- Email: [email protected]
Factors influencing the prevalence of trypanosome infection of Glossina pallidipes on the Ruvu flood plain of Eastern Tanzania
Acta Tropica 70 (1998) 143 – 155
Factors influencing the prevalence of trypanosome infection of Glossina pallidipes on the Ruvu flood plain of Eastern Tanzania A.R. Msangi a, C.J. Whitaker b, M.J. Lehane c,* a
Tsetse and Trypanosomiasis Research Institute, P.O. Box 1029, Tanga, Tanzania b Center for Applied Statistics, Uni6ersity of Wales, Bangor LL57 TUT, UK c School of Biological Sciences, Uni6ersity of Wales, Bangor LL57 2UW, UK
Received 6 September 1997; received in revised form 15 December 1997; accepted 28 January 1998
Abstract We report the pattern of infection of Glossina pallidipes with Trypanosoma 6i6ax and T. congolense at a site in the Coast region of eastern Tanzania, studied between November 1993 and December 1994. Of the 2315 flies dissected 114 (4.9%) were T. congolense positive, 77 (3.3%) were T. 6i6ax positive and 2 (0.1%) were T. brucei positive. Fly age was determined by the pteridine fluorescence method. Prevalence of infection was most strongly affected by month and the linear effect of age with the interaction of month and age having an effect for T. congolense-type infections. Sex and sex by month also have some predictive capacity when data for T. congolense and T. vivax-type infections are combined. In contrast to other similar studies our results suggest that the infection rate is non-linearly related to age of the tsetse fly, with older flies having progressively more chance of infection. The potential biological factors underpinning these interactions are discussed. © 1998 Published by Elsevier Science B.V. All rights reserved. Keywords: Glossina pallidipes; Age-prevalence; Trypanosoma congolense; Trypanosoma 6i6ax; Tanzania
* Corresponding author. E-mail: [email protected] 0001-706X/98/$19.00 © 1998 Published by Elsevier Science B.V. All rights reserved. PII S0001-706X(98)00013-8
144
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
1. Introduction It is still a widely held view that tsetse susceptibility to trypanosomes rapidly diminishes with fly age (Jordan, 1974; Molyneux, 1977; Maudlin, 1991). While this assertion describes the situation commonly observed under laboratory conditions there are several publications which suggest it may not accurately reflect the situation in the field. For example, Both Harley (1966) and Ryan et al. (1982) reported that prevalences of mature infections increased with fly age. One of the perennial difficulties in such studies in the field has been accurately determining the age of field caught tsetse flies. The wing fray method commonly used for male flies is not a reliable indicator of age under field conditions particularly if the age of individual insects is required in the study (Vale et al., 1976; Ryan et al., 1980). The ovarian dissection method used for females (Challier, 1965) produces accurate estimates of age through the first eight ovarian cycles—that is until the fly is approximately 70 days post-emergence. However, female tsetse flies in the field commonly achieve more than twice this age (Jordan, 1986) and these flies are wrongly ascribed to much younger age categories when the ovarian dissection age determination method is used. Consequently, a field trial was begun in Tanzania in 1993 in which we looked at the age-prevalence curve of field caught Glossina pallidipes infected with trypanosomes in a study where fly age was determined by the pteridine fluorescence method (Lehane and Mail, 1985; Lehane and Hargrove, 1988). and the results are reported here. These results are compared to two similar studies, which have been reported since we began our work, which used ovarian dissection and wing fray methods of age determination (Woolhouse et al., 1993, 1994). The study area was in a remote, sparsely populated part of the Coast region of Tanzania in which there were no cattle during the period of the study. Animal trypanosomiasis is common in cattle ranches adjoining the area (Msangi, unpublished results), human disease is rarely recorded.
2. Materials and methods
2.1. Study site and fly collection Field work was based at the laboratories of the Animal Diseases Research Institute (ADRI), Dar es Salaam. The field site was in the Kigogo area, Kisarawe district, Coast region, 75 km west of Dar es Salaam on the edge of the flood plain of the Ruvu river (6 49% S 38 37% W). The area is grassland, near to the river, with woodland and isolated dense thickets becoming increasingly common farther from the river. The site supports a small population of game animals including elephant, buffalo, lion, hyena, warthogs, bushpig, bushbuck, dik dik and a variety of other antelope; no cattle were present during the period of our investigations. Flies were collected from a single 500-m stretch of an earth road running though the mixed grassland/woodland. They were collected using three F3 traps and two Ngu traps (Flint, 1985) each baited with a 4-ml sachet containing 3-n-propylphenol, 4-
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
145
methylphenol and octenol at a ratio of 1:8:4 (Hargrove and Langley, 1990). To avoid undue fly mortality traps were set in the shade of trees and flies were collected within 2 – 3 h. G. morsitans morsitans, G. bre6ipalpis and G. pallidipes were trapped. Only G. pallidipes, which were caught in greater numbers in the traps, were retained for further study. Flies were stored alive at 4–5°C in the dark and returned to the laboratory for analysis. They were collected at least once a month from November 1993 to December 1994, except for April 1994, when the road was impassable due to rain. Rainfall and temperature records for the period were obtained from the meteorological station at Kibaha Sugarcane Research Institute, Coast Region.
2.2. Fly dissection and head collection Flies were dissected within 24 h of capture. Mouthparts were removed from the head capsule which was placed in a gelatin capsule containing dry silica gel wrapped in tissue and posted to the UK for pteridine analysis. Mouthparts, salivary glands and midguts were dissected in normal saline and screened under a light microscope for the presence of trypanosomes. Only mouthpart positive flies had their midgut dissected. Trypanosome infections were characterized as Trypanosoma brucei-type if the salivary gland was infected, T. congolense-type if salivary glands were negative but midgut and mouthparts were positive or T. 6i6ax-type if the salivary glands and midgut were negative but mouthparts were positive (Lloyd and Johnson, 1924). We are aware that the procedures may underestimate all infections because light infections may not be apparent and also that the technique will overestimate T. congolense-type infections and underestimate T. 6i6ax-type infections because of the confounding effect of immature midgut infections (Jordan, 1974).
2.3. Pteridine analysis Pteridine analysis was conducted on field flies within 6 weeks of their capture. The calibration curve was produced on flies of known age bought from the Tsetse Research Laboratories, Bristol, UK. Heads were homogenized in 3 ml of 0.1 N NaOH adjusted to pH 10 with glycine ( 11.5 g/l) and fluorescence was assayed using a Perkin-Elmer LS-3 fluorescence spectrometer with the excitation monochromator set at 360 nm and the emission monochromator set at 450 nm. Age of field caught flies was calculated using the linear regression equation of pteridine content on calendar age of the Bristol flies (Lehane and Mail, 1985; Lehane and Hargrove, 1988; Langley et al., 1988).
