Insecticide-treated nets and malaria prevalence, Papua New Guinea, 2008–2014

Abstract Objective To investigate changes in malaria prevalence in Papua New Guinea after the distribution of long-lasting Insecticide-treated nets, starting in 2004, and the introduction of artemisinin-based combination therapy in 2011. Methods Two malaria surveys were conducted in 2010–2011 and 2013–2014. They included 77 and 92 randomly selected villages, respectively. In each village, all members of 30 randomly selected households gave blood samples and were assessed for malaria infection by light microscopy. In addition, data were obtained from a malaria survey performed in 2008–2009. Results The prevalence of malaria below 1600 m in altitude decreased from 11.1% (95% confidence interval, CI: 8.5–14.3) in 2008–2009 to 5.1% (95% CI 3.6–7.4) in 2010–2011 and 0.9% (95% CI 0.6–1.5) in 2013–2014. Prevalence decreased with altitude. Plasmodium falciparum was more common than P. vivax overall, but not everywhere, and initially the prevalence of P. vivax infection decreased more slowly than P. falciparum infection. Malaria infections were clustered in households. In contrast to findings in 2008–2009, no significant association between net use and prevalence was found in the later two surveys. The prevalence of both fever and splenomegaly also decreased but their association with malaria infection became stronger. Conclusion Large-scale insecticide-treated net distribution was associated with an unprecedented decline in malaria prevalence throughout Papua New Guinea, including epidemic-prone highland areas. The decline was accompanied by broader health benefits, such as decreased morbidity. Better clinical management of nonmalarial fever and research into residual malaria transmission are required.


Malaria in Papua New Guinea
Manuel W Hetzel et al.
census -the most up-to-date. 11 Not all provinces or selected villages could be included because of problems with access and security. The pre-2012 province structure was adopted to ensure comparability over time: Hela Province was considered part of Southern Highlands Province and Jiwaka Province, part of Western Highlands Province. For each village, the survey team leader selected a random sample of 30 households using a list compiled by village leaders. All members of sampled households were eligible for inclusion. The sample size, which took into account financial and operational constraints, was adequate for detecting a 25% reduction in parasitaemia from 2008-2009 to 2010-2011 at the regional level at a 95% level of significance with 80% power. The first survey in 2008-2009, whose results are presented for comparison, included villages from only districts covered by the long-lasting insecticide-treated net campaign, but the method of selecting households and their members was identical to that in subsequent surveys. 7 Data were collected using an adapted Malaria Indicator Survey questionnaire. 12 Household heads provided details of each household member's demographic characteristics and coverage by malaria interventions. A capillary blood sample was collected by fingerstick from each available, consenting household member aged over 5 months. Trained study nurses prepared one thick and one thin blood film for light microscopy. The haemoglobin concentration was measured using a portable He-moCue Hb 201 + photometric analyser (HemoCue AB, Ängelholm, Sweden). Symptomatic household members were offered a malaria rapid diagnostic test and treatment or referral to the nearest health-care facility, where appropriate. Axillary temperature was measured using an electronic thermometer and children aged between 2 and 9 years had their spleen palpated. Each patient's blood sample was accompanied by information on recent travel. The locations of the survey villages were determined using a hand-held Garmin eTrex Global Positioning System device (Garmin Ltd., Olathe, United States of America).
Malaria was diagnosed by light microscopy at the Papua New Guinean Institute of Medical Research following established procedures. 7,13 Each slide was examined independently by two microscopists, each viewing a minimum of 200 thick film fields. Slides with discordant results were examined by a third microscopist, who was certified at World Health Organization (WHO) level 1 or 2. A slide was considered positive for malaria if judged positive by at least two microscopists. For the 2010-2011 survey, additional assessments of unclear species identifications were performed at the Australian Army Malaria Institute in Australia by WHOcertified level-1 malaria microscopists. The number of parasites per 200 white blood cells was determined. The study was approved by the Papua New Guinea Medical Research Advisory Committee (MRAC no. 07.30 and no. 10.12).

