The impact of gastrointestinal nematodes on wild reindeer: experimental and cross-sectional studies

Authors


Dr Audun Stien, Centre for Ecology & Hydrology, Hill of Brathens, Banchory AB31 4BW, UK. Tel: 01330 826336; Fax: 01330 823303; E-mail: a.stien@ceh.ac.uk

Summary

  • 1It is well known that gastrointestinal nematodes can have a significant impact on the growth of farmed ruminants. The clinical signs of these parasites are often subtle, with production losses mainly due to reductions in host appetite and gut function. However, little is known about the impact of this widespread group of parasites on wild ruminants. We use experiments and cross-sectional data to investigate the effects of gastrointestinal nematodes on a wild host, the Svalbard reindeer.
  • 2Individually marked reindeer were treated for parasites. Their body mass, back fat depth and fecundity were compared with the control group one year later. The effect of treatment on worm burdens was investigated in a subsample of animals that were culled 3 and 6 months after treatment. Also, the relationship between the intensity of infection and body mass, back fat depth and host pregnancy rates was investigated in cross-sectional data from culled reindeer.
  • 3The anthelmintic treatment caused an increase in the body mass, back fat depth and fecundity of the reindeer. Treatment depressed the abundance of adult parasites of Ostertagia gruehneri for at least 6 months, but had no significant effect on the abundance of adults of the other dominant parasite species, Marshallagi marshalli, or the abundance of larval stages of either species.
  • 4In the cross-sectional study, the probability of a reindeer being pregnant in late winter was negatively related to the abundance of adult O. gruehneri when controlling for host body mass. However, no clear evidence were found for an effect of parasitism on host condition in the cross-sectional study.
  • 5Our experimental results show for the first time in a natural ruminant host population that gastrointestinal nematodes can have a significant effect on host condition and fecundity. However, effects of infection on host condition was not detectable in the cross-sectional study. Also, we show that the experimental effects on host condition and fecundity is most likely to be due to a negative effect of O. gruehneri, while the experimental design did not allow detection of potential negative effects of M. marshalli.

Introduction

The understanding of the interacting effects of density-dependent and climatic factors on the population dynamics of ruminants has improved considerably in recent years (Gaillard et al. 2000; Coulson et al. 2001). The factors usually identified as the causes of variation in ruminant demographic rates are food availability, direct effects of climate, predation and viral and bacterial diseases (Gaillard et al. 2000). However, gastrointestinal nematodes are abundant in wild ruminants (Hoberg, Kocan & Richard 2001), and theory suggests that they may have significant effects on plant–herbivore interactions (Grenfell 1988, 1992). Gastrointestinal nematode infections of domestic ruminants are well known to be responsible for large production losses worldwide (Coyne & Smith 1994). The main effect of these parasites on domestic ruminant productivity is through subclinical infections, which are difficult to detect without controlled experiments, but cause reduced appetite, food assimilation and growth (Soulsby 1982; Xiao & Gibbs 1992; Arneberg, Folstad & Karter 1996; Fox 1997; Forbes et al. 2000). As in domestic systems, gastrointestinal nematodes would be expected to have a negative impact on host growth and body condition and associated negative effects on host fecundity and survival. In support of this, Gulland (1992) showed in an experimental study of feral Soay sheep that animals treated with anthelmintics to reduce their gastrointestinal nematode infections had higher survival than untreated controls. No other studies have shown a clear effect of this widespread group of parasites on the body condition or demographic rates of wild ruminant hosts (Gulland 1995). We suspect that this may be because preliminary cross-sectional studies have suggested that these parasites are of little importance and therefore research groups have not persued the matter experimentally. Here we report on the first long-term study of a gastrointestinal nematode–wild ruminant interaction. We investigate the impact of gastrointestinal nematodes on the body condition and fecundity of Svalbard reindeer (Rangifer tarandus platyrhynchus Vrolik) using both longitudinal experimental data on individual reindeer, and cross-sectional data from culled animals.

