancroftian filariasis endemic areas. The aims were two-fold; first, to determine whether there is a negative association between the occurrence of chronic disease (hydrocele, lymphoedema and the two combined) and patent infection, as suggested by the immunological model; second, to determine whether the form of this association varies between studies, and whether this heterogeneity is attributable to variations in the local infection prevalence, as suggested by a dynamic model of disease (Bundy et al., 1991).

The analysis required information on the numbers of individuals in a given community with (a) microfilaraemia (mf) alone, (b) disease alone and (c) both mf and disease signs. An extensive literature survey located a total of 25 studies meeting this data requirement, although only 14 studies provided enough information to undertake separate analyses for hydrocele and lymphoedema. These surveys encompassed the major filariasis endemic regions (Indian subcontinent, Africa, the South Pacific islands and Brazil) and vector species, as well as a broad range of local infection prevalences. For each community, the association between mf and clinical disease was assessed via the 2x2 contingency table using odds ratio analysis (Fleiss, 1993). The odds ratio (OR) is a measure of the degree of association between mf and disease status, and denotes the odds of disease occurring in mf-positives relative to mf-negatives. The x2 test is employed as a test of independence, an OR of 1 indicating an equal chance of disease in mf-positives and mf-negatives. The immunological model of filarial infection and disease implies a significantly lower odds of disease in mf-positives (OR less than 1). A fixed effects meta-analysis was undertaken to compare and aggregate results from different studies and to evaluate the global evidence for the observed association between mf and disease. Succinctly, let 0I be the odds of disease occurring in mf-positives relative to mf-negatives, and let wi denote the reciprocal of its variance in the ith study (see standard formulae given in Hedges and Olkin, 1985; Fleiss, 1993). Then a good estimator of the assumed common underlying effect size is:

With an approximate 95% confidence interval for the estimate given by:

Michael and colleagues (1994) also tested for the existence of significant between-study heterogeneity in the relationship by constructing the statistic:

When effect sizes are homogeneous, Q follows a X2 distribution with (k-1) degrees of freedom, where k denotes the number of studies.

The results of the meta-analyses of the occurrence of combined chronic disease, hydrocele only (for males) and lymphoedema only in mf-positives, are displayed graphically in Figures 2.8-2.10. In each figure, the estimated ORs from

Fig 2.8 Odds ratios (mf-positives to mf-negatives) and 95% confidence intervals for 23 studies of the relation between the presence of mf and combined chronic disease (hydrocele and lymphoedema) in bancroftian filariasis. Two studies (marked with asterisk) provided data only for males. See text for explanation of the figure and interpretation of the results. aIvory Coast; bKoupela; cMali; dTingrela. For references, see source. From Michael et al. (1994), with permission

Fig 2.8 Odds ratios (mf-positives to mf-negatives) and 95% confidence intervals for 23 studies of the relation between the presence of mf and combined chronic disease (hydrocele and lymphoedema) in bancroftian filariasis. Two studies (marked with asterisk) provided data only for males. See text for explanation of the figure and interpretation of the results. aIvory Coast; bKoupela; cMali; dTingrela. For references, see source. From Michael et al. (1994), with permission individual studies are plotted in descending order of magnitude, together with their respective 95% confidence intervals. Ratios lying to the right of the unity line (OR> 1) denote a positive association, or a higher observed probability of disease in mf-positives. By contrast, an OR located to the left (OR< 1) represent a negative relationship for that study, with a higher chance of disease in mf-negatives. Studies in which the 95% confidence interval of the estimated OR include 1 signify equal chance of disease in their respective mf-positive and mf-negative populations. The results show that, contrary to the expectation of a negative association between mf and chronic disease, most studies had ORs that did not differ significantly from unity (12/21 for combined chronic disease, 8/14 for hydrocele, and 8/12 for lymphoedema), and thus provide no evidence for a significant association between the presence or absence of patent infection and the occurrence of disease. Indeed, the overall results suggest a bias towards a positive association, with more studies in each disease category showing significantly higher rather than lower odds of disease in mf-positives (Figures 2.8-2.10). However, for all three meta-analyses, there was significant between-study variability which precluded the computation of a common OR for these studies. Michael and colleagues, however, showed that although there could be regional effects, the observed between-study variability could be explained by the local incidence of infection; in general, there was a trend for the odds of patent infection in diseased individuals to increase positively with increasing prevalence of infection (Figure 2.11). The authors concluded that, on balance, these results supported the prediction of the dynamic model of disease (proportion of individuals with both chronic disease and microfilaraemia increase with increasing prevalence of infection because of higher probabilities of reinfection) rather than the immunological model of infection and disease development in lymphatic filariasis.

Fig 2.9 Meta-analysis of 14 studies of the relation between the presence of mf and hydroceles in males. The individual study odds ratios (mf-positives to mf-negatives) are plotted together with their 95% confidence intervals. aTingrela; bKoupela; cMali. For references, see source. From Michael et al. (1994), with permission

Fig 2.10 Meta-analysis of 14 studies. Two studies provided data for males only (marked with asterisk) of the relation between the presence of mf and lymphoedema. aKoupela; aTingrela. For references, see source. From Michael et al. (1994), with permission cö

Was this article helpful?

## Post a comment