2.4. Data analysis All analyses were performed using (Genstat, 1993). Risk factors affecting infection were determined using stepwise logistic regression. The risk factors used here were categorical factors (e.g. sex, month) and metric variables (e.g. age, rainfall, temperature). The factor×factor× factor× variable interactions were also used.
146
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
Fig. 1. Mean monthly rainfall and temperature at Kibaha Sugar cane Research Institute 1993 – 1994.
The goodness of fit of the model is expressed as the deviance; the deviance is taken as having a x 2 distribution with the degrees of freedom shown. Here, no further risk factor was added to the stepwise model if the deviance explained was less than the degrees of freedom.
3. Results and analysis
3.1. Raw data The rainfall and temperature pattern for the study period is presented in Fig. 1. Infection rates are given in Table 1. Clearly no further analysis is possible for T. brucei type infections. Changes in the monthly prevalence of infection for T. congolense-type and T. 6i6ax-type infections are given in Fig. 2. The prevalence of infection with age is presented in Fig. 3. For the purposes of the analysis we have rounded gestation up to the 40th day, and pupation period (Jordan, 1986) to 2 months. Using this period along with fly age and date of capture we have separated the flies into cohorts dependent on their predicted month of conception. The prevalence of infection dependent on cohort is give in Fig. 4. Table 1 A summary of the infections discovered in the 1822 female and 493 male (total 2315) G. pallidipes dissected in the study
Number of infected flies Female flies infected (%) Male flies infected (%)
T. congolense
T. 6i6ax
T. brucei
114 (4 9%) 4.6 6.1
77 (3.3%) 3.3 3.4
2 (0.1%) 0.1
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
147
Fig. 2. Monthly prevalences of infection of G. pallidipes with T. congolense and T. 6i6ax infections. April data missing—site inaccessible.
3.2. Analysis A stepwise logistic regression of infection status on month, sex, month by sex interaction, the linear, quadratic and cubic effects of age, and the interactions of these age components with month and sex was performed. For all infections the analysis of deviance table produced is given in Table 2. This analysis suggests that month and the linear component of age and their interaction are the most important effects with sex and the sex by month interaction having a lesser effect. The remaining terms do not reach the 5% significance level. Repeating the analysis for just T. congolense-type infections gives the analysis of deviance results presented in Table 3. This analysis suggests that again month, the linear effect of age and their interaction have the greatest effect, with sex and the month by sex interaction having a lesser role. Repeating the analysis for just T. 6i6ax-type infections gives the analysis of deviance results presented in Table 4. This analysis suggests that only month and the linear effect of age affect infection.
Fig. 3. The prevalence of infection with age of G. pallidipes with T. congolense and T. 6i6ax.
148
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
Fig. 4. The prevalence of infection with cohort of G. pallidipes with T. congolense and T. 6i6ax.
Given that month and age are both risk factors in acquiring infection for T. congolense type and T. 6i6ax-type infections we decided to pursue the analysis further by looking at monthly birth cohort and the results are presented in Tables 5 and 6. While these analyses account for a larger deviance the df has increased for both T. congolense-type and T. 6i6ax-type infections. Also, as the month by cohort interaction is required for T. congolense-type infections the model is arguably not even simpler. We now tried to examine the effect of month and cohort more fully. We replaced the month factor with sine and cosine variables to model a cyclic pattern based on a period of 12 months. Equally, differences in the infection rates of each cohort are interesting but may be better explained if taken as a proxy variable for the climatic conditions occurring and so we looked at the climatic conditions in the month of conception and in the month of eclosion as well. When we replaced the month and cohort effects by the cyclic effect of month and by the climatic conditions in the month of eclosion and 2 months prior to eclosion (conception) we produce the analysis presented in part (b) of Tables 5 and 6. This shows that while these simpler variables can explain some of the variability due to month and cohort there is still a significant amount of variability that is unexplained. For the purposes of analysis, it was decided to stay with the models suggested from Tables 3 and 4. The major influence on infection is month. Month may be a proxy variable for a range of factors among which are temperature and rainfall. To investigate this we looked at rainfall and temperature in each of the 5 months prior to capture of the fly. In addition, we included a sine and cosine term in an attempt to model any periodic effect further to rainfall and temperature. Results are presented in Table 7. Clearly these factors do not sufficiently explain all the variability accounted for by month.
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
149
3.3. Model From the above analyses it is clear that month, the linear effect of age and its interaction with month best models T. congolense-type infections, while for T. 6i6ax-type infections month and the linear effect of age only have the best predictive power. From the logistic regression analyses developed here, the logit of the infection rate is linearly related to age. This means that the infection rate is not linearly related to age but shows an increasing rate of infection over the age range found here. This contrasts with the essentially linear effects noted by (Woolhouse et al., 1993, 1994). To see that our result is not just a consequence of using the logit transformation in the logistic regression analysis we present Fig. 5. This uses a generalized additive model (Hastie and Tibshirani, 1990) to visualize the data. We Table 2 Analysis of deviance table for a stepwise logistic regression analysis of all infections on month, sex and the linear, quadratic and cubic components of age Risk factor Month +Age (1) +Month. Age (1) +Sex +Sex.month +Age (2) +Sex.age(2) Error Total
Deviance 41 24 27 3.5 29 2.1 2.2
P
df
B0.001 B0.001 B0.01 \0.05 B0.01 \0.05 \0.05
1204 1328
12 1 12 1 12 1 1 2274 2314
Table 3 Analysis of deviance table for a stepwise logistic regression analysis of T. congolense type infections on month, sex and the linear, quadratic and cubic components of age Risk factor
Deviance
P
df
Month +Age (1) +Month. Age (1)
31 17 32
B0.01 B0.001 B0.01
12 1 12
Sum +Sex +Sex.month +Age (2) +Month. Age (2) +Sex.age(2)
80 1.8 26 1.6 14 2.1
\0.05 B0.05 \0.05 \0.05 \0.05
25 1 12 1 12 1
Error Total
783 909
2262 2314
150
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
Table 4 Analysis of deviance table for a stepwise logistic regression analysis of T. vivax type infections on month, sex and the linear, quadratic and cubic components of age Risk factor Month +Age (1)
Deviance
df
P B0.001 B0.05
40 5.0
12 1
Sum Error
45 630
13 2301
Total
675
2314
fitted a cubic spline function for age with ten effective degrees of freedom, together with a linear regression model. The graphs show the relationship between infection and age without the restriction of the logit transformation.