Reindeer are the only wild ruminant on the high Arctic archipelago of Svalbard (76–81°N, 9–33°E) and are therefore the only host for a species depauperate helminth community dominated by directly transmitted strongyle nematodes of the abomasum (Bye & Halvorsen 1983; Halvorsen & Bye 1986). Recent studies using molecular techniques have also shown that just two species of nematodes, Ostertagia gruehneri Skrjabin and Marshallagia marshalli Ransom (Dallas, Irvine & Halvorsen 2000a, 2001), contribute more than 99% of the infrapopulations of gastrointestinal nematodes (Irvine et al. 2000). The simplicity of this parasite community makes this system ideal for the study of host–parasite interactions. The abundance of these abomasal nematodes is higher in Svalbard reindeer than in reindeer populations on the Norwegian mainland (Bye 1987), and previous studies of culled animals have suggested a negative relationship between measures of host fat stores and the intensity of infection (Halvorsen & Bye 1986, 1999).

We analyse data from a 6-year study of the impact of nematode infections on reindeer pregnancy rates, body mass and back fat depth. First, we analyse the performance of individually marked reindeer where gastrointestinal nematodes were removed experimentally from a subsample of animals each year. We include an analysis of the effects of the anthelmintic treatment on worm burdens using animals culled 3 and 6 months after treatment. Second, we analyse cross-sectional data on reindeer performance and parasite intensities of infection obtained from culled animals, to evaluate the sensitivity of this more traditional approach in detecting impact of nematode infections on wildlife hosts.

Materials and methods

experimental study

The study was undertaken in Nordenskiöldland, Spitsbergen (77°50′−78°20′N, 15°00′−17°30′E). Female reindeer were caught and marked in the Colesdalen–Semmeldalen–Reindalen valley system in late April/early May each year from 1995 with in total 642 observations on body mass for 337 individual reindeer and 325 observations on pregnancy or presence of a calf for 207 individuals one year after treatment. The age of reindeer caught for the first time as calves or yearlings was known with certainty. Otherwise, animals caught for the first time were 2 years or older. At capture, body mass was measured and, since 1998, the back fat depth has been measured on a subset of animals (299 observations on 228 individuals) using ultrasound scanning. Svalbard reindeer give birth in the first 2 weeks of June (Tyler 1987). Pregnancy status in April/May was determined using the progesterone (P4) concentration in blood samples and ultrasound diagnosis (Ropstad et al. 1999). In addition, the calving success was recorded by presence of a calf between 25 June and 25 August.

Three anthelmintic treatments were used across the 5 years. In 1995 we used only a slow-release albendazole capsule (Proftril, Pfizer Ltd, Sandwich, Kent, UK) delivered orally to the rumen with an anticipated period of efficacy of about 3 months. This was discontinued in subsequent years because the size of the capsule was thought to be inappropriate for Svalbard reindeer. In 1996, we used only a single dose injection of moxidectin: Cydectin (Wyeth, Madison, NJ, USA), 0·2 mg per kg live mass. In both these years every other animal caught was treated, with the controls receiving nothing. Since 1997 we have treated one-third of animals with both an ivermectin capsule (Ivomec Maximizer, Merial Animal Health Ltd, Harlow, Essex, UK) and the injectable moxidectin, one-third with moxidectin only, with the remaining one-third being controls. The administration of a slow-release capsule was re-introduced because the ivermectin capsule is smaller and therefore easier to administer. Moxidectin was given with the ivermectin capsule as a back-up in the case the animal managed to regurgitate the capsule when released. The group treated with only moxidectin was maintained to control for any treatment-induced differences between the experimental groups. The moxidectin treatment was expected to give a relatively short period of protection from parasite infections (6 weeks), while the slow-release capsule was expected to keep the reindeer parasite free for approximately 3 months. In the analysis we only distinguish between treated and untreated animals since we found no significant differences in body mass, back fat or fecundity 1 year after treatment between animals that received injectable moxidectin and the ivermectin capsule in any of the analyses (P > 0·6) although there tended to be greater positive effects in the capsule treated group. The treatment effect was investigated as an impact on reindeer performance and fitness rather than a direct test of the anthelmintics on nematode abundance. However, we measured re-infection in adult reindeer treated with moxidectin alone and culled three (n = 8) and six (n = 7) months post-treatment (late July and October, respectively) in 1999. At the same time untreated adult reindeer were sampled for comparison (n = 7 at 3 months, n = 7 at 6 months).

