Background
Malaria has been a major cause of morbidity and mortality in tropical areas particularly in sub-Saharan Africa (SSA), and was estimated to cause 300 – 350 million cases and 0.5 - 1 million deaths annually between 2000 and 2007 [
1]. However, recent reports have shown that the burden of malaria has declined in several countries with remarkable differences between different regions of Africa [
1,
2]. In southern Africa, a significant decline of malaria burden has been reported in the South African Republic [
3‐
6], Zambia [
7,
8], Swaziland and Southern part of Mozambique [
9,
10]. A similar decline has also been reported in the Horn of Africa (Ethiopia and Eritrea) [
11‐
15] and in isolated islands of Zanzibar [
16,
17], Sao Tome and Principe [
18] and Bioko Island in Equatorial Guinea [
19,
20]. Reports from other parts of Africa show inconsistent results, although a declining burden of malaria has been reported in some countries, including Kenya [
21‐
24], Tanzania [
25‐
28], Rwanda [
15,
29], the Gambia [
30,
31] and Gabon [
32]. These changes have been attributed to scaling-up of control interventions such as deployment of artemisinin-based combination therapy (ACT), insecticide-treated bed nets (ITNs), indoor residual spraying (IRS) and larval control [
2,
33]; yet, there are indications that the declining malaria burden in some of the countries actually started before implementation of such major intervention programmes [
2]. A recent study conducted in Kenya revealed that drivers of changing malaria epidemiology in some areas might not be entirely attributed to claimed interventions [
34].
In Tanzania, malaria has until the mid-2000s been responsible for over 18 million cases, 100,000 deaths/year and accounting for over 43% of outpatient cases, 30% of all hospital admissions and 40% of deaths occurring at health facilities [
35]. Despite few reports which have indicated that malaria has declined in most parts of the country [
36], most of the data obtained from health facilities through the health management information system (HMIS) have revealed that the burden of malaria in Tanzania has remained largely unchanged with high and similar proportions of hospital attendances, admissions and deaths between 2003 and 2010, as documented in the World Health Organization (WHO) report of 2011 [
1]. However, the most recent Tanzania HIV and malaria indicator surveys (THMIS) conducted in 2011/2012 have shown that malaria parasite prevalence has declined by >50% since the latest surveys of 2007 (from 18% to 9%), with the highest parasite prevalence of 33% in the North-western region of Geita near Lake Victoria and the lowest of <1% in several regions of North-western, Central, Southern highlands and Zanzibar Island (MoH, unpublished data). Furthermore, studies conducted in Muheza district of north-eastern Tanzania have showed a concurrent decline of malaria incidence among hospitalized children, and interestingly, also of community acquired bacteraemia, suggesting that the decline in malaria burden is also affecting trends of bacterial infections [
26].
The data used in the WHO report [
1] were derived from hospital records through HMIS which in most cases are over-estimated by including majority of patients diagnosed based on symptoms only. Even in areas with the capacity for malaria diagnosis by microscopy, previous studies showed that malaria was commonly over-diagnosed especially in areas of low transmission, mainly attributed to limited skills of microscopists for preparation and examination of blood smears [
37]. Thus, more detailed studies are required to determine the current malaria burden in Tanzania particularly after introduction of malaria rapid diagnostic tests (RDTs) in 2012.
Apart from the mentioned national surveys, the observed decline in the burden of malaria in Tanzania is based on few detailed studies [
25‐
28] and thus, there is limited information from other malaria endemic areas in the country. Furthermore, the reasons for the decline of malaria burden in these particular areas are not clear, although some of the interventions deployed in recent years are believed to be the main cause of the current changes in malaria epidemiology [
1,
25,
27]. This indicates that more detailed studies are required to document the trend of malaria in areas with variable endemicity and to provide reliable information on the possible association between interventions in place and resulting effects on malaria burden, for supporting formulation of new policies for malaria control. Finally, the current decline in the burden of malaria, when confirmed, will demand for a significant changes in malaria control to identify and treat the remaining cases and eventually attempt to block malaria transmission as well as improve management of non-malaria fevers.
A long-term presence in the two rural villages of Muheza district enabled the present study to assess the long-term trends of malaria burden over a period of 20 years (1992 – 2012). In these villages, which were until recently holo/hyper-endemic to malaria, different inventions (including ITNs, passive case detection and treatment of malaria, and different anti-malarial drug trials) have been deployed, reflecting changing local demands and national policies. Their potential contribution to the observed trends of decreasing malaria is discussed, while the remarkable changes of parasite prevalence are given in more details, including differences between the two nearby villages.