Data analysis
Measures of the prevalence of malaria infection and morbidity were age-standardized using the standard population for Asia given by the International Network for the Demographic Evaluation of Populations and Their Health (INDEPTH). 14 Results are presented separately for villages below 1600 m in altitude and include comparisons with data from the 2008-2009 survey. For villages at 1600 m or higher, we compared data from the 2010-2011 and 2013-2014 surveys only as the 2008-2009 survey included few highland villages. To account for stratified sampling, national estimates were weighted, as described elsewhere. 5 Splenomegaly was defined as a palpable spleen (i.e. Hackett grade 1 to 5) and anaemia was defined according to WHO recommendations, which include age-specific cut-offs and altitude corrections. 15 Living in a high-quality house served as a proxy for having both good sanitation and a relatively high socioeconomic status, as defined elsewhere. 10 Binary variables were compared using χ 2 tests and logistic regression, and non-normally distributed variables were compared using the non-parametric Mann-Whitney U test. Data analyses were conducted using Stata/IC v. 14.0 (StataCorp LP., College Station, USA) and the survey design was taken into account by using Statas set of commands for survey data analysis (svy).

Research
Malaria in Papua New Guinea Manuel W Hetzel et al.

Predictors of infection
Regression analysis findings are presented in Table 5. Univariable logistic regression found that, in 2010-2011, malaria infection was significantly less likely above 1600 m (odds ratio, OR: 0.15; 95% CI: 0.06-0.38); in 2013-2014, the corresponding OR was 0.04 (95% CI: 0.00-0.33). The prevalence of infection below 1600 m was significantly    16 In the 2008-2009 survey, a significant association was found between net use and a lower risk of malaria infection (adjusted odds ratio, aOR: 0.64; 95% CI: 0.54-0.76). 7 However, no corresponding association was found in the 2010-2011 survey (aOR: 1.09; 95% CI: 0.82-1.45), in an analysis that adjusted for altitude, age and housing quality, or in the 2013-2014 survey (aOR: 1.08; 95% CI: 0.64-1.82), in an analysis that adjusted for altitude and age (Table 5). In 2010-2011, people living in high-quality houses were significantly less likely to be infected (aOR: 0.25; 95% CI: 0.08-0.79). Malaria cases were clustered in households. Univariable analysis found that, in 2010-2011, the odds of infection were over 26 times higher for individuals living with an infected person than for those who were not (OR: 25

Morbidity
In all surveys, individuals infected with malaria were significantly more likely than those without to report a recent fever episode, to show symptoms of acute fever (i.e. an axillary temperature over 37.5 °C), to be anaemic or, in those aged 2 to 9 years, to have splenomegaly (P < 0.01 for all). Although the proportion of the population with a recent   Notes: The black line shows the proportion for each whole survey population. Differences between subsequent years in the proportion of the overall population were significant (P < 0.01).

Research
Malaria in Papua New Guinea Manuel W Hetzel et al.
history of fever decreased over time (Fig. 6) 37.4% of all infected individuals reported a recent fever episode (Fig. 6) and 3.7% had an acute fever (Fig. 7). The prevalence of splenomegaly in participants aged 2 to 9 years also decreased over time (Fig. 8) and again the association with infection tended to become stronger: the OR adjusted for age