3.4. Type of infection The 191 flies with either T. congolense- or T. 6i6ax-type infection were examined. We analyzed the data to see if the proportion of T. congolense-type infections (as a proportion of total infections) was related to month, age, sex and their interactions. We also tried replacing month by the climatic effects in the month of capture and the previous month. The results are presented in Table 8. The analysis shows that month is the only significant predictor of the proportion of T. congolense-type infections. However, in this analysis it is possible to replace month by a cyclic pattern while still explaining a large part of the deviance. Table 5 Analysis of deviance table for a stepwise logistic regression analysis of T. congolense-type infections Risk factor
Deviance
P
df
(a) Month +Cohort +Month. Cohort Sum
31 38 43 112
B0.01 B0.01 B0.01
12 17 41 70
B0.001 B0.01 B0.05 \0.05 \0.05 B0.05
1 1 1 1 1 1 6
(b) +Age (1) +(Rain birth) +Sex +(Rain birth).sex +(Temp eclosion) +(Temp.eclosion). sex Sum
14 10 4.5 3.3 1.5 5.1 39
(a) month, sex, cohort and the linear, quadratic and cubic components of age and (b) sex, monthly cycle, the linear components of age, the rain and temperature at conception and the rain and temperature at eclosion.
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
151
Table 6 Analysis of deviance table for a stepwise logistic regression analysis of T. 6i6ax-type infections Risk factor
Deviance
P
df
(a) Month +Cohort Sum
40 26 66
B0.001 \0.05
12 17 29
(b) Cyclic +(Rain birth) +Sex. Cyclic Sum
22 1.7 2.6 27
B0.001 \0.05 \0.05
2 1 2 5
(a) month, sex, cohort and the linear, quadratic and cubic components of age and (b) sex, monthly cycle, the linear components of age, the rain and temperature at conception and the rain and temperature at eclosion.
4. Discussion We are aware that the trypanosome identification procedures used will overestimate T. congolense-type infections and underestimate T. 6i6ax-type infections because of the confounding effect of immature midgut infections. In a series of comparable studies in Zimbabwe by Woolhouse et al. (1993) and in Zambia by Woolhouse et al. (1994) molecular techniques were used to confirm identification. It was estimated that 84 and 94%, respectively, of midgut positive flies supported T. congolense infections. So the available evidence suggests the technique adopted here will provide reasonable estimates of infection type. In the Zimbabwe and Zambia Table 7 Analysis of deviance table for a stepwise logistic regression analysis of infection types infections on the linear componentsa Risk factor
Deviance
P
df
Age (1) +Rain (2) +Sex +Rain (0) Sum
14 17 4.9 3.2 39
B0.001 B0.001 B0.05 \0.05
1 1 1 1 4
(b) Cyclic +Temp (2) +Age (1) +Rain (5) Sum
22 6.2 3.6 3.5 35
B0.01 B0.05 \0.05 \0.05
2 1 1 1 5
a Age, sex, monthly cycle and rain and temperature during the 5 months prior to capture for (a) T. congolense type and (b) T. 6i6ax type infections.
152
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
Fig. 5. The smoothed prevalences of infection with age of G. pallidipes with T. congolense (-) and T. 6i6ax (--).
studies, the tsetse fly investigated was also G. pallidipes. The overall infection pattern in these studies showed that T. 6i6ax-type infections (3.1 and 6.2%, respectively) were more common than T. congolense-type infections (2.4 and 3.0%, Table 8 Analysis of deviance table for a stepwise logistic regression analysis of T. congolense type infections as a proportion of total T. congolense and T. 6i6ax type infections Risk factor
Deviance
P
df
(a) Month +Age (2) +Month.age (2) Sum Error Total
32 1.3 15 48 209 258
B0.001 \0.05 \0.05
11 1 11 23 167 190
19 4.5
B0.001 \0.05
2 2
(b) Cyclic +Sex.cyclic
(a) Sex, month and the linear, quadratic and cubic components of age and (b) sex, monthly cycle and the linear components of age.
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
153
respectively). In contrast, in this study, T. congolense-type infections (4.9%) were more common than T. 6i6ax-type infections (3.3%). T. brucei-type infections were least common in all three studies ( 0.1% in all studies). The studies performed here are comparable to those of Woolhouse et al. (1993, 1994) with the major exception that these authors divided age into ovarian categories and we used the pteridine method of age determination. In consequence the maximum age category considered here is 185 days with 836 (36%) of the ages being over 70 days. This compares to the maximum age category of 62–70 days for the studies mentioned above. The reason for the difference is that the ovarian age determination technique assigns all female flies older than ovarian category 7 to a younger, and hence wrong ovarian category. For comparison purposes, we grouped our measure of age into the eight categories of Woolhouse et al. (1993, 1994) and repeated the analyses. Under these circumstances for T. congolense-type infections only month was found to be a significant factor, accounting for a deviance of 31 with 12 df., for T. 6i6ax-type infections month and age were found to be significant factors, accounting for deviances of 31 and 5, respectively, with 12 and 1 df, respectively. From these differences seen in the T. congolense-type infections it is clear that use of the ovarian age determination method, with the attendant misclassification of many older flies, can hide the effects of age on prevalence of infection. The major effect on prevalence of infection in this study is month and the linear effect of age for both T. congolense-type and T. 6i6ax-type infections. In addition for T. congolense-type infections the interaction of month and age has an effect. However, there is evidence that sex and the sex by month interaction have some capacity to predict infection when considering the combined data for T. 6i6ax and T. congolense (Table 2). It is interesting to notice, as just explained above, that if we condense our age data into categories typical of studies using ovarian age grading techniques (Woolhouse et al., 1993, 1994) we lose most effects except month with age remaining as a marginally significant factor for only T. 6i6ax-type infections. This change in the relative importance of age in the analysis suggests that age may have been a more significant factor influencing infection in the previous studies mentioned above than suggested in the analysis of those studies because of the artificial placement of older flies into younger age categories. Thus, in this study, 41% of female flies are aged more than 71 days and would be wrongly age graded when using ovarian age grading methods. Similarly, it is possible that the difference in age determination techniques used may explain why, in this study, we see evidence for some effects related to fly sex in contrast to previous studies (Mohamed-Ahmed et al., 1989; Woolhouse et al., 1993, 1994). This would occur because ovarian-based age determination methods artificially condense older flies into younger age categories hiding the longer life span of female compared to male flies. Thus, in this study 41% of females were aged more than 71 days compared to only 18% of males. Evidence for an effect of sex on infection rate has been gathered in laboratory studies (Maudlin, 1991). The accuracy of the pteridine age grading method has been established in field trials (Lehane and Hargrove, 1988; Langley et al., 1988). These studies suggest a S.D. of 9 5–10 days is to be expected with the