Methods
Study site and interventions between 1992 and 2012
This study utilized assembled data collected during repeated cross-section surveys (CSS) conducted between September 1992 and June 2012 in the village of Magoda and between 1997 and June 2012 in the neighbouring village of Mpapayu both in Muheza district in Tanga region, North-eastern Tanzania. The villages were involved in malaria and filariasis studies as described elsewhere [
38‐
40]. Currently, the two villages have a population of 2,934 individuals in 678 households and under-fives account for 13.2% of the entire population. The data generated by previous studies, therefore, provide additional value in assessing the historical evidence to demonstrate the impact of different interventions and policies over the period.
After an initial screening of parasite prevalence of the entire population of Magoda village in 1992, and based on the promising results from the Gambia [
41], a Maloprim
® (3.125 mg pyrimethamine and 25 mg dapsone, GlaxoSmithKline, UK) trial was the first study to be implemented in Magoda in 1993–94 to test the prophylactic effects of weekly administration of the drug on parasite infections, anaemia and splenomegaly [
40]. Children aged <10 years old were given either one or two tablet(s) of Maloprim
® to under-fives and those aged 5–9 years, respectively and followed-up weekly to determine their general condition and malaria infection status [
40]. Because of increasing resistance to chloroquine (CQ) in Muheza district which was then the government policy as first line treatment of uncomplicated malaria [
42], patients who developed parasitaemia during follow-up under this trial were treated with either sulphadoxine/pyrimethamine (SP) or amodiaquine (AQ) [
40].
Subsequently, a mobile clinic, visiting Magoda village twice weekly was introduced in 1994 to estimate if a rebound of malaria may arise after termination of the Maloprim
® trial, and the mobile clinic was continued till November 2004 to provide medical services to the community. The mobile services and malaria monitoring were extended to the neighbouring village of Mpapaya from 1997. During the visits by the mobile clinic, suspected cases of malaria had blood smears taken, and treatment was given on clinical grounds using either SP or AQ, because blood smears were only examined after returning to the research laboratory, for practical reasons. Patients not symptomatically treated but with malaria parasites in blood smears, were given antimalarials on the following day after microscopy results were obtained. Furthermore, repeated
in vivo drug efficacy trials were carried out and involved patients with uncomplicated malaria (confirmed by microscopy) in 1994–1996 [
43‐
45], 1997 [
46] and 1998–1999 [
38] testing SP, and later AQ in 2003 [
47]. In December 1998, each sleeping bed in all households in Magoda (on request from the villagers), received a permethrin-treated net (ITN) while in Mpapayu, deltamethrin-treated nets were distributed in March 2001; all ITNs were re-impregnated twice a year [
38] until 2003. Throughout the observation period (up to 2004), children and adults with fever and other ailments requiring hospitalization were offered admission to the local designated district hospital of Muheza located 10 km from the village. By 2005, a local dispensary was built to replace the mobile clinic while artemisinin-combination therapy (ACT, with artemether/lumefantrine, AL) was introduced in January 2007, following changes of malaria treatment policy [
48]; and rapid diagnostic test for malaria (RDTs) were introduced in 2012 (NMCP, Personal Communication).
Recruitment of participants and data collection
For this paper, complete data sets with personal details which were recorded on a morbidity questionnaire including age, sex and residence, parasite infection status for asexual and sexual stages, haemoglobin (Hb) levels, splenomegaly and ITN use and insecticide impregnation status were available for CSS conducted between 1999 and 2012 (Table
1). Additionally, summarized data were retrieved from reports covering the surveys conducted in 1992 [
49] and for other years (1995, 1996, 1997 and 1998), quarterly/annual reports of Amani Centre of the National Institute for Medical Research (NIMR) were used to get the prevalence and density of
P. falciparum infections. The CSS involved individuals aged between six months and 20 years. The surveys were conducted during (April/June) or after (July/September) the long rain season between September 1992 and June 2012 as described elsewhere [
38,
40,
49,
50].