cross-sectional study

Two-year-old or older female reindeer were culled every year in the autumn, October (1994–99) and in the winter, either February (1997–98), March (1996–98) or April–early May (1995–2000). Svalbard reindeer are in peak body condition in October and lose body mass and fat from then onwards until the following June, when they start building up their fat reserves again (Reimers & Ringberg 1983). A total of 69 animals were culled in the Colesdalen–Semmeldalen–Reindalen valley system where we also followed individually marked animals, while 111 were culled in Sassendalen approximately 50 km East of Colesdalen. Movement of reindeer between these valleys is thought to be uncommon (Øritsland & Arendal 1986; Tyler & Øritsland 1989), as recently conferred by molecular techniques showing genetic separation between them (Côtéet al. 2002). The total body mass (kg) and back fat depth (mm) was measured on all culled reindeer. For reindeer culled in October, analyses of the ovaries were used to determine whether the animal had ovulated or not (Langvatn 1992). In the winter samples, pregnancy was determined by the presence of a foetus. Accurate ages of culled animals were derived from annuli in the cementum of extracted (I1) incisors (Reimers & Nordby 1968).

In the culled animals, nematodes were extracted from abomasum and counted following the methodology described in Halvorsen et al. (1999). The taxon of adult male nematodes was determined from their morphology (Drózdz 1995), and minor and major morphs of the same species were grouped according to recent molecular findings (Dallas et al. 2000a, 2001). The intensity of infection of reindeer by adult nematodes of the two species was estimated using the proportion of males of each species and assuming the same proportion of the respective species in the female fraction. Larval stages extracted from abomasum mucosa and abomasum lumen were distinguished but were not speciated due to a lack of species specific morphological traits prior to the recent development of molecular tools (Dallas et al. 2000b).

statistical analyses

The repeated observations of body mass and back fat depth in April–May of the individually marked reindeer in the experimental study were analysed using linear and non-linear mixed effects models (Davidian & Giltinan 1995). To correct for between individual variation in these variables, reindeer identity was fitted as a random effect. This approach reduced the degrees of freedom in the analysis substantially since animals that were measured only once in the course of the study contributed only to the estimation of the random effect. Age was coded as a two-level categorical variable (yearlings, and adult (age > 2 years) reindeer) and, together with year of sampling and status with respect to anthelmintic treatment the previous year, was fitted as a fixed effect. In any year a reindeer could belong to one of three groups with respect to experimental treatment: (1) treated, (2) control or (3) not caught the previous year and therefore not part of the experiment. To focus the analysis on the animals that took part in the experiment, a separate parameter was fitted for each year for the animals in group (3) in all models. To stabilize the variance, the observations on body mass were log-transformed before analysis and the variance assumed to be constant. The models for back fat depth were fitted assuming a log link function and a power relationship between the variance and the mean back fat depth. The statistical significance of fixed effects were tested using Wald F-tests. The models were fitted in Splus (using the lme and nlme functions, Venables & Ripley 1999).

