Epidemiology of malaria in the Papua New Guinean highlands

Authors


Corresponding Author Ivo Mueller, PNG Institute of Medical Research, PO Box 60, Goroka, EHP 441, Papua New Guinea. Tel.: +61 3 9345 2555; Fax: +61 3 9347 0852; E-mail: ivomueller@fastmail.fm

Abstract

Objectives  To conduct an in-depth investigation of the epidemiology of malaria in the Papua New Guinea (PNG) highlands and provide a basis for evidence-based planning and monitoring of intensified malaria control activities.

Methods  Between December 2000 and July 2005, 153 household-based, rapid malaria population surveys were conducted in 112 villages throughout the central PNG highlands. The presence of malaria infections was determined by light microscopy and risk factors assessed using a structured questionnaire. The combined dataset from all individually published surveys was reanalysed.

Results  The prevalence of malaria infections in the different surveys ranged from 0.0% to 41.8% (median 4.3%) in non-epidemic surveys and 6.6% to 63.2% (median 21.2%, P < 0.001) during epidemics. Plasmodium falciparum was the predominant infection below 1400 m and during epidemics, Plasmodium vivax at altitudes >1600 m. Outside epidemics, prevalence decreased significantly with altitude, was reduced in people using bed nets [odds ratio (OR) = 0.8, P < 0.001] but increased in those sleeping in garden houses (OR = 1.34, P < 0.001) and travelling to highly endemic lowlands (OR = 1.80, P < 0.001). Below 1400 m, malaria was a significant source of febrile illness. At higher altitudes, malaria was only a significant source of febrile illness during epidemic outbreaks, but asymptomatic malaria infections were common in non-epidemic times.

Conclusions  Malaria is once again endemic throughout the PNG highlands in areas below 1400–1500 m of altitude with a significant risk of seasonal malaria outbreaks in most area between 1400–1650 m. Ongoing control efforts are likely to result in a substantial reduction in malaria transmission and may even result in local elimination of malaria in higher lying areas.

Abstract

Objectifs:  Pour mener une surveillance approfondie de l’épidémiologie du paludisme dans les régions montagneuses de la Papouasie Nouvelle Guinée (PNG) et procurer une base pour la planification basée sur des données et le suivi de l’intensification des activités de lutte antipaludique.

Méthodes:  Entre décembre 2000 et juillet 2005, 153 enquêtes de population rapides sur le paludisme, basées sur les ménages, ont été menées dans 112 villages à travers les régions montagneuses centrales de la PNG. La présence des infections paludiques a été déterminée par la microscopie optique et les facteurs de risque ont étéévalués à l’aide d’un questionnaire structuré. L’ensemble des données combinées de toutes les enquêtes publiées individuellement a été réanalysé.

Résultats:  La prévalence des infections paludiques dans les différentes enquêtes variait de 0,0 à 41,8% (médiane: 4,3%) dans les enquêtes non-épidémiques et de 6,6 à 63,2% (médiane: 21,2%, p <0,001) au cours des épidémies. P. falciparumétait l’infection prédominante en dessous de 1400 m et au cours des épidémies, P. vivaxà des altitudes > 1600m. En dehors des épidémies, la prévalence diminuait de façon significative avec l’altitude, était réduite chez les personnes utilisant des moustiquaires (odds ratio (OR) = 0,8, p <0,001) mais augmentait chez ceux qui dormaient dans des abris de jardin (OR = 1,34, p <0,001) et voyageaient dans les régions de basse altitude à forte endémicité (OR = 1,80, p <0,001). Au dessous de 1400 m, le paludisme était une source importante de maladie fébrile. Sur les altitudes plus élevées, le paludisme était seulement une source importante de maladie fébrile pendant les flambées épidémiques, mais les infections asymptomatiques du paludisme étaient fréquentes dans les périodes non-épidémiques.

Conclusions:  Le paludisme est une fois de plus endémique dans les régions montagneuses de la PNG dans les zones au dessous de 1400-1500 m d’altitude avec un risque important d’épidémies de paludisme saisonnier dans la plupart des régions entre 1400 et 1650 m. Les efforts de lutte en cours sont susceptibles d’entraîner une réduction substantielle de la transmission du paludisme et peuvent même conduire à l’élimination locale du paludisme dans des régions plus élevées.

Abstract

Objetivos:  Realizar un estudio a fondo sobre la epidemiología de la malaria en las tierras altas de Papúa Nueva Guinea (PNG), y proveer así la base para una planeación basada en la evidencia y la monitorización de actividades de control de la malaria.

Métodos:  Se realizaron encuestas rápidas sobre malaria en 153 hogares de 112 poblados a lo largo de la zona central de las tierras altas de PNG entre Diciembre del 2000 y Julio del 2005. Se determinó la presencia de infecciones de malaria mediante microscopía óptica, y se evaluaron los factores de riesgo utilizando un cuestionario estructurado. Se realizó un nuevo análisis de los datos combinados de todos los estudios publicados individualmente.