Table 1
Prevalence of
P. falciparum
gametocytes, splenomegaly and geometric mean gametocyte density among individuals who participated in malaria cross sectional surveys in Magoda and Mpapayu villages in Muheza district from 1999 – 2012
1999 | 383 (40.4) | 38.6 | 12.1 | 31 | 232 (39.2) | 40.5 | 22 | 40 |
2000 | 352 (27.8) | 21.6 | 6 | 55 | 226 (34.5) | 43.4 | 8 | 85 |
2001 | 352 (42.4) | 45.6 | 7.8 | 95 | 202 (39.1) | 52 | 5 | 62 |
2004 | 392 (36.0) | 23.7 | 5.6 | 21 | 237 (38.7) | 23.6 | 1.7 | 16 |
2008 | 422 (41.2) | 0.7 | 5 | 32 | 268 (36.2) | 1.6 | 2.2 | 49 |
2009 | 512 (33.4) | 2.5 | 3.1 | 23 | 267 (38.2) | 1.9 | 3 | 64 |
2010 | 480 (33.7) | 0.2 | 0.4 | 121 | 273 (34.4) | 1.1 | 0.7 | 22 |
2011 | 482 (26.6) | 0.2 | 0.2 | 560 | 259 (28.2) | 0.8 | 0 | 0 |
2012 | 459 (26.0) | 0.7 | 0.4 | 52 | 252 (26.6) | 0 | 0.4 | 32 |
Each recruited person was examined by a medical doctor and had axillary temperature taken using a digital thermometer, and splenomegaly was assessed as previously described [
51]. Blood samples were collected from each of the study participants by venous bleeding or finger prick for parasitological examination and other laboratory analyses. Thick and thin blood smears were prepared and dried in the field, and later brought to the laboratory for further processing. For the surveys conducted before 2004, blood smears from participants with fever (axillary temperature ≥37.5°C) were examined after the end of each day and those with malaria parasites were treated with SP or AQ on the following day. For the CSS conducted between 2008 and 2012, all participants were tested with RDTs and those with positive test results and other symptoms of uncomplicated malaria were treated with AL as described by Ishengoma
et al.[
50]. Asymptomatic individuals with positive RDT results were not treated with anti-malarials, but were advised to report to the nearest dispensary in Magoda in case they fell sick.
Laboratory analysis
Blood smears collected in the CSS were stained using 10% Giemsa solution for 30 minutes and examined at a magnification of 1,000× to detect parasite species and to determine parasitaemia. Quality control of results of microscopic examination of blood smears were performed as described elsewhere [
50,
52]. Briefly, each blood smear was examined by two technicians blinded of the patient status and RDT results. Parasite density was calculated as the average counts of the two technicians if their results did not differ by more than 50% for blood smears with ≥400 asexual parasites/μl of blood. For blood smears with < 400 asexual parasites/μl, any counts of each of the two technicians was accepted and used to calculate the average parasite density. Blood smears with discordant results were re-examined by a third technician and the results of any two technicians was accepted as explained above. Further discordant smears were resolved by a team of three technicians who re-examined such smears at the same time. For the studies done between 1999 and 2004, haemoglobin (Hb) levels were measured by PCV and later converted into Hb levels (g/dl) as described elsewhere [
53]. In the CSS conducted from 2008 to 2012, Hb was measured using a HemoCue (HemoCue, Ångelholm, Sweden) and anaemia was described as Hb < 11 g/dl.
Ethical considerations
The studies which provided data for this paper were approved by the Medical Research Coordination Committee of NIMR. Verbal and written informed consent was sought from patients or parents/guardians in case of children. Village meetings were held to explain and discuss the study plans with community members and feedback including results of previous surveys was given to the communities through the above meetings together with a written report.
Data analysis
Microsoft Access database was utilized for data management with double entry, validation and cleaning of data collected from 2008 to 2012; followed by analysis using STATA version 11 (STATA Corp Inc., TX, USA). Data collected in 1999–2004 were managed using Epi-Info (CDC, Atlanta, USA) and Microsoft Excel, and later transferred to STATA version 11 (STATA Corp Inc., TX, USA) for analysis. The analysis involved comparison of different variables such as parasite positivity, parasite density, gametocyte carriage, gametocyte density, splenomegaly, mean Hb, anaemia and ITN use in order to test for the differences across years and between the two villages. Categorical data were compared using chi-square test while continuous variables were tested using Students t-test. Logistic regression analysis was used to assess the effects of different interventions (using data collected from 1999 to 2012), such as deployment and distribution of ITNs, anti-malarial drug trials and changes of malaria treatment policy (chloroquine in 1999 – 2001, SP from 2001 to 2004 and AL in 2008 – 2012) on the risk of being infected with malaria parasites after adjusting for age and fever status (if study participants had fever or not). The different interventions which had specific point of deployment such as use of ITNs were coded as binary variables and interventions which covered a particular period were similarly labeled with reference to the duration of such interventions; and these variables were used in logistic regression analysis. Non-normal continuous variables such as parasite density were log transformed to normality. Linear regression models were used in modeling the relationship between Hb concentration and parasite density as response variables against explanatory variables. P-value <0.05 was considered significant.