For the culled animals we did not find a good model that described well both the within year variation and the between year variation in body mass and back fat depth. We therefore analysed the samples from October separately from the winter samples (February–May). The age was known for all culled animals. The increase in body mass with reindeer age could therefore be modelled using the von Bartallanfy growth function:

ŷ = b1[1 − emath image](eqn 1)

where ŷ is the expected value of the response variable, b1 is the expected asymptotic value of the response variable, b2 is the rate of increase towards b1, and b3 is the age where ŷ is zero. The model was fitted assuming constant variance. Predictor variables other than age were fitted as effects on the asymptotic body mass with age (b1). For the animals culled in autumn the categorical predictor variables used, were whether the reindeer was still lactating or not (Lactation), year and site (Colesdalen and Sassendalen) of sampling. For the animals culled in winter the categorical predictor variables used were year, month and site of sampling. Lactation was not included in the model for winter body mass since only 4 reindeer of the 82 sampled were found lactating in the winter. In addition, the intensity of adult O. gruehneri, M. marshalli, nematode larvae in abomasum lumen, nematode larvae in abomasum mucosa and total infection were fitted as continuous predictor variables. The back fat depth of culled animals was analysed assuming a log link function and a power relationship between the variance and the mean back fat depth. The same predictor variables were used as in the models for body mass, except that the effect of reindeer age was modelled both as a continuous variable (Age) and as a class variable indicating 2-year-old-reindeer (Age2).

For the experimental study, we attempted a generalized mixed model approach also for the analysis of pregnancy. However, it was difficult to get these models to converge and the estimates of the individual level random effect were zero or very close to zero. This suggested that very limited information was present for estimating this random effect, probably due to the discrete nature of the binary data and relatively few observations available from the same animal across years (Breslow & Clayton 1993). We therefore ignored the possible effect of between individual variation in fecundity and developed a model that maximized the use of the available data on reindeer fecundity by incorporating information on both pregnancy in April–May and the presence of a calf in the summer. Information on both pregnancy and the presence of a calf in the summer was obtained on 35% of the individuals observed in a given year, while only one of these variables was observed for the rest of the females. The data were analysed using a multinomial model with a logit link (McCullagh & Nelder 1989) to estimate the two parameters, p1 = the probability of a female reindeer being pregnant, and p2 = the probability of a reindeer having a calf given it was pregnant. Here we report the results from the analysis of p1, since this was the parameter that was affected by anthelmintic treatment. The full analysis is given in Albon et al. 2002). Likelihood ratio tests and profile likelihood confidence limits were used to identify significant effects of the explanatory variables (treatment and year) on p1 (McCullagh & Nelder 1989). The ovulation rate in reindeer culled in October and pregnancy rate in reindeer culled in the winter were analysed using logistic regression.

The abundance of nematode infections in animals culled in July and October was analysed using models with a negative binomial error structure and a log link function (Wilson & Grenfell 1997). The models were fitted using maximum likelihood as described in Irvine et al. (2000). In the model for adult O. gruehneri the negative binomial variance parameter k was found to vary between treatment groups, and in the model for M. marshalli, k was modelled as an exponential function of the mean (see Irvine et al. 2000). A constant k was found to be sufficient when modelling the abundance of nematode larvae in abomasum lumen and mucosa.

Results

experimental study

Body mass

The average body mass of marked female reindeer increased with increasing age from yearlings to 2 years old or greater and differed between years of sampling (Table 1). After controlling for these fixed effects, there was a positive effect of treatment on average body mass (estimated mean treatment effect over years = 1·9 kg, SE = 0·42, Fig. 1a). This effect of treatment was not found to differ significantly between age classes or years (Table 1).

Table 1.  Wald tests for the fixed effects in mixed effects models for reindeer body mass and subcutaneous back fat, with reindeer identity fitted as a random effect. The fixed effects included in the best fitting models are highlighted in bold. Age was coded as a two level categorical variable (yearlings and adult reindeer; age ≥ 2 years). Measurements of subcutaneous back fat were carried out on a subset of animals in year 1998–2000. The back fat was only measured on one yearling in 1998, allowing only one parameter to be fitted for the Age * Year interaction. The estimated standard deviation of the random effect for between individual variation was for the best body mass model σ = 0·060, and for the back fat model σ = 0·25
Fixed effectBody massSubcutaneous back fat
Fd.f.PFd.f.P
Age545·21, 294<0·0001 9·901, 640·003
Year72·424, 294<0·000112·212, 64<0·0001
Treatment 8·171, 2940·005 9·381, 640·003
+Age * Year 1·874, 2900·11 0·201, 620·66
+Treatment * Year 0·244, 2900·91 2·232, 620·12
+Treatment * Age 0·041, 2930·85 0·391, 630·54
Figure 1.