154
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
technique which, we believe, explains the presence of infected flies in our earliest age groupings. Because there are inaccuracies in age grading the effect of age on prevalence of infection is likely to be greater than predicted here. The lowest proportion of infected flies occurs at the end of the long rains and an essentially similar situation was found in Zimbabwe and Zambia (Woolhouse et al., 1993, 1994) but not in the Ivory Coast (Ryan et al., 1982). The possible significance of this observation is open to debate (for discussion see Woolhouse et al., 1993). In addition to temperature and rainfall month may also act as a proxy variable for several other factors. One of the most important may be host movements which may well have a profound impact on the trypanosome type and the rate of encounter which is likely to influence the pattern of infection. A more general cycling in host numbers, irrespective of infection levels in these hosts, will also influence fly feeding rates and it has been reported that depriving normally refractory G. m. morsitans of a meal for 3 – 4 days prior to an infected meal, greatly enhances the chances of the fly becoming infected (Gingrich et al., 1982; Makumyaviri et al., 1984). Thus, changes in the availability of blood meals through the year may also influence the pattern of infection and may underlie the month effect. The temporal pattern of infection of flies with both trypanosomes shows a similar, but not identical, pattern. The increase of prevalence of infection with age (Fig. 5) is consistent with the findings in other field studies (Harley, 1966; Ryan et al., 1982; Mohamed-Ahmed et al., 1989; Woolhouse et al., 1993, 1994). Fig. 5 shows that there is a noticeable increase in the prevalence of infection in flies over 125 and 150 days of age for T. congolense and T. 6i6ax type infections, respectively. While only small numbers of flies are found at these ages (90 and 27 aged over 125 and 150 days, respectively) there is still the finding that there is a higher prevalence of infection. How much the shape of this curve is influenced by the different age determination methods employed remains to be determined. Two obvious means by which increasing rates of infection with age may be caused are through increased survivorship of infected flies or through a gradual increase in the proportion of the population which is susceptible to infection. This latter seems a possibility either through selective loss of refractory flies or through erosion of the refractory status of the surviving population. Further studies are required to determine the mechanisms underlying the patterns of infection presented.
References Challier, 1965. AUTHOR PLEASE CITE FULL REFERENCE. Flint, S., 1985. A comparison of various traps for Glossina spp. (Glossinidae) and other Dipteral. Bull. Entomol. Res. 75, 529–534. Genstat, 1993. Genstat 5 Release 3 Reference Manual. Oxford Scientific, London. Gingrich, J.B., Ward, R.A., Macken, L.M., Esser, K.M., 1982. African sleeping sickness: new evidence that mature tsetse flies Glossina morsitans can become potent vectors. Trans. R. Soc. Trop. Med. Hyg. 76, 479–481.
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
155
Hargrove, J.W., Langley, P., 1990. Sterilizing tsetse in the field: a successful field trial. Bull. Entomol. Res. 80, 397–403. Harley, J. M. B., 1966. Further studies on age and trypanosome infection rates in Glossinapallidipes Aust., G. palpalisfuscipes Newst. and G. bre6ipalpis Newst. in Uganda. Bull. Entomol. Res. 57, 459–477. Hastie, T.J., Tibshirani, R.J., 1990. Generalised Additive Models. Chapman and Hall, London. Jordan, A.M., 1974. Recent developments in the ecology and methods of control of tsetse flies (Glossina spp.) (Dipt., Glossinidae)—a review. Bull. Entomol. Res. 63, 361 – 399. Jordan, A. M., 1986. Trypanosomiasis Control and African Rural Development. Longman, London. Langley, P.A., Hall, M.J.R., Felton, T., Ceesay, M., 1988. Determining the age of tsetse flies, Glossina spp. Diptera: Glossinidae: an appraisal of the pteridine fluorescence technique. Bull. Entomol. Res. 78, 387–395. Lehane, M.J., Hargrove, J.W., 1988. Field experiments on a new method for determining age in tsetse flies Diptera: Glossinidae. Ecol. Entomol. 13, 319 – 322. Lehane, M.J., Mail, T.S., 1985. Determining the age of adult male and female Glossina morsitans morsitans using a new technique. Ecol. Entomol. 10, 219 – 224. Lloyd, L.L., Johnson, W.B., 1924. The trypanosome infections of tsetse flies in Northern Nigeria and a method of estimation. Bull. Entomol. Res. 14, 265 – 288. Makumyaviri, A.M., Demey, F., Claes, Y., Verhulst, A., Le Ray, D., 1984. Caracterisation de la capacite vectorielle de Glossina morsitans morsitans Diptera: Glossinidae vis a vis de Trypanosoma brucei brucei EATRO 1125 AnTAR 1. Ann. Soc. Belg. Med. Trop. 64, 365 – 372. Maudlin, I., 1991. Transmission of African trypanosomiasis: interactions among tsetse immune system, symbionts and parasites. Adv. Dis. Vector Res. 7, 117 – 148. Mohamed-Ahmed, M.M., Ahmed, A.I., Ishag, A., 1989. Trypanosome infection rate of Glossina morsitans submorsitans in Bahr el Arab, South Darfur Province, Sudan. Trop. Ann. Health Prod. 21, 239–244. Molyneux, D. H., 1977. Vector relationships in the Trypanosomatidae. Adv. Parasitol. 15, 1 – 82. Ryan, L., Molyneux, D.H., Kuzoe, F.A.S., 1980. Differences in rate of wing fray between Glossina species. Tropenmed. Parasitol. 31, 111 – 116. Ryan, L., Kupper, W., Goff, S. L., Molyneux, D. H., Clair, M., 1982. Differences in rates of acquisition of trypanosome infections between Glossina species in the field. Annal. Soc. Belg. Med. Trop. 41, 355–359. Vale, G.A., Hargrove, J.W., Jordan, A.M., Langley, P.A., Mews, A.R., 1976. Survival and behaviour of tsetse flies (Diptera: Glossinidae) released in the field: a comparison between wild flies and and animal-fed and in vitro-fed laboratory reared flies. Bull. Entomol. Res. 66, 731 – 744. Woolhouse, M.E.J., Hargrove, J.W., McNamara, J.J., 1993. Epidemiology of trypanosome infections of the tsetse fly Glossina pallidipes in the Zambezi valley. Parasitology 106, 479 – 485. Woolhouse, M.E.J., Bealby, K., McNamara, J.J., Silutongwe, J., 1994. Trypanosome infections of the tsetse fly Glossina pallidipes in the Luangwa valley. Int. J. Parasitol. 24, 987 – 993.