Resultados:  La prevalencia de infecciones de malaria en los diferentes estudios estaba entre el 0.0% y el 41.8% (mediana 4.3%) en estudios no epidémicos, y entre el 6.6% y el 63.2% (mediana 21.2%, p<0.001) durante las epidemias. P.falciparum era la infección predominante por debajo de 1400 m y durante las epidemias, P.vivax en altitudes >1600m. Fuera de las epidemias, la prevalencia disminuía significativamente con la altitud, era reducida entre personas que utilizaban redes mosquiteras (odds ratio (OR)=0.8, p<0.001) pero aumentaba entre aquellos que dormían en cobertizos (OR=1.34, p<0.001) y viajaban a las áreas bajas, altamente endémicas (OR=1.80, p<0.001). Por debajo de los 1400 m, la malaria era una fuente significativa de fiebre. A mayores alturas la malaria era la única causa significativa de fiebre durante los brotes epidémicos, pero las infecciones asintomáticas eran comunes en épocas no epidémicas.

Conclusiones:  La malaria vuelve a ser endémica en las tierras altas de PNG en áreas por debajo de los 1400-1500 m de altitud, con un riesgo significativo de brotes estacionales de malaria en la mayoría de las áreas entre 1400 y 1650 m. Los actuales esfuerzos de control podrían resultar en una disminución sustancial de la transmisión de malaria, e incluso se podría lograr una eliminación local de la malaria en las áreas a mayor altitud.

Introduction

Although malaria is likely to have been present in low-lying intermountain valleys and in communities with links to the highly malarious lowlands prior to European contact (Radford et al. 1976), the problem of malaria in Papua New Guinea (PNG) highlands was first formally investigated after severe epidemic outbreaks in newly established tea and coffee plantations in Western Highlands Province (WHP) in the late 1940s and early 1950s (Spencer & Spencer 1955; Spencer et al. 1956). Of particular concern was the risk of introducing malaria into newly established coffee plantations through the extensive recruitment of highland labourers to work in lowland plantations (Radford et al. 1976).

The in-depth studies that followed these outbreaks (Peters et al. 1958; Peters & Christian 1960a,b) found seasonal low-level malaria and regular epidemics with significant morbidity and mortality in many of the densely populated highlands areas. Infections with Plasmodium vivax were predominant (68%), Plasmodium falciparum accounted for only 17% and Plasmodium malariae for 15% of all infections. Malaria transmission was highly seasonal, with regular epidemics occurring at the end of the rainy season (i.e. April–July). Overall, it was judged that the elimination of malaria from the PNG highlands was highly feasible (Peters 1960).

Subsequently, an elimination program based on indoor dichlorodiphenyltrichloroethane (DDT) spraying was started that by the early 1970s covered most of the economically important, highly populated areas of the central PNG highlands (Parkinson 1974). Surveys conducted prior to the start of the DDT spraying operation found overall parasite rates of 5–10% in many highlands areas (Ewers & Jeffrey 1971), with strong seasonal fluctuation and predominance of P. vivax (Crane & Pryor 1971). As predicted, the DDT program was very successful, and by the end of the 1960s, parasite rates in the sprayed areas had dropped below 1% and malaria was considered close to eliminated (Black 1969; Ewers & Jeffrey 1971; Parkinson 1974; Crane et al. 1985). In remoter, lower-lying areas (<1200 m) such as in South Simbu, control was less successful and even after several years of control, parasite rates remained at 5–10% (McMahon 1973). When control programs ceased, rates rebounded quickly to levels near or above those seen before control (Crane et al. 1985; Mueller et al. 2005a). The control program never extended into the remote Southern and Western Highlands, and surveys in Enga (Sharp 1982) and Southern Highlands (Hii et al. 1997) in the late 1970s to early 1990s found rates of malaria comparable to those observed in pre-control surveys. In both areas, malaria prevalence strongly decreased with altitude, while the proportion of infection owing to P. vivax increased.

Since the stopping of large-scale vector control in the 1980s, the malaria situation in the PNG highlands has received very little attention, despite increasing reports of epidemic outbreaks. The large-scale epidemic that was observed during the 1997 El Niño event in the highlands of West Papua, Indonesia (Bangs & Subinato 1999), which also caused a massive increase in people seeking malaria treatment in all parts of the PNG highlands, once again alerted PNG health authorities of the malaria problem in the highland areas. In response, a series of rapid malariological surveys was conducted between 2000 and 2005 to assess the extent of malaria transmission throughout the central PNG highlands (Mueller et al. 2003a,b, 2004, 2006, 2007a,b; Maraga et al. 2011). Although primarily aimed at providing a basis for evidence-based planning and improved allocation of program resources at provincial level, together, these surveys allow the first in-depth investigation of the epidemiology of malaria in the PNG highlands for 40 years. In addition, they provide an essential pre-control baseline for the ongoing malaria control efforts undertaken by the PNG national malaria control program, with support from the Global Fund to Fight AIDS, Tuberculosis and Malaria. Here we present a thorough reanalysis of the combined dataset from all surveys that were previously published on a province-by-province basis (Mueller et al. 2003a,b, 2004, 2006, 2007a,b; Maraga et al. 2011), thereby providing a comprehensive summary of all data and additional comparative information on important aspects of PNG highlands malaria.