The (a) average body mass and (b) estimated pregnancy rate in April–May of adult Svalbard reindeer not treated (open bars) and treated with anthelmintics (shaded bars) the previous April–May. Error bars give 95% confidence limits of the means.

Back fat

The subcutaneous back fat thickness of marked female reindeer increased with increasing age from yearlings to adult reindeer (age ≥ 2 years) and differed between years of sampling (Table 1). After controlling for these fixed effects there was a positive effect of treatment on average back fat thickness (estimated mean treatment effect over years = 3·3 mm, SE = 1·8, Fig. 2). This effect of treatment was not found to differ significantly between age classes or years (Table 1).

Figure 2.

The average subcutaneous back fat depth in April–May of adult Svalbard reindeer not treated (open bars) and treated with anthelmintics (shaded bars) the previous April–May. Error bars give 95% confidence limits of the means.

Pregnancy

There was a significant positive effect of anthelmintic treatment on reindeer pregnancy rates (estimated log odds ratio = 0·63, 95% CI = [0·04, 1·24], Fig. 1b). In addition, the pregnancy rate in marked adult female reindeer varied significantly between years (χ2 = 22·16, d.f. = 4, P = 0·0002). The average treatment effect across years increased the probability of pregnancy by 11% (SE = 3%, Fig. 1b). After controlling for the effect of body mass, the estimated effect of treatment on pregnancy was reduced and not significantly different from zero (estimated log odds ratio = 0·17, 95% CI = [−0·79, 1·12]). Similarly, by controlling for the effect of back fat thickness the effect of treatment on pregnancy was removed (estimated log odds ratio = 0·04, 95% CI = [−0·95, 1·01]), suggesting that most of the positive effect of treatment on the pregnancy status of female reindeer was through a positive effect on their body condition with an associated increased likelihood of pregnancy.

Worm burdens

In late July, 3 months after treatment, all culled animals treated with moxidectin alone had become re-infected with nematodes. The abundance of adult O. gruehneri did not increase over the next 3 months (from July to October) in either the moxidectin treated or control group (χ2 = 0·10, d.f. = 1, P = 0·75, Fig. 3a). In the pooled sample (July and October), the moxidectin-treated animals had significantly lower burdens (63%) of adult O. gruehneri than the untreated controls (χ2 = 4·11, d.f. = 1, P < 0·05). In contrast, there was no effect of moxidectin treatment on the total abundance of M. marshalli2 = 0·25, d.f. = 1, P = 0·62, Fig. 3b), or larvae either in the abomasum lumen (χ2 = 0·12, d.f. = 1, P = 0·73, Fig. 3c) or its mucosa (χ2 = 1·31, d.f. = 1, P = 0·25, Fig. 3d). Marshallagia marshalli abundances were close to zero in both the treated and control groups in July (3 months post-treatment) and rose from July to October (6 months post-treatment). The larval abundances also all increased from July to October (P < 0·003, Fig. 3b–d).

Figure 3.

The estimated abundance, in thousand worms per host, in July and October of (a) adult Ostertagia gruehneri, (b) adult Marshallagia marshalli, (c) nematode larvae in abomasum lumen, and (d) nematode larvae in abomasum mucosa, in Svalbard reindeer not treated (open bars) and treated with anthelmintics (shaded bars) the previous April–May. Error bars give 95% confidence limits of the means.

cross-sectional study

Body mass

The body mass of adult female reindeer culled in October increased asymptotically with age, was lower in lactating than non-lactating females (estimate = −5·36, SE = 1·18), and varied with year and site of sampling with a year * site interaction (Table 2). In the February–May samples, body mass depended on reindeer age, the year and month of sampling. Also there was a significant three-way interaction between year, month and site of sampling (Table 2). After controlling for these factors there was no significant relationship between female body mass and the intensity of parasites in the reindeer (Table 2) even though body mass showed a negative trend with all measures of parasite burdens in October and with the intensity of M. marshalli, larvae in abomasum lumen and total infection in the samples from February to May.