.
Factors influencing the prevalence of trypanosome infection of Glossina pallidipes on the Ruvu flood plain of Eastern Tanzania A.R. Msangi a, C.J. Whitaker b, M.J. Lehane c,* a
Tsetse and Trypanosomiasis Research Institute, P.O. Box 1029, Tanga, Tanzania b Center for Applied Statistics, Uni6ersity of Wales, Bangor LL57 TUT, UK c School of Biological Sciences, Uni6ersity of Wales, Bangor LL57 2UW, UK
Received 6 September 1997; received in revised form 15 December 1997; accepted 28 January 1998
Abstract We report the pattern of infection of Glossina pallidipes with Trypanosoma 6i6ax and T. congolense at a site in the Coast region of eastern Tanzania, studied between November 1993 and December 1994. Of the 2315 flies dissected 114 (4.9%) were T. congolense positive, 77 (3.3%) were T. 6i6ax positive and 2 (0.1%) were T. brucei positive. Fly age was determined by the pteridine fluorescence method. Prevalence of infection was most strongly affected by month and the linear effect of age with the interaction of month and age having an effect for T. congolense-type infections. Sex and sex by month also have some predictive capacity when data for T. congolense and T. vivax-type infections are combined. In contrast to other similar studies our results suggest that the infection rate is non-linearly related to age of the tsetse fly, with older flies having progressively more chance of infection. The potential biological factors underpinning these interactions are discussed. © 1998 Published by Elsevier Science B.V. All rights reserved. Keywords: Glossina pallidipes; Age-prevalence; Trypanosoma congolense; Trypanosoma 6i6ax; Tanzania
* Corresponding author. E-mail: [email protected] 0001-706X/98/$19.00 © 1998 Published by Elsevier Science B.V. All rights reserved. PII S0001-706X(98)00013-8
144
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
1. Introduction It is still a widely held view that tsetse susceptibility to trypanosomes rapidly diminishes with fly age (Jordan, 1974; Molyneux, 1977; Maudlin, 1991). While this assertion describes the situation commonly observed under laboratory conditions there are several publications which suggest it may not accurately reflect the situation in the field. For example, Both Harley (1966) and Ryan et al. (1982) reported that prevalences of mature infections increased with fly age. One of the perennial difficulties in such studies in the field has been accurately determining the age of field caught tsetse flies. The wing fray method commonly used for male flies is not a reliable indicator of age under field conditions particularly if the age of individual insects is required in the study (Vale et al., 1976; Ryan et al., 1980). The ovarian dissection method used for females (Challier, 1965) produces accurate estimates of age through the first eight ovarian cycles—that is until the fly is approximately 70 days post-emergence. However, female tsetse flies in the field commonly achieve more than twice this age (Jordan, 1986) and these flies are wrongly ascribed to much younger age categories when the ovarian dissection age determination method is used. Consequently, a field trial was begun in Tanzania in 1993 in which we looked at the age-prevalence curve of field caught Glossina pallidipes infected with trypanosomes in a study where fly age was determined by the pteridine fluorescence method (Lehane and Mail, 1985; Lehane and Hargrove, 1988). and the results are reported here. These results are compared to two similar studies, which have been reported since we began our work, which used ovarian dissection and wing fray methods of age determination (Woolhouse et al., 1993, 1994). The study area was in a remote, sparsely populated part of the Coast region of Tanzania in which there were no cattle during the period of the study. Animal trypanosomiasis is common in cattle ranches adjoining the area (Msangi, unpublished results), human disease is rarely recorded.
2. Materials and methods
2.1. Study site and fly collection Field work was based at the laboratories of the Animal Diseases Research Institute (ADRI), Dar es Salaam. The field site was in the Kigogo area, Kisarawe district, Coast region, 75 km west of Dar es Salaam on the edge of the flood plain of the Ruvu river (6 49% S 38 37% W). The area is grassland, near to the river, with woodland and isolated dense thickets becoming increasingly common farther from the river. The site supports a small population of game animals including elephant, buffalo, lion, hyena, warthogs, bushpig, bushbuck, dik dik and a variety of other antelope; no cattle were present during the period of our investigations. Flies were collected from a single 500-m stretch of an earth road running though the mixed grassland/woodland. They were collected using three F3 traps and two Ngu traps (Flint, 1985) each baited with a 4-ml sachet containing 3-n-propylphenol, 4-
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
145
methylphenol and octenol at a ratio of 1:8:4 (Hargrove and Langley, 1990). To avoid undue fly mortality traps were set in the shade of trees and flies were collected within 2 – 3 h. G. morsitans morsitans, G. bre6ipalpis and G. pallidipes were trapped. Only G. pallidipes, which were caught in greater numbers in the traps, were retained for further study. Flies were stored alive at 4–5°C in the dark and returned to the laboratory for analysis. They were collected at least once a month from November 1993 to December 1994, except for April 1994, when the road was impassable due to rain. Rainfall and temperature records for the period were obtained from the meteorological station at Kibaha Sugarcane Research Institute, Coast Region.
2.2. Fly dissection and head collection Flies were dissected within 24 h of capture. Mouthparts were removed from the head capsule which was placed in a gelatin capsule containing dry silica gel wrapped in tissue and posted to the UK for pteridine analysis. Mouthparts, salivary glands and midguts were dissected in normal saline and screened under a light microscope for the presence of trypanosomes. Only mouthpart positive flies had their midgut dissected. Trypanosome infections were characterized as Trypanosoma brucei-type if the salivary gland was infected, T. congolense-type if salivary glands were negative but midgut and mouthparts were positive or T. 6i6ax-type if the salivary glands and midgut were negative but mouthparts were positive (Lloyd and Johnson, 1924). We are aware that the procedures may underestimate all infections because light infections may not be apparent and also that the technique will overestimate T. congolense-type infections and underestimate T. 6i6ax-type infections because of the confounding effect of immature midgut infections (Jordan, 1974).
2.3. Pteridine analysis Pteridine analysis was conducted on field flies within 6 weeks of their capture. The calibration curve was produced on flies of known age bought from the Tsetse Research Laboratories, Bristol, UK. Heads were homogenized in 3 ml of 0.1 N NaOH adjusted to pH 10 with glycine ( 11.5 g/l) and fluorescence was assayed using a Perkin-Elmer LS-3 fluorescence spectrometer with the excitation monochromator set at 360 nm and the emission monochromator set at 450 nm. Age of field caught flies was calculated using the linear regression equation of pteridine content on calendar age of the Bristol flies (Lehane and Mail, 1985; Lehane and Hargrove, 1988; Langley et al., 1988).