Discussion
Although the information on the current burden of malaria in most of the malaria endemic countries is fragmented or lacking, there is clear evidence that the burden of malaria has declined in some endemic countries [
2]. For Tanzania, a significant decline of malaria burden has been confirmed on the islands of Zanzibar after deployment of intensive malaria interventions including ACT, ITNs and IRS [
16,
17], whereas data on the changing epidemiology of malaria on the mainland are less well documented and relatively few reports exist except from some parts including north-eastern Tanzania. Furthermore, previous [
28] and the most recent Tanzania HIV and malaria indicator surveys (THMIS) of 2011/2012 showed a significant decline of malaria burden, whereby the average parasite prevalence in under-fives decreased from 18% to 9% between 2007 and 2011 (MoH, unpublished data). These nation-wide surveys are aggregated at the regional level and cannot give detailed patterns of malaria at small geographic coverage such as the district.
Studies of malaria (and other parasitic diseases, such as lymphatic filariasis) including repeated cross-sectional studies were conducted in the villages of Magoda (since 1992) and Mpapayu (since 1997) in Muheza district of north-eastern Tanzania and have enabled long-term observations of trends in malaria for over 15–20 years. The studies also showed the possible impact of the deployed interventions at the community level and changing national malaria control policies. These villages and other parts of Muheza have been reported to have high burden of malaria, as well as high levels of resistance to different anti-malarial drugs [
43‐
47]. Treatment-wise, in Magoda village, SP and AQ were deployed early in 1994 during the Maloprim
® trial [
40], and CQ was withdrawn (due to high level of resistance) even before the official change of malaria treatment policy replacing CQ with SP were effected in 2001 [
54].
Overall, the study reported here observed a significant decline of malaria burden over the study period. Although the major decline in the prevalence of
P. falciparum infections occurred between 1997 and 2004 (Magoda) and 1998–2004 (Mpapayu) when more interventions were being implemented, further changes were also observed between 2008 and 2012 in both villages. The peak prevalence of malaria shifted from children aged 5–9 to those aged 10–19 years old in 2008–2012, a trend also observed in the highlands of the neighbouring district of Korogwe which generally have low malaria transmission [
51]. The prevalence of splenomegaly and gametocytes also declined during the same period.
The decline in prevalence observed before 1999 could possibly be due to better services provided by the research team while the changes observed between 1999 and 2004 could also be attributed to drug trials, which were implemented from 1994 [
43‐
47], and later deployment of free ITNs [
38]. During these trials, the research teams were always available in the communities and provided health services to other patients including treatment of malaria, regardless of their involvement in the studies. The deployment of ITNs at different time points in the two villages, while other services were otherwise equal, enabled an estimation of the impact of ITNs on malaria and prevalence of drug resistance markers [
38]. High bed net coverage has also been associated with reduced malaria morbidity and mortality as reported in previous studies [
55,
56].
Presence of research team in the villages, deployment of a mobile clinic and other services provided by village health workers would also possibly reduce malaria burden through prompt diagnosis and treatment and referral to Muheza district hospital for free services, when needed. Furthermore, the records collected between 1992 and 2004 showed that the number of deaths declined significantly and cerebral malaria virtually disappeared from Magoda and Mpapayu village (Ib C. Bygbjerg, personal communication). Similar studies conducted in the neighbouring district of Korogwe also showed that improved health services at the community level through deployment and use of village health workers under the supervision of trained medical personnel was associated with a significant reduction in the burden of malaria [
25,
27]. Although each of the interventions deployed during the aforementioned period might be attributed to the reduction in malaria burden observed by 2004, the contribution of each of the interventions implemented could not be fully quantified by this study.