Table 2. anova table for models of the total body mass of adult female reindeer culled in October or February–May. Residual sums of squares (RSS), residual degrees of freedom (d.f.) and the P values from F-tests are given. The significant terms are highlighted in bold. The effect of age was modelled using a von Bartallanfy growth function, other terms were fitted as effects on the asymptotic with age mass. Year * Site denote the interaction term between year and site of sampling, and Year * Month * Site the three-way interaction between year, month and site of sampling which was significant also after fitting the two-way interaction terms between the variables
TermOctoberTermFebruary–May
RSSd.f.PRSSd.f.P
Intercept3441·1697 Intercept5142·4881 
Age2698·2295<0·0001Age4211·6179<0·0001
Lactation2411·85940·0002Year (Y)3374·73740·0002
Year1881·63890·0002Month (M)2692·8372<0·0001
Site1794·20880·03Site (S)2654·38710·30
Year * Site1591·08840·04Year * Site2336·56670·07
+O. gruehneri1579·36830·43Year * Month2271·86650·12
+M. marshalli1546·42830·13Site * Month2147·65630·09
+larvae in lumen1580·87830·47Year * Month * Site1831·24620·004
+larva in mucosa1577·88830·41+O. gruehneri1817·55610·50
+total infection1553·52830·16+M. marshalli1810·15610·40
    +larvae in lumen1806·76610·37
    +larva in mucosa 1830·87610·91
    +total infection1827·85610·74

Back fat

The subcutaneous back fat depth of female reindeer culled in October was lower in 2-year-old animals than 3-year-olds (estimate = −0·18, SE = 0·06). However, after this age, the back fat depth decreased with increasing age (estimate = −0·014, SE = 0·006). Back fat depth was lower in lactating vs. non-lactating females (estimate = −0·18, SE = 0·03) and showed significant between year variation. In reindeer sampled in February–May, there was a significant reduction in back fat depth as the winter proceeded and with increasing reindeer age (estimate = −0·057, SE = 0·019). Reindeer culled in Sassendalen also had lower back fat depth than those culled in Colesdalen (estimate = −0·28, SE = 0·09). After controlling for these factors there was no significant relationship between measures of back fat depth and worm burden (Table 3).

Table 3. F-tests for models for the subcutaneous back fat of adult female reindeer culled in October and February–May. In bold, the statistics for the significant terms. The effect of age was modelled using a both a class variable indicating 2-year-old animals (Age2) and a continuous variable (Age)
TermOctoberTermFebruary–May
Fd.f.PFd.f.P
Age 5·881, 900·02Month32·572, 78<0·0001
Age210·671, 900·002Age 9·311, 780·003
Lactation29·951, 90<0·0001Site 9·231, 780·003
Year 8·185, 90<0·0001+Age2 2·231, 770·14
Site 2·221, 890·14+Year 0·955, 730·45
+O. gruehneri 1·101, 890·30+O. gruehneri 0·201, 770·66
+M. marshalli 2·461, 890·12+M. marshalli 0·251, 770·62
+Larvae in lumen 0·011, 890·96+Larvae in lumen 0·021, 770·89
+Larva in mucosa 0·661, 890·42+Larva in mucosa 0·881, 770·35
+Total infection 0·301, 890·59+Total infection 0·781, 770·38

Pregnancy

The ovulation rate in reindeer culled in October increased with reindeer body mass (slope = 0·253, SE = 0·077, P = 0·001). After controlling for this effect no other variable was found to be significant. Also, there was a strong positive relationship between the probability of pregnancy among animals culled in February–May and body mass (slope = 0·246, SE = 0·063, P < 0·001). However, after correcting for this effect there was also a strong negative effect of the adult O. gruehneri infection (slope = −0·00027, SE = 0·00010, P = 0·004, Fig. 4). Pregnancy did not show a statistically significant relationship with other measures of parasite intensities, year and site of sampling after controlling for body mass (P > 0·4).