2.4. Data analysis All analyses were performed using (Genstat, 1993). Risk factors affecting infection were determined using stepwise logistic regression. The risk factors used here were categorical factors (e.g. sex, month) and metric variables (e.g. age, rainfall, temperature). The factor×factor× factor× variable interactions were also used.
146
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
Fig. 1. Mean monthly rainfall and temperature at Kibaha Sugar cane Research Institute 1993 – 1994.
The goodness of fit of the model is expressed as the deviance; the deviance is taken as having a x 2 distribution with the degrees of freedom shown. Here, no further risk factor was added to the stepwise model if the deviance explained was less than the degrees of freedom.
3. Results and analysis
3.1. Raw data The rainfall and temperature pattern for the study period is presented in Fig. 1. Infection rates are given in Table 1. Clearly no further analysis is possible for T. brucei type infections. Changes in the monthly prevalence of infection for T. congolense-type and T. 6i6ax-type infections are given in Fig. 2. The prevalence of infection with age is presented in Fig. 3. For the purposes of the analysis we have rounded gestation up to the 40th day, and pupation period (Jordan, 1986) to 2 months. Using this period along with fly age and date of capture we have separated the flies into cohorts dependent on their predicted month of conception. The prevalence of infection dependent on cohort is give in Fig. 4. Table 1 A summary of the infections discovered in the 1822 female and 493 male (total 2315) G. pallidipes dissected in the study
Number of infected flies Female flies infected (%) Male flies infected (%)
T. congolense
T. 6i6ax
T. brucei
114 (4 9%) 4.6 6.1
77 (3.3%) 3.3 3.4
2 (0.1%) 0.1
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
147
Fig. 2. Monthly prevalences of infection of G. pallidipes with T. congolense and T. 6i6ax infections. April data missing—site inaccessible.
3.2. Analysis A stepwise logistic regression of infection status on month, sex, month by sex interaction, the linear, quadratic and cubic effects of age, and the interactions of these age components with month and sex was performed. For all infections the analysis of deviance table produced is given in Table 2. This analysis suggests that month and the linear component of age and their interaction are the most important effects with sex and the sex by month interaction having a lesser effect. The remaining terms do not reach the 5% significance level. Repeating the analysis for just T. congolense-type infections gives the analysis of deviance results presented in Table 3. This analysis suggests that again month, the linear effect of age and their interaction have the greatest effect, with sex and the month by sex interaction having a lesser role. Repeating the analysis for just T. 6i6ax-type infections gives the analysis of deviance results presented in Table 4. This analysis suggests that only month and the linear effect of age affect infection.
Fig. 3. The prevalence of infection with age of G. pallidipes with T. congolense and T. 6i6ax.
148
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
Fig. 4. The prevalence of infection with cohort of G. pallidipes with T. congolense and T. 6i6ax.
Given that month and age are both risk factors in acquiring infection for T. congolense type and T. 6i6ax-type infections we decided to pursue the analysis further by looking at monthly birth cohort and the results are presented in Tables 5 and 6. While these analyses account for a larger deviance the df has increased for both T. congolense-type and T. 6i6ax-type infections. Also, as the month by cohort interaction is required for T. congolense-type infections the model is arguably not even simpler. We now tried to examine the effect of month and cohort more fully. We replaced the month factor with sine and cosine variables to model a cyclic pattern based on a period of 12 months. Equally, differences in the infection rates of each cohort are interesting but may be better explained if taken as a proxy variable for the climatic conditions occurring and so we looked at the climatic conditions in the month of conception and in the month of eclosion as well. When we replaced the month and cohort effects by the cyclic effect of month and by the climatic conditions in the month of eclosion and 2 months prior to eclosion (conception) we produce the analysis presented in part (b) of Tables 5 and 6. This shows that while these simpler variables can explain some of the variability due to month and cohort there is still a significant amount of variability that is unexplained. For the purposes of analysis, it was decided to stay with the models suggested from Tables 3 and 4. The major influence on infection is month. Month may be a proxy variable for a range of factors among which are temperature and rainfall. To investigate this we looked at rainfall and temperature in each of the 5 months prior to capture of the fly. In addition, we included a sine and cosine term in an attempt to model any periodic effect further to rainfall and temperature. Results are presented in Table 7. Clearly these factors do not sufficiently explain all the variability accounted for by month.
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
149
3.3. Model From the above analyses it is clear that month, the linear effect of age and its interaction with month best models T. congolense-type infections, while for T. 6i6ax-type infections month and the linear effect of age only have the best predictive power. From the logistic regression analyses developed here, the logit of the infection rate is linearly related to age. This means that the infection rate is not linearly related to age but shows an increasing rate of infection over the age range found here. This contrasts with the essentially linear effects noted by (Woolhouse et al., 1993, 1994). To see that our result is not just a consequence of using the logit transformation in the logistic regression analysis we present Fig. 5. This uses a generalized additive model (Hastie and Tibshirani, 1990) to visualize the data. We Table 2 Analysis of deviance table for a stepwise logistic regression analysis of all infections on month, sex and the linear, quadratic and cubic components of age Risk factor Month +Age (1) +Month. Age (1) +Sex +Sex.month +Age (2) +Sex.age(2) Error Total
Deviance 41 24 27 3.5 29 2.1 2.2
P
df
B0.001 B0.001 B0.01 \0.05 B0.01 \0.05 \0.05
1204 1328
12 1 12 1 12 1 1 2274 2314
Table 3 Analysis of deviance table for a stepwise logistic regression analysis of T. congolense type infections on month, sex and the linear, quadratic and cubic components of age Risk factor
Deviance
P
df
Month +Age (1) +Month. Age (1)
31 17 32
B0.01 B0.001 B0.01
12 1 12
Sum +Sex +Sex.month +Age (2) +Month. Age (2) +Sex.age(2)
80 1.8 26 1.6 14 2.1
\0.05 B0.05 \0.05 \0.05 \0.05
25 1 12 1 12 1
Error Total
783 909
2262 2314
150
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
Table 4 Analysis of deviance table for a stepwise logistic regression analysis of T. vivax type infections on month, sex and the linear, quadratic and cubic components of age Risk factor Month +Age (1)
Deviance
df
P B0.001 B0.05
40 5.0
12 1
Sum Error
45 630
13 2301
Total
675
2314
fitted a cubic spline function for age with ten effective degrees of freedom, together with a linear regression model. The graphs show the relationship between infection and age without the restriction of the logit transformation.