Discussion
Within 5 years, the prevalence of malaria in Papua New Guinea decreased from 11.1% to 0.9% and during 2013-2014 no parasites were detected by light microscopy in most provinces. This is a greater reduction than the 26% observed in Africa between 2000 and 2016. 17 Moreover, the prevalence in 2014 was lower than that in other countries in the Asia-Pacific region, including the neighbouring Papua province of Indonesia. [18][19][20][21] An initial shift towards proportionally more P. vivax than P. falciparum infections appeared to be transient and was followed by a clear reduction in both species, as observed elsewhere. 22,23 These trends are in line with previously documented declines in malaria following the introduction of long-lasting insecticidetreated nets. 8,9,24 Provinces in which no malaria parasites were found should not be considered malaria-free because, as parasite density decreases, an increasing proportion of infections becomes submicroscopic, 13,18,25 particularly if transmission decreases faster than the loss of Notes: The black line shows the proportion for each whole survey population. Differences between subsequent years in the proportion of the overall population were significant (P < 0.01). Notes: The black line shows the proportion for each whole survey population. Differences between subsequent years in the proportion of the overall population were significant (P < 0.01). Notes: The black line shows the proportion for each whole survey population. Differences between subsequent years in the proportion of the overall population were significant (P < 0.01). Anaemia was defined according to WHO recommendations, which included age-specific cut-offs and altitude corrections. 15 Malaria in Papua New Guinea Manuel W Hetzel et al.
immunity. In three provinces with zero prevalence, rapid diagnostic tests found that people with fever who had not left the province had a current or recent infection. Consequently, maintaining a high level of intervention coverage is crucial for avoiding resurgence. The notion that climatic change might have increased malaria in the highlands could not be substantiated. 26 In locations above 1600 m, malaria prevalence was lower in 2010-2011 and 2013-2014 than between 2000 and 2005. 2 The protective effect of insecticide-treated nets in both the highlands and lowlands, from where infections are often imported, 27 may have outweighed the impact of changing weather patterns or increased people movement. Unlike in previous years, 2,3 P. falciparum was the dominant species in the highlands. The prevalence of fever and splenomegaly declined with that of parasite infection. However, the association between infection and symptoms became stronger over time, perhaps because the proportion of microscopically detectable infections that were symptomatic increased as transmission and immunity declined. The decrease in splenomegaly was most marked, which may reflect a reduction in chronic malaria infection. 7 On the other hand, anaemia remained common, indicating that the cause is multifactorial. 28 Anaemia may not, therefore, be useful for monitoring rapid changes in malaria prevalence. 29 As severe anaemia, in particular, affects children's health and development, its causes and appropriate mitigating measures should be investigated. 28,30 Between 2004 and 2012, the distribution of insecticide-treated nets to households was the only large-scale malaria intervention in Papua New Guinea. 10 The baseline survey demonstrated a strong negative association between net coverage and malaria prevalence. 7 In the absence of other factors, such as major economic developments or a prolonged drought, 9 it is plausible that the drop in prevalence between the 2008-2009 and 2010-2011 surveys resulted from increased provision of nets and measures promoting their use. The lack of an association between net use and malaria prevalence in the last two surveys may have been due to factors such as outdoor biting, which sustained disease transmission, and the mass effect of net use on all community members. 31 With our survey design, it was not possible to quantify the relative contributions of net use and artemisinin-based combination therapy to the reduction in prevalence. Combination therapy was introduced in November 2011 and, by late 2012, was available at approximately half of health-care facilities. 32 Nevertheless, although the treatment's gametocidal effect can reduce transmission from patients, its prophylactic effect is limited. Moreover, in 2014, only 45% of patients with confirmed or suspected malaria who attended health-care facilities were treated with artemisinin-based combination therapy, 33 which corresponds to a population coverage of 19% at best. The community benefits of combination therapy can be maximized by prompt diagnosis and treatment. 34 With decreasing malaria prevalence, clinicians across Papua New Guinea should be encouraged to administer antimalarials only to people with a positive test result, which has proven to be a safe approach, 35 and to thoroughly investigate the causes of nonmalarial fevers. Better guidance on differential diagnosis and on fever management is warranted. 36 We found that individuals cohabiting with another infected person were more likely to carry parasites, possibly due to similar exposure patterns. There were fewer infections in high-quality houses occupied by better-off households, possibly because of economic factors or the building's structure or location -most high-quality houses were in urban areas. Earlier studies in Papua New Guinea found conflicting evidence of the impact of housing, namely raised structures, on mosquito exposure. 37,38 As indoor exposure to malaria vectors has been reduced by nets, people's outdoor behaviour may be an increasingly important determinant of exposure as many vectors tend to bite outdoors. 9,39 The investigation of residual malaria transmission is crucial for eliminating the disease and should take into account human and mosquito behaviour patterns, including the distribution of different Anopheles populations, their biting preferences and their susceptibility to interventions. 8,9,37 In our study, we used age-standardization to account for differences in the age-composition of participants between surveys and between participants who gave blood samples and the general population. However, as the 2008-2009 survey included only districts where nets were distributed, the national prevalence of parasite infection may have been underestimated. Data from sentinel sites showed that the prevalence after net distribution was 4.8% compared with 15.7% before. 8 In addition, the estimated prevalence in the 2010-2011 and 2013-2014 surveys may have been too low because, due to security concerns, they excluded West New Britain Province, where the prevalence is traditionally high. 40 In conclusion, increased use of long-lasting insecticide-treated nets in Papua New Guinea was associated with a rapid and significant decline in malaria prevalence -the lowest prevalence ever recorded was in 2013-2014. The decline also occurred in the epidemic-prone highlands. Light microscopy showed that P. falciparum remained more common than P. vivax. Notes: The black line shows the proportion for each whole survey population. Differences between subsequent years in the proportion of the overall population were significant (P < 0.01). Severe anaemia was defined according to WHO recommendations, which included age-specific cut-offs and altitude corrections. 15 Research Malaria