Figure 4.

The distribution of pregnant (closed circles) and non-pregnant (open circles) adult female Svalbard reindeer culled in February–May 1995–2000 in relation to their body mass and intensity of Ostertagia gruehneri infection. The logistic regression estimates of a probability of pregnancy of 0·1, 0·3, 0·5, 0·7 and 0·9 are given using contour lines.

Discussion

Our experimental study shows that gastrointestinal nematodes have a negative effect on the body condition, body mass and fecundity of Svalbard reindeer. The negative effect on reindeer body mass is consistent with the results from domestic ruminants, where most of the effect is attributable to reduced appetite (Fox 1993, 1997; Arneberg et al. 1996; Forbes et al. 2000). The effect of the anthelmintic treatment on the average body mass (1·9 kg) and back fat depth (3·3 mm) 1 year after treatment was relatively small but consistently positive every year. Nonetheless, the average treatment effect on the pregnancy rate of adult female reindeer was substantial (11%). This effect could be explained by the differences in host back fat depth and body mass, suggesting that the effect of the nematodes on pregnancy works through an effect on host condition.

We suspect that the effects of the parasites on host back fat and body mass were underestimated in the experimental study because of the long time-span from treatment until we measured the response. Reindeer treated with moxidectin alone had all become infected with nematodes 3 months post-treatment, and these infections had a substantial time to cause a negative effect on treated hosts by the time of measurement, a further 9 months later. Also, the effects of treatment were measured at the seasonal low for back fat and body mass. Any difference at peak body mass and back fat in the autumn has therefore been modulated by the loss of body condition experienced by all animals over the winter (see also Mitchell, McCowan & Nicholson 1976).

Albon et al. (2002) evaluate the potential importance of gastrointestinal nematodes in the population dynamics of Svalbard reindeer. They show that the effect of treatment on reindeer fecundity is positively related to the autumn abundance of O. gruehneri, and suggest that O. gruehneri may be a significant factor in the regulation of Svalbard reindeer populations. Here we show that 3 months post-treatment, treated animals had similar levels of infection with M. marshalli compared to controls, but had significantly lower levels of adult O. gruhneri. This effect of treatment on the abundance of adult O. gruehneri was persistent 6 months after treatment. This finding supports the hypothesis that the effects of the anthelmintic treatment on reindeer condition, body mass and pregnancy is mainly due to a reduction in the levels of O. gruehneri. It is also consistent with the finding of a negative relationship between pregnancy rates and adult O. gruehneri intensities in the reindeer culled in the winter. In contrast, M. marshalli has been found to have no negative impact on infected sheep (Al-Khalidi et al. 1989), and our results suggests that this may also be true for reindeer. However, owing to the seasonal pattern in the population dynamics of M. marshalli, our experimental design did not allow detection of potential negative effects of M. marshalli on the host. Svalbard reindeer have large infections of larvae in the abomasum mucosa in April–May (range = 2000–29 000, Halvorsen et al. 1999) and a recent genetic study of these larvae has shown that they are predominantly O. gruehneri (Irvine 2001). The reduction in the number of adults of O. gruehneri due to treatment is therefore likely to be due to both the direct removal of adults and the removal of O. gruhneri larvae that otherwise would have developed into adults.

Since no effect on the abundance of adult M. marshalli or the abundance of the larval nematode stages was detectable 3 months after treatment, turnover rates of these groups of parasites are probably relatively high. The average life expectancy of M. marshalli must be significantly less than 3 months, and similarly, the combined effect of the survival and developmental rate of the larval stages must cause the average period the parasite stays as larvae in abomasum mucosa and lumen during spring to be less than 3 months. A short life expectancy of adult M. marshalli is also consistent with the observation that its population dynamics show strong seasonal fluctuations that would only be possible with a relatively short adult life expectancy (El-Azazy 1995; Irvine et al. 2000). In contrast, the population dynamics of O. gruehneri show a less pronounced seasonal fluctuation, suggesting a longer life expectancy (Irvine et al. 2000).