Materials and Methods

As part of an extensive assessment of the extent of malaria risk in PNG highlands, 153 surveys were conducted in 112 villages distributed through the central highlands range of PNG (Figure 1). Surveys were conducted on a province-by-province basis starting with Western Highlands (Mueller et al. 2003a) in 2000/2001, Eastern Highlands (Mueller et al. 2003b) in 2000–2002, Simbu in 2001/2002 (Mueller et al. 2004), Enga (Mueller et al. 2006) and highlands areas of Morobe (Mueller et al. 2007a) and Madang (Mueller et al. 2007b) in 2003/2004 and Southern Highlands in 2003–2005 (Maraga et al. 2011). Within each province, survey areas were selected based on altitudinal strata and accessibility, with individual villages within each area then selected at random. In all provinces, surveys were conducted both in the wet and dry season albeit not always in the same villages. Village locations were recorded by GPS, and elevations above sea level determined using a barometric altimeter. A detailed description of survey locations and underlying selection criteria are given elsewhere (Mueller et al. 2003a,b, 2004, 2006, 2007a,b; Maraga et al. 2011).

Figure 1.

 Location of all rapid malarialogical surveys conducted in Papua New Guinea highlands 2000–2004. For more detailed maps, see Mueller et al. (2003a,b, 2004, 2006, 2007a,b), Maraga et al. (2011).

Despite the staged conduct of the surveys over 5 years, all surveys were conducted using a uniform method (Mueller et al. 2003a,b, 2004, 2006, 2007a,b; Maraga et al. 2011). To achieve a representative sample of the entire village population, a household-based sampling strategy was used. A number of households with a total population of approximately 200 people were selected. From each selected household, every member, who could be reached during the stay in the village, was included in the survey. If the village had fewer than 200 inhabitants, sampling of every resident was attempted. This approach allowed us to sample approximately 75% of all selected household members. Compared to data from demographic surveillance sites in PNG (Mueller & Hetzel, unpublished data), adolescents and infants are likely to be moderately under-represented in the sample thus obtained.

Demographical data were recorded from all household residents. From each household member that gave consent to participate, a thick and thin blood film was prepared, the spleen palpated in lying position and axillary temperature taken. Haemoglobin levels were measured using the HemoCue system (HemoCue AB, Ängelholm, Sweden). Symptomatic individuals were tested with rapid diagnostic test (Diamed, Cressier, Switzerland, or ICT Diagnostics, Durban, South Africa) and those found positive for malaria were treated according to standard treatment guidelines. A short questionnaire on current symptoms, past malaria episodes and treatment, recent travel and other behavioural patterns (sleeping in garden houses, hunting and fishing) was administered to each participant or their guardian.

Giemsa-stained blood films were examined by light microscopy for 100 thick film fields under oil immersion before being declared negative. The parasite species in positive films were identified and densities recorded as the number of parasites per 200 white blood cells (WBC). Densities were converted to the number of parasites per μl of blood assuming 8000 WBC per μl. The slides were read independently by two experienced microscopists.

Statistical analyses were carried out using STATA 8.0 (Stata Corp., College Station, TX, USA) statistical software. Chi-square tests and logistic regression analyses were used for binary and categorical variables. Individual differences in haemoglobin levels were investigated using analyses of variance (anova), while non-parametric tests were used to analyse differences in mean prevalence of infection and rates of enlarged spleens between surveys conducted at different altitude and season.

Results

Between December 2000 and July 2005, 153 surveys with a total of 22 485 participants were conducted in 112 villages distributed through the central highlands range of PNG (Figure 1). Thirty-six villages were surveyed twice, two villages three times. The majority of surveys were conducted in either the mid to late wet season [March–May: 55/153 (36%)] or early to mid dry season [June–August: 70/153 (46%)]. Nineteen surveys (12%) were in villages <1000 m, 22 (14%) at 1000–1199 m, 20 (13%) at 1200–1399 m, 47 (31%) at 1400–1599 and 45 (29%) at altitudes ≥1600 m. The number of participants ranged from 51 to 258 (median = 151) for non-epidemic (n = 135) and 38 to 249 (median = 121) for epidemic surveys (n = 19). Epidemic outbreaks were only observed at altitudes of 1250–1960 m [interquartile range (IQR) = 1520–1640] and between March and July.