3.4. Type of infection The 191 flies with either T. congolense- or T. 6i6ax-type infection were examined. We analyzed the data to see if the proportion of T. congolense-type infections (as a proportion of total infections) was related to month, age, sex and their interactions. We also tried replacing month by the climatic effects in the month of capture and the previous month. The results are presented in Table 8. The analysis shows that month is the only significant predictor of the proportion of T. congolense-type infections. However, in this analysis it is possible to replace month by a cyclic pattern while still explaining a large part of the deviance. Table 5 Analysis of deviance table for a stepwise logistic regression analysis of T. congolense-type infections Risk factor
Deviance
P
df
(a) Month +Cohort +Month. Cohort Sum
31 38 43 112
B0.01 B0.01 B0.01
12 17 41 70
B0.001 B0.01 B0.05 \0.05 \0.05 B0.05
1 1 1 1 1 1 6
(b) +Age (1) +(Rain birth) +Sex +(Rain birth).sex +(Temp eclosion) +(Temp.eclosion). sex Sum
14 10 4.5 3.3 1.5 5.1 39
(a) month, sex, cohort and the linear, quadratic and cubic components of age and (b) sex, monthly cycle, the linear components of age, the rain and temperature at conception and the rain and temperature at eclosion.
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
151
Table 6 Analysis of deviance table for a stepwise logistic regression analysis of T. 6i6ax-type infections Risk factor
Deviance
P
df
(a) Month +Cohort Sum
40 26 66
B0.001 \0.05
12 17 29
(b) Cyclic +(Rain birth) +Sex. Cyclic Sum
22 1.7 2.6 27
B0.001 \0.05 \0.05
2 1 2 5
(a) month, sex, cohort and the linear, quadratic and cubic components of age and (b) sex, monthly cycle, the linear components of age, the rain and temperature at conception and the rain and temperature at eclosion.
4. Discussion We are aware that the trypanosome identification procedures used will overestimate T. congolense-type infections and underestimate T. 6i6ax-type infections because of the confounding effect of immature midgut infections. In a series of comparable studies in Zimbabwe by Woolhouse et al. (1993) and in Zambia by Woolhouse et al. (1994) molecular techniques were used to confirm identification. It was estimated that 84 and 94%, respectively, of midgut positive flies supported T. congolense infections. So the available evidence suggests the technique adopted here will provide reasonable estimates of infection type. In the Zimbabwe and Zambia Table 7 Analysis of deviance table for a stepwise logistic regression analysis of infection types infections on the linear componentsa Risk factor
Deviance
P
df
Age (1) +Rain (2) +Sex +Rain (0) Sum
14 17 4.9 3.2 39
B0.001 B0.001 B0.05 \0.05
1 1 1 1 4
(b) Cyclic +Temp (2) +Age (1) +Rain (5) Sum
22 6.2 3.6 3.5 35
B0.01 B0.05 \0.05 \0.05
2 1 1 1 5
a Age, sex, monthly cycle and rain and temperature during the 5 months prior to capture for (a) T. congolense type and (b) T. 6i6ax type infections.
152
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
Fig. 5. The smoothed prevalences of infection with age of G. pallidipes with T. congolense (-) and T. 6i6ax (--).
studies, the tsetse fly investigated was also G. pallidipes. The overall infection pattern in these studies showed that T. 6i6ax-type infections (3.1 and 6.2%, respectively) were more common than T. congolense-type infections (2.4 and 3.0%, Table 8 Analysis of deviance table for a stepwise logistic regression analysis of T. congolense type infections as a proportion of total T. congolense and T. 6i6ax type infections Risk factor
Deviance
P
df
(a) Month +Age (2) +Month.age (2) Sum Error Total
32 1.3 15 48 209 258
B0.001 \0.05 \0.05
11 1 11 23 167 190
19 4.5
B0.001 \0.05
2 2
(b) Cyclic +Sex.cyclic
(a) Sex, month and the linear, quadratic and cubic components of age and (b) sex, monthly cycle and the linear components of age.
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
153
respectively). In contrast, in this study, T. congolense-type infections (4.9%) were more common than T. 6i6ax-type infections (3.3%). T. brucei-type infections were least common in all three studies ( 0.1% in all studies). The studies performed here are comparable to those of Woolhouse et al. (1993, 1994) with the major exception that these authors divided age into ovarian categories and we used the pteridine method of age determination. In consequence the maximum age category considered here is 185 days with 836 (36%) of the ages being over 70 days. This compares to the maximum age category of 62–70 days for the studies mentioned above. The reason for the difference is that the ovarian age determination technique assigns all female flies older than ovarian category 7 to a younger, and hence wrong ovarian category. For comparison purposes, we grouped our measure of age into the eight categories of Woolhouse et al. (1993, 1994) and repeated the analyses. Under these circumstances for T. congolense-type infections only month was found to be a significant factor, accounting for a deviance of 31 with 12 df., for T. 6i6ax-type infections month and age were found to be significant factors, accounting for deviances of 31 and 5, respectively, with 12 and 1 df, respectively. From these differences seen in the T. congolense-type infections it is clear that use of the ovarian age determination method, with the attendant misclassification of many older flies, can hide the effects of age on prevalence of infection. The major effect on prevalence of infection in this study is month and the linear effect of age for both T. congolense-type and T. 6i6ax-type infections. In addition for T. congolense-type infections the interaction of month and age has an effect. However, there is evidence that sex and the sex by month interaction have some capacity to predict infection when considering the combined data for T. 6i6ax and T. congolense (Table 2). It is interesting to notice, as just explained above, that if we condense our age data into categories typical of studies using ovarian age grading techniques (Woolhouse et al., 1993, 1994) we lose most effects except month with age remaining as a marginally significant factor for only T. 6i6ax-type infections. This change in the relative importance of age in the analysis suggests that age may have been a more significant factor influencing infection in the previous studies mentioned above than suggested in the analysis of those studies because of the artificial placement of older flies into younger age categories. Thus, in this study, 41% of female flies are aged more than 71 days and would be wrongly age graded when using ovarian age grading methods. Similarly, it is possible that the difference in age determination techniques used may explain why, in this study, we see evidence for some effects related to fly sex in contrast to previous studies (Mohamed-Ahmed et al., 1989; Woolhouse et al., 1993, 1994). This would occur because ovarian-based age determination methods artificially condense older flies into younger age categories hiding the longer life span of female compared to male flies. Thus, in this study 41% of females were aged more than 71 days compared to only 18% of males. Evidence for an effect of sex on infection rate has been gathered in laboratory studies (Maudlin, 1991). The accuracy of the pteridine age grading method has been established in field trials (Lehane and Hargrove, 1988; Langley et al., 1988). These studies suggest a S.D. of 9 5–10 days is to be expected with the