In the cross-sectional data, the pregnancy rate of the reindeer in winter showed a negative relationship with the abundance of adult O. gruehneri although we found no strong relationship between the back fat depth, body mass and ovulation rate of the host and our measures of the intensity of nematode infections. Nonetheless, the experimental results show that the parasites have a negative effect on the back fat depth and body mass of the host. The reason for our inability to detect this effect in the cross-sectional data is likely to be both ecological and statistical. The different ecological processes involved in generating the observed worm burdens, host back fat depths and body masses may obscure simple relationships and tend to cause both positive and negative covariation between these variables (Gulland 1995). Since gastrointestinal nematodes are transmitted via food, their numbers may be especially likely to show positive covariation with host growth and condition since animals that have a high feeding rate may ingest more infective larvae and also have a higher growth rate (Halvorsen & Bye 1986, 1999; Hutchings et al. 2000). The effect of parasites with low pathogenicity is also likely to be the result of slow accumulative processes, where the parasite burden at the time of sampling may be a poor representation of the infection history of the animal. These factors may obscure a negative relationship between host fitness measures and parasite burdens when cross-sectional data are analysed. Our cross-sectional data set is also affected by two statistical problems common to this type of data. First, because of ethical considerations, a relatively small number of hosts could be culled at any one time and site combination. The necessity of sampling different sites introduces variation in the data set that needs to be accounted for by the statistical model and together with the small sample sizes, reduces the likelihood of detecting relatively small effects of the parasites. Second, since the animals were sampled only when culled, causes of between individual variation in body mass and back fat depth independent of parasite burdens add to the error variance. In contrast, it is possible to control for this variation in the individual-based longitudinal study. Clearly, a manipulative longitudinal individual-based approach is preferable and more likely to succeed in giving reasonable estimates of the effect of parasites in the wild (McCallum 2000).

The widespread occurrence of gastrointestinal nematodes in ruminants suggest that they may be of importance in the population dynamics of many species (Arneberg et al. 1996). Our study shows it may be difficult to detect the negative effect of this group of parasites in cross-sectional studies even when these parasites are of importance to host condition and demographic rates. Experimental manipulations of parasite loads will therefore have to be performed to get a proper evaluation of the significance of this group of parasites in natural systems (Tompkins & Begon 1999). Hopefully, this approach will be adopted by other long-term studies of wild ruminants.

Acknowledgements

We are grateful to the Governor of Svalbard for permission to work on Spitsbergen, and the support of his environmental staff, in particular, Jon Ove Scheie. At the outset essential logistical support and equipment hire were provided by the Norwegian Polar Institute. Our spring catching trips involved additional help from Erling Meisingset, Inge Engeland, Irma Oskam, Jan Licke, Leif Egil Loe, Glenn Roar Berge, Yvonne Halle, Steeve Côté, Chantal Beaudoin and Elke Lindner, and the subsequent summer follow-up included Steve Wilkinson, Erling Meisingset, Veibjørn Veiberg, Leif Egil Loe, Johan Andersen, Chris Mcfarlane and Jules Jones. Jørn Dybdahl, Fred Skanche Hansen, Stein Lier-Hansen and members of Longyearbyen Hunting and Fishing Association have helped with the culling. We very much appreciate all their contributions. Andy Forbes and Barry McPherson at Merial New Zealand Ltd. kindly supplied the ivermectin boluses and Larry Parker at Fort Dodge Animal Health Ltd. the moxidectin. The work was funded both by the Research Council of Norway (TERRØK programme 1994–96, and Arktisk Lys programme 1996–99) and the Natural Environment Research Council, UK (1997–2000: GR3/10811).

Ancillary