Of all participants, 11 497 (51.1%) were female, 6442 (28.7%) children <10 years of age, 4256 (18.9%) adolescents (10–19 years) and 11 197 (49.8%) adults. The age of 560 (2.6%) participants could not be determined. Although there were significant differences in age distributions between non-epidemic and epidemic surveys (Table 1) or among surveys at different altitudes (data not shown), these differences were small (<3%) and thus unlikely to affect cross-survey comparisons. A more detailed description of survey characteristics is given elsewhere (Mueller et al. 2003a,b, 2004, 2006, 2007a,b; Maraga et al. 2011).

Table 1.   Basic demographic characteristics of study participants
 Non-epidemic (N = 20 082) (%)Epidemic (N = 2403) (%) P-value
Gender
 Female10 139 (51.6)1358 (50.4)0.26
Age groups (years)
 <2985 (4.9)90 (3.7)0.001
 2–31351 (6.7)155 (6.5)
 4–61869 (9.3)216 (9.0)
 7–91575 (7.8)201 (8.4)
 10–193819 (19.0)437 (18.2)
 20–396266 (31.2)799 (33.3)
 40+3664 (18.2)468 (19.5)
 NA553 (2.8)37 (1.5)

Altitude and malaria endemicity

The prevalence of malaria infections in the different surveys ranged from 0.0% to 41.8% (median 4.3%, IQR [0.1, 5.3]) in non-epidemic surveys and 6.6% to 63.2% (median 21.2%, IQR [10.3, 35.2], P < 0.001) in surveys conducted during epidemic outbreaks. Prevalence rates were highly negatively correlated with altitude rates in non-epidemic surveys (Figure 2, ρ = −0.55, n = 134, P < 0.001) but not in endemic surveys (ρ = −0.16, n = 18, P = 0.52). The decrease was less pronounced for P. vivax than P. falciparum (Table 2) and the proportion of infections caused by P. vivax significantly increased with altitude (ρ = 0.26, P = 0.002). Consequently, while in areas below 1400 m, P. falciparum was more prevalent, and P. vivax dominated above 1600 m. While there was only limited seasonal difference in surveys conducted at altitudes <1200 m (P = 0.15), marked seasonality was observed at higher altitudes with the higher prevalence rates observed during the second half of the rainy season and the early dry season (i.e. March to July, median PR: 7.6% IQR [3.7, 21.9] compared to the main dry [August–October, 3.7%, IQR [2.2, 5.0] and the early wet season (November–February, 2.3%, IQR [1.3, 4.7]), P = 0.008]. These seasonal differences were exclusively found among P. falciparum infections (March–July: 4.3%, August–October: 1.5%, November–February: 0.8%, P = 0.009, P. vivax, P = 0.22) and owing to the high rates of P. falciparum infection during epidemic surveys (65.0%vs. 40%, P = 0.1). No significant seasonal differences were found if only non-endemic surveys >1200 m were considered (P = 0.34).

Figure 2.

 Association of altitude of survey location with prevalence of malarial infections and rate of enlarged spleens in children 2–9 years.

Table 2.   Median prevalence and species composition of malarial infections in surveys conducted at different altitudes
Altitude (m) N Prevalence of infectionSpecies composition†Enlarged spleen
PfPvPmPoAllPfPvPmSR 2–9‡IQR
  1. N, number of non-epidemic and epidemic surveys conducted at each altitude; IQR, Interquartile range. *P < 0.001, **P < 0.01.

  2. †As a proportion of all infections.

  3. ‡SR 2–9: Proportion of children 2–9 years with enlarged spleen.

Non-epidemic
 <10001913.17.51.9022.860.9%27.3%10.0%34.414.2, 66.2
 1000–1199228.13.10.3013.467.3%25.5%3.4%8.00, 25.5
 1200–1399194.13.4009.252.8%36.5%0.0%8.90, 28.5
 1400–1599371.51.5002.550.0%50.0%0.0%00, 10.5
 >1600370.61.0002.533.3%66.6%0.0%00, 4.2
 Spearman’s rho −0.61*−0.42*−0.45*−0.30*−0.55*−0.24**0.26**−0.36*−0.56* 
Epidemic
 1200–139919.716.30022.236.0%59.6%4.4%25.613.0, 40.0
 1400–159978.85.60011.367.5%30.0%0.0%11.70, 32.6
 >160089.89.41.4016.155.5%42.2%3.5%9.80.9, 21.5
 Spearman’s rho −0.05−0.270.250.02−0.16−0.01−0.070.22−0.37 

Enlarged spleens were commonly observed in children 2–9 years living in villages below 1000 m (34.4, IQR [14.2, 66.2]) indicating mesoendemic transmission. Spleen rates (SR) decreased significantly with increasing altitude (Table 2, ρ = −0.56, P < 0.001), but even at altitudes of 1000–1399 m, 18 of 41 (43.9%) surveys observed spleen rates >10%. Above 1400 m, enlarged spleens were rarely observed except during epidemic outbreaks (non-epidemic: 0, IQR [0, 4.5], epidemic 13.3, IQR [2.8, 26.7], P < 0.001). The prevalence of enlarged spleens was highly correlated with prevalence of malarial infections in non-epidemic (Figure 2, ρ = 0.73, n = 134, P < 0.001) but not in epidemic surveys (ρ = 0.35, n = 18, P = 0.16). Although the prevalence of infection was comparable in epidemic surveys (P = 0.65) and those conducted at altitudes below 1000 m, spleen rates were significantly lower in epidemic surveys (13.3%vs. 34.4%, P = 0.02).