154
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
technique which, we believe, explains the presence of infected flies in our earliest age groupings. Because there are inaccuracies in age grading the effect of age on prevalence of infection is likely to be greater than predicted here. The lowest proportion of infected flies occurs at the end of the long rains and an essentially similar situation was found in Zimbabwe and Zambia (Woolhouse et al., 1993, 1994) but not in the Ivory Coast (Ryan et al., 1982). The possible significance of this observation is open to debate (for discussion see Woolhouse et al., 1993). In addition to temperature and rainfall month may also act as a proxy variable for several other factors. One of the most important may be host movements which may well have a profound impact on the trypanosome type and the rate of encounter which is likely to influence the pattern of infection. A more general cycling in host numbers, irrespective of infection levels in these hosts, will also influence fly feeding rates and it has been reported that depriving normally refractory G. m. morsitans of a meal for 3 – 4 days prior to an infected meal, greatly enhances the chances of the fly becoming infected (Gingrich et al., 1982; Makumyaviri et al., 1984). Thus, changes in the availability of blood meals through the year may also influence the pattern of infection and may underlie the month effect. The temporal pattern of infection of flies with both trypanosomes shows a similar, but not identical, pattern. The increase of prevalence of infection with age (Fig. 5) is consistent with the findings in other field studies (Harley, 1966; Ryan et al., 1982; Mohamed-Ahmed et al., 1989; Woolhouse et al., 1993, 1994). Fig. 5 shows that there is a noticeable increase in the prevalence of infection in flies over 125 and 150 days of age for T. congolense and T. 6i6ax type infections, respectively. While only small numbers of flies are found at these ages (90 and 27 aged over 125 and 150 days, respectively) there is still the finding that there is a higher prevalence of infection. How much the shape of this curve is influenced by the different age determination methods employed remains to be determined. Two obvious means by which increasing rates of infection with age may be caused are through increased survivorship of infected flies or through a gradual increase in the proportion of the population which is susceptible to infection. This latter seems a possibility either through selective loss of refractory flies or through erosion of the refractory status of the surviving population. Further studies are required to determine the mechanisms underlying the patterns of infection presented.
References Challier, 1965. AUTHOR PLEASE CITE FULL REFERENCE. Flint, S., 1985. A comparison of various traps for Glossina spp. (Glossinidae) and other Dipteral. Bull. Entomol. Res. 75, 529–534. Genstat, 1993. Genstat 5 Release 3 Reference Manual. Oxford Scientific, London. Gingrich, J.B., Ward, R.A., Macken, L.M., Esser, K.M., 1982. African sleeping sickness: new evidence that mature tsetse flies Glossina morsitans can become potent vectors. Trans. R. Soc. Trop. Med. Hyg. 76, 479–481.
A.R. Msangi et al. / Acta Tropica 70 (1998) 143–155
155
Hargrove, J.W., Langley, P., 1990. Sterilizing tsetse in the field: a successful field trial. Bull. Entomol. Res. 80, 397–403. Harley, J. M. B., 1966. Further studies on age and trypanosome infection rates in Glossinapallidipes Aust., G. palpalisfuscipes Newst. and G. bre6ipalpis Newst. in Uganda. Bull. Entomol. Res. 57, 459–477. Hastie, T.J., Tibshirani, R.J., 1990. Generalised Additive Models. Chapman and Hall, London. Jordan, A.M., 1974. Recent developments in the ecology and methods of control of tsetse flies (Glossina spp.) (Dipt., Glossinidae)—a review. Bull. Entomol. Res. 63, 361 – 399. Jordan, A. M., 1986. Trypanosomiasis Control and African Rural Development. Longman, London. Langley, P.A., Hall, M.J.R., Felton, T., Ceesay, M., 1988. Determining the age of tsetse flies, Glossina spp. Diptera: Glossinidae: an appraisal of the pteridine fluorescence technique. Bull. Entomol. Res. 78, 387–395. Lehane, M.J., Hargrove, J.W., 1988. Field experiments on a new method for determining age in tsetse flies Diptera: Glossinidae. Ecol. Entomol. 13, 319 – 322. Lehane, M.J., Mail, T.S., 1985. Determining the age of adult male and female Glossina morsitans morsitans using a new technique. Ecol. Entomol. 10, 219 – 224. Lloyd, L.L., Johnson, W.B., 1924. The trypanosome infections of tsetse flies in Northern Nigeria and a method of estimation. Bull. Entomol. Res. 14, 265 – 288. Makumyaviri, A.M., Demey, F., Claes, Y., Verhulst, A., Le Ray, D., 1984. Caracterisation de la capacite vectorielle de Glossina morsitans morsitans Diptera: Glossinidae vis a vis de Trypanosoma brucei brucei EATRO 1125 AnTAR 1. Ann. Soc. Belg. Med. Trop. 64, 365 – 372. Maudlin, I., 1991. Transmission of African trypanosomiasis: interactions among tsetse immune system, symbionts and parasites. Adv. Dis. Vector Res. 7, 117 – 148. Mohamed-Ahmed, M.M., Ahmed, A.I., Ishag, A., 1989. Trypanosome infection rate of Glossina morsitans submorsitans in Bahr el Arab, South Darfur Province, Sudan. Trop. Ann. Health Prod. 21, 239–244. Molyneux, D. H., 1977. Vector relationships in the Trypanosomatidae. Adv. Parasitol. 15, 1 – 82. Ryan, L., Molyneux, D.H., Kuzoe, F.A.S., 1980. Differences in rate of wing fray between Glossina species. Tropenmed. Parasitol. 31, 111 – 116. Ryan, L., Kupper, W., Goff, S. L., Molyneux, D. H., Clair, M., 1982. Differences in rates of acquisition of trypanosome infections between Glossina species in the field. Annal. Soc. Belg. Med. Trop. 41, 355–359. Vale, G.A., Hargrove, J.W., Jordan, A.M., Langley, P.A., Mews, A.R., 1976. Survival and behaviour of tsetse flies (Diptera: Glossinidae) released in the field: a comparison between wild flies and and animal-fed and in vitro-fed laboratory reared flies. Bull. Entomol. Res. 66, 731 – 744. Woolhouse, M.E.J., Hargrove, J.W., McNamara, J.J., 1993. Epidemiology of trypanosome infections of the tsetse fly Glossina pallidipes in the Zambezi valley. Parasitology 106, 479 – 485. Woolhouse, M.E.J., Bealby, K., McNamara, J.J., Silutongwe, J., 1994. Trypanosome infections of the tsetse fly Glossina pallidipes in the Luangwa valley. Int. J. Parasitol. 24, 987 – 993.
.