Based on prevalence and spleen rates, the following altitudinal limits of malaria endemicity can be defined: >1000 m: moderately endemic (mesoendemic), 1000–1399: low endemic (hypoendemic) with risk of epidemic outbreaks, 1400–1599 m: epidemic malaria, and above 1600 m: no local malaria transmission although certain areas may be at risk of epidemics and imported cases may be present. Detailed maps and estimated proportion of villages in each stratum are given on a province-by-province basis in (Mueller et al. 2003a,b, 2004, 2006, 2007a,b; Maraga et al. 2011).

Individual risk factors for Plasmodium infection

The risk of malaria infection was very strongly age dependent. Overall (Figure 3), malarial infections were most common in children 2.0–3.9 (20.5%) and 4.0–6.9 (20.1%) years of age and lowest in adults (20–39: 8.0%, 40+: 6.1%, P < 0.001). With increasing altitude, a moderate shift in peak prevalence to older age groups was observed. In areas below 1400 m, peak prevalence was observed in children 2.0–3.9, shifting to children 4.0–6.9 at higher altitudes (Figure 3, test for age differences for individual age groups, χ2-values 36.0–229.2, df = 6, P < 0.001). Significant differences in age-specific prevalence rates were observed for both P. falciparum (Figure 3, all P-values < 0.001, except >1600 m: P = 0.064) and P. vivax (all P-values < 0.001).

Figure 3.

 Age-specific prevalence of malarial infections at different altitudes.

Whereas the relative age distribution in P. vivax infection was comparable during epidemic and non-epidemic surveys (χ2 = 9.9, df = 6, P = 0.13) in areas above 1200 m, significantly more P. falciparum infections were observed in adults during epidemics (χ2 = 13.5, df = 6, P = 0.036) with the median age of P. falciparum-positive participants rising from 12 years (IQR [5.0, 25]) in non-epidemic surveys to 16 years (IQR [6.4, 30], P = 0.002) in epidemic surveys.

At the time of the surveys, bed net use was relatively low overall with only 11.7% of participants having slept under a net the previous night. Use was most common in villages <1000 m (31.1%) and least common in those >1600 m (4.9%, P < 0.001). When taking into account differences in prevalence with altitude and age, bed net use was independently associated with a significant reduction in risk of malarial infections [odds ratio (OR) = 0.80, P = 0.002] in non-epidemic surveys. The protective effect of bed nets was exclusively against infection with P. falciparum (OR = 0.73, P < 0.001) with no effect at all against infection with P. vivax (P = 0.97). During epidemic surveys, there was also a tendency for insecticide treated nets (ITN) use to be associated with protection against P. falciparum infections (OR = 0.49, P = 0.055) but not against P. vivax infections (P = 0.25).

In non-epidemic surveys, people reporting sleeping regularly in a garden house had a higher risk of both P. falciparum (Table 3, OR = 1.37, P < 0.001) and P. vivax infections (OR = 1.21, P = 0.032), whereas a recent history of travel to the lowlands was associated only with significantly higher risk of P. falciparum (OR = 2.38, P < 0.001) but not P. vivax infections (P = 0.12).

Table 3.   Individual, multivariate risk factors for malaria infections in non-epidemic and epidemics surveys
 Any infections Plasmodium falciparum Plasmodium vivax
OR [CI95] P OR [CI95] P OR [CI95] P
Non-epidemic surveys
 Altitude (m)
  <1000       
  1000–11990.60 [0.52, 0.69] 0.69 [0.59, 0.81] 0.56 [0.45, 0.70] 
  1200–13990.38 [0.32, 0.44] 0.38 [0.32, 0.46] 0.59 [0.46, 0.75] 
  1400–15990.15 [0.12, 0.17] 0.11 [0.08, 0.13] 0.35 [0.28, 0.45] 
  ≥ 16000.09 [0.08, 0.11]<0.0010.06 [0.04, 0.08]<0.0010.25 [0.19, 0.32]<0.001
 Age (years)
  <2      
  2.0–3.92.05 [1.57, 2.69] 1.85 [1.32, 2.59] 1.83 [1.27, 2.63] 
  4.0–6.91.99 [1.57, 2.57] 1.76 [1.27, 2.44] 1.63 [1.57, 2.31] 
  7.0–9.91.62 [1.24, 2.12] 1.63 [1.17, 2.28] 1.04 [0.71, 1.53] 
  10.0–19.91.18 [0.93, 1.52] 1.26 [0.93, 1.71] 0.78 [0.55, 1.11] 
  20.0–39.90.64 [0.50, 0.82] 0.72 [0.53, 0.98] 0.42 [0.29, 0.60] 
  ≥400.53 [0.41, 0.70]<0.0010.60 [0.43, 0.84]<0.0010.36 [0.24, 0.54]<0.001
 Garden house1.34 [1.20, 1.50]<0.0011.37 [1.20, 1.57]<0.0011.21 [1.02, 1.45]0.032
 Lowland travel1.80 [1.34, 2.37]<0.0012.38 [1.74, 3.24]<0.001  
 ITN use0.80 [0.70, 0.93]0.0020.73 [0.61, 0.87]<0.001  
Epidemic surveys
 Altitude (m)
  <1600
  ≥16000.75 [0.58, 0.95]0.020  0.44 [0.30, 0.65]<0.001
 Age (years)
  <2      
  2.0–3.91.93 [1.03, 3.62] 2.42 [0.94, 6.20] 1.25 [0.59, 2.64] 
  4.0–6.92.50 [1.38, 4.55] 3.64 [1.49, 8.90] 1.49 [0.73, 3.03] 
  7.0–9.91.96 [1.07, 3.60] 2.72 [1.09, 6.74] 1.22 [0.59, 2.50] 
  10.0–19.91.19 [0.67, 2.11] 1.93 [0.80, 4.64] 0.77 [0.39, 1.52] 
  20.0–39.90.98 [0.56, 1.71] 1.77 [0.75, 4.18] 0.54 [0.28, 1.05] 
  ≥400.65 [0.36, 1.17]<0.0011.17 [0.48, 2.86]<0.0010.33 [0.16, 0.71]<0.001
 Garden house    0.53 [0.36, 0.77]0.001
 ITN use  0.49 [0.22, 1.06]0.055  

Malaria-associated morbidity

Although malaria infections were associated with substantial morbidity throughout the highlands, the burden of malaria-attributable fevers (MAF) was strongly related to transmission levels (Table 4). Whereas malaria was a significant source of febrile illness in moderate to low endemic areas (i.e. altitude <1400 m), at higher altitude malaria was only a significant source of febrile illness during epidemic outbreaks. Outside epidemics, 29.5% of participants living below 1400 m with a history of fever during the last 72 h had concurrent malarial infections, compared to only 7.7% (P < 0.001) among those reporting a febrile illness in surveys conducted above 1400 m. Similarly, if only cases with manifest fever (i.e. axillary temperature >37.5 °C) were considered, 40.0% and 15.6% (P < 0.001), respectively, were positive for malaria upon blood slide examination. In children <10 years, malaria-attributable cases of febrile illness increased to 42.8% (history) and 54.7% (temp >37.5 °C) for low to moderately endemic areas (<1400) and 13.1% and 19.6% at altitudes above 1400 m (P < 0.001). In the high transmission season (i.e. March to August), malaria was a more common cause of illness in both endemic (35.0%vs. 12.1%, P < 0.001) and non-endemic areas (8.7%vs. 4.3%, P = 0.010). During epidemic outbreaks, malaria was, however, the major source of morbidity. Not only do significantly more people report a febrile illness (1400 m: 24.4%vs. 12.2%, P < 0.001), the fevers are also much more likely to have concurrent malarial infections (38.8%vs. 7.7%, P < 0.001).

Table 4.   Morbidity associated with malarial infections at different altitudes: observed fevers, reported febrile illness and haemoglobin levels
Altitude (m) N Axillary temperature >37.5 °CReported fever lasts 3 daysAnaemia
Febrile*MAF†Sympt‡Febrile*MAF†Sympt‡Mean g/dlHb < 8
g/dl
ΔHB (adj)§
g/dlCI95
  1. *Febrile: Prevalence observed fevers (axillary temperature >37.5°C) and reported febrile illness (in last 3 days) among participants in surveys.

  2. †Malaria-attributable fevers (MAF): Proportion of participants with observed fevers and reported febrile illness that have concurrent Plasmodium spp. infections (positive on thick film).

  3. ‡Sympt: Proportion of participants with Plasmodium spp. infections that have observed or reported febrile symptoms.

  4. §Reduction in mean haemoglobin levels associated with concurrent Plasmodium spp. infection.

Non-epidemic
 <100030185.6%40.5%8.9%15.0%35.4%20.9%11.17.4%0.580.42, 0.73
 1000–119933403.8%34.7%8.0%13.6%26.4%21.9%12.04.8%0.950.76, 1.14
 1200–139930722.2%49.2%9.3%11.9%25.7%26.3%12.23.4%1.291.07, 1.24
 1400–159952721.8%18.1%7.0%12.8%9.6%24.9%12.91.4%1.060.84, 1.29
 >160055510.9%10.6%3.1%11.7%5.8%22.3%13.50.5%0.900.66, 1.14
   inline image = 215.6 inline image = 31.9 inline image = 7.2 inline image = 22.6 inline image = 211.7 inline image = 4.7 F 4,17950 = 702.3 inline image = 389.5LR4,17939 = 8.5 
   P < 0.001 P < 0.001 P = 0.13 P < 0.001 P < 0.001 P = 0.32 P < 0.001 P < 0.001 P < 0.001 
Epidemic
 1200–13991715.9%60.0%15.8%25.2%46.5%52.6%11.91.8%1.340.74, 1.94
 1400–159912604.5%50.1%10.0%26.2%37.0%42.1%12.34.8%1.421.13, 1.71
 >16009723.1%63.3%9.6%21.6%41.6%44.1%12.72.7%1.481.15, 1.81
   inline image = 4.5 inline image = 1.3 inline image = 2.1 inline image = 6.6 inline image = 2.1 inline image = 1.5 F 2,2158 = 11.5LR2 = 8.4 F 2,2118 = 0.8 
   P = 0.11 P = 0.52 P = 0.50 P = 0.04 P = 0.36 P = 0.46 P < 0.001 P = 0.02 P = 0.32 

While malaria was an important source of illness, the majority of Plasmodium spp. infections were asymptomatic with only 26.9% people with malarial infection reporting to be suffering from febrile illness (Table 4, ‘sympt’). In non-epidemic surveys, there was no difference in rate of reported febrile illness (in last 72 h) associated with a malaria infection among people living at different altitudes (P = 0.326). The proportion was significantly higher during epidemic (43.6%) than non-epidemic surveys (22.6%, P < 0001) and for P. falciparum (29.1%) than P. vivax (25.2%, P = 0.033). An additional 16.9% reported having suffered from ‘sik malaria’ (local vernacular for febrile illness without cough) during the last 2 weeks. In total 37.7% and 65.2% (P < 0.001) of people with concurrent Plasmodium infections thus reported ongoing or recent febrile illness during non-epidemic and epidemic surveys, respectively.

Irrespective of endemicity, malaria infections were associated with an increased risk of anaemia (Table 4). In non-epidemic surveys, concurrent malarial infections resulted in a 0.58–1.29 g/dl reduction (all P-values < 0.001) in age- and sex-specific mean haemoglobin (Hb) levels. The effect of concurrent infections on Hb levels was significantly smaller at low altitudes (P < 0.001). However, as population mean Hb levels increased significantly with increasing altitude (and thus decreasing malaria endemicity, Table 4), moderate-to-severe anaemia (i.e. Hb < 8 g/dl) was only commonly found in moderate and low endemic area with 7.4% and 4.2% of participants, respectively.

Discussion

The surveys conducted from 2000 to 2005 show that malaria is once again endemic throughout the PNG highlands in areas below 1400–1500 m of altitude and that a significant risk of seasonal malaria outbreaks exists in most areas between 1400–1650 m. As such, the malaria situation at the start of the 20th century is remarkably similar to that of the 1950s and 1960s prior to the last elimination campaign (Black 1954; Spencer & Spencer 1955; Spencer et al. 1956; Peters et al. 1958; Peters & Christian 1960a,b; Ewers & Jeffrey 1971; Sharp 1982) with the same areas once again endemic for malaria and current transmission and risk of epidemic equally peaking in the late dry to early wet season (i.e. March to June).

The one major change is the increased prevalence of P. falciparum both in endemic areas as well as a cause of potentially severe epidemic outbreaks (Mueller et al. 2005b). This is most likely due to the increased travel between the highlands and highly endemic areas in Madang and Morobe provinces after the construction of a sealed highway (Radford et al. 1976). Recent travel to the lowlands, often to market highland grown vegetable in coastal towns, was found to be a significant risk factor for infection with P. falciparum but not P. vivax. Such travel constitutes a source of introduction of coastal parasites into highlands area that can cause epidemic outbreaks of P. falciparum malaria in areas that are otherwise climatically suited only for endemic transmission of P. vivax (Mueller et al. 2002).

Altitude and seasonal rainfall patterns are the major determinants of malaria transmission in the PNG highlands. At altitudes below 1400 m, temperatures allow endemic malaria transmission with P. falciparum as the predominant parasite. Given its shorter developmental cycles in the mosquito and thus lower minimum temperature (Gilles & Warrell 2002), endemic P. vivax transmission can occur locally until at least 1600 m, where P. falciparum transmission happens mainly during epidemic outbreaks (Mueller et al. 2005b). As a consequence, the relative proportion of P. vivax infection increased with altitude. Interestingly, the altitudinal limits for stable malaria transmission in PNG are lower than those reported from the East African (Ghebreyesus et al. 2000; Shanks et al. 2005; Gahutu et al. 2011) but comparable to those in the Malagasy highlands (Romi et al. 2002). The reasons for this are unclear.

As already shown by Peters et al. in the late 1950s (Peters & Christian 1960b), transmission in highlands areas is closely linked to seasonal rainfall patterns. Following the onset of the rainy season in November, it takes several months for mosquito numbers to build up sufficiently for malaria transmission to take off. As the rains wane and breeding sites start drying up in April–May, mosquito numbers also start dropping and malaria transmission falls until it practically stops at the height of the dry season. The seasonal pattern of malaria transmission thus showed a 2-month lag compared to the observed rainfall patterns (Peters & Christian 1960a). Although recent entomological studies in the PNG highlands are lacking, the seasonal difference in prevalence as well as the timing of epidemic outbreaks indicate the same still holds true today.

Despite its wide geographical distribution, malaria is not a major source of febrile illness in PNG highland areas above 1400 m. Except during epidemic outbreaks that can carry a very high burden of morbidity (Mueller et al. 2005b), fewer than one in 10 participants with reported and one in six with measured axillary temperature >37.5 had concurrent Plasmodium infections. Given the syndromic nature of fever treatment at PNG health facilities, most of these febrile, parasite-negative patients would nevertheless be treated with antimalarials. The new 2009 PNG malaria treatment guidelines (PNG Department of Health 2009) for the first time include parasitological diagnosis of all fever cases with rapid diagnostic tests (RDTs) or light microscopy and advocate treatment only for parasitologically confirmed cases. If properly implemented, these new guidelines could reduce the number of malaria treatments dispensed in areas >1400 m by up to 90%, thus limiting the amount of overtreatment and reducing the risk of drug resistance.

As transmission levels are low at altitudes >1400 m, it was surprising that even at these higher altitudes the majority of Plasmodium infections were not associated with febrile symptoms. Equally intriguing, although the age of peak prevalence shifting into older age groups with increasing altitude, at all altitudes children had a higher risk of being infected than adults (Figure 3). Both observations challenge the notion that PNG highlanders are ‘non-immune’ to malaria and together with recent observations of predominantly asymptomatic communities in remote island (Harris et al. 2010) and East African Highlands (Baliraine et al. 2009) indicate that a significant level of clinical immunity to malaria may be achieved even at low levels of transmission. Several factors could contribute to the apparent level of antimalarial immunity. Firstly, studies of Indonesian transmigrants found that adults acquire significant clinical immunity after as few as three P. falciparum infections, whereas children require many more infections to acquire a comparable level of immunity (Baird et al. 1991, 2003). Secondly, as a result of the high levels of overtreatment with antimalarials, many highlanders are likely to have residual antimalarial drug levels in their blood, which may suppress infecting Plasmodium parasites at low levels allowing the development of strong, protective immune responses (Roestenberg et al. 2009). Last but not least, parasites in low transmission settings tend to be genetically less diverse (Anthony et al. 2005) making it easier to acquire immunity to locally circulating parasites. Irrespective of the reasons, the presence of large number of asymptomatic infections will limit the effectiveness of a malaria control strategy that is primarily based on case-management.

Future perspective for malaria in the PNG Highlands

In 2004, PNG received a first grant from the Global Fund to Fight AIDS, Tuberculosis and Malaria that allowed a country-wide distribution of 1.35 million long-lasting insecticide-treated bed nets (Global Fund to Fight AIDS, Tuberculosis and Malaria, 2011). With continued Global Fund support for 2009–2013, the national malaria control program is conducting additional long-last insecticide treated nets (LLIN) distributions and will roll out of artemisinin-based combination therapy and RDT throughout PNG. The experience from the DDT control program in the 1960s and 70s (Black 1969; Parkinson 1974) as well as the significant protection associated with the use of (mostly untreated) bed nets in the current study indicates that the ongoing control efforts are likely to result in a substantial reduction in malaria transmission and may even lead to local elimination of malaria in specific areas. Given the increasing mobility of many highlands populations, there thus remains a significant risk of re-introduction of malaria with potential for epidemic outbreaks (Mueller et al. 2005b) should malaria control measures be neglected or abandoned. Continued monitoring of the malaria transmission, identification of residual transmission hotspots as well as surveillance for potential outbreaks therefore need to be important components of any malaria control and elimination program in the PNG Highlands.

Acknowledgements

First and foremost we would like to thank all study participants, the provincial health authorities and mission health services of the PNG highlands provinces without whose support this work would have been impossible. We also thank Rex Ivivi, John Taime, Gimana Poigeno, Moses Ousari, Jonah Iga, Bianca Pluess and Sonja Schoepflin for the conduct of the field studies and Leo Makita, Steve Bjorge, Luo Dapeng and Ian Riley for their contributions to design and interpretation of the study results. This work was supported by grants from the WHO Western Pacific Regional Office, Rotary Against Malaria and the PNG National Department of Health.

Ancillary