In the CSS conducted in 2008 – 2012, the prevalence of
P. falciparum infections dropped to <10% in Magoda and <5% in Mpapayu from 2011. Of the recent interventions, AL was introduced as first-line drug for treatment of uncomplicated malaria in January 2007and free ITNs were distributed in Magoda in 2009 through the US President Malaria Initiative (US PMI), followed by long-lasting nets in 2012 (Zuberi Mohamedi, personal communication). However, the coverage and usage of bed nets remained below 74%, i.e. far below the levels attained by 2004. It is thought-provoking that coverage and use of ITNs were much higher before 2004, when ITNs were distributed by the project compared to the most recent coverage under the national universal campaigns. Factors other than ITNs and deployment of AL could possibly be driving the most recent changes as also reported by Okiro
et al.[
34]. These include changes in climatic variables, such as rainfall and temperature, improvements of socio-economic status, health care and human land use activities, which could possibly be associated with declining malaria transmission. However, such factors were not assessed in this study and are worth taking into account in future studies.
The decline in other indices of malaria transmission intensity such as the prevalence of splenomegaly and gametocytes indicate that these areas formerly known as holo/hyper-endemic have now become low malaria transmission settings within very few years as also recently shown in the neighbouring district of Korogwe [
25,
27]. The shift of parasite prevalence to older children and adolescents in 2008–2012 might be attributed to delayed development of immunity among study participants due to reduced exposure to infections [
57,
58].
The significant decline in the prevalence of
P. falciparum gametocytes observed in the CSS done between 2008 and 2012 could partly be attributed to the effects of ACT together with ITNs which reduce malaria transmission. Previous studies have shown that artemisinins reduce the parasite biomass drastically and thus reducing the rate of development and gametocyte carriage [
59]. Such effects could be responsible for the low prevalence of gametocytes in the study areas observed in the recent CSS. Despite an increase in mean Hb levels and a decrease in the prevalence of anaemia before 2000, there were no significant changes over the study period. However, a slight increase in mean Hb and a decline in the prevalence of anaemia were observed in 2012 which could possibly indicate that there was a slight lag between the decline of malaria prevalence and downward trends of anaemia in these communities. Further monitoring will be crucial to document these important changes in malaria prevalence and their influence on other malariometric indices.
Although vector population studies have not been done in the study villages in recent years, studies conducted in the neighbouring villages in Muheza and Tanga districts showed that the density of malaria vectors has progressively declined leading to low transmission in this part of the country [
60]. Furthermore, the composition of malaria vector populations in Muheza has significantly changed whereby the anthropophilic
Anopheles gambiae s.s. have been replaced with the zoophilic
Anopheles arabiensis[
61]. These studies suggest that the changes in vector density and composition could be responsible for the low malaria burden in the area and other parts of Tanzania as recently revealed by the nation – wide malaria indicator surveys of 2011(MoH, Unpublished data). Thus, the observed trends of declining malaria burden could be linked to the interventions deployed in the study communities including ITNs, drug trials, better health care through mobile clinic, changing national malaria control policies and other unknown factors which could have indirectly caused the dramatic reduction in malaria transmission. However, the low burden of malaria observed in these villages cannot be generalised to reflect the situation in other parts of Tanzania due to high level of interventions which were deployed in the area over the study period.
Acknowledgements
The authors wish to thank members of the survey teams (Samuel Gesase, Julius Massaga, Daniel Minja, Hamis Msangeni, Julius Mhina, Juma Akida, Samuel Sembuche, Celine Mandara, Fredy Saguti, Vito Baraka, Sudi Lwitiho, Deo Maiga, Rashidi Madebe, Ezekiel Malecela, Johari Sadi, Juma Tupa, Magreth Munisi, Zaina Maumba, Beatrice Semng’indo, Neema Barua, Benson Swai, Silas Msangi, Sylivia Masawe, Tilaus Gustav, Lydia Masawe, Thompson Mwampamba, Hatibu Athumani, Halima Mpambile, Stella Mkandawile, Allen Mrango, Frank Mnango, Seth Nguhu, Zahabu Kauzeni, Simba Athumani and Salimu Tembo) and the data management team at Tanga (Filbert Francis, Benson Sway, Chris Msokame, Mary Lukindo and Fides Mumburi). Study participants, community members, field workers in (Magoda and Mpapayu) and health authorities in Muheza district are acknowledged for their cooperation and support provided to the study teams. Funding for the CSS from 1992 and 2004 was provided by DANIDA through the ENRECA Programme. Surveys done between 2008 and 2012 were supported by DANIDA through grant No. N_08_KU and the National Institute for Medical Research (NIMR) through the Health Research Users Trust Fund (HRTF), and the Thorvald Madsen’s Foundation. The Director General of the National Institute for Medical Research (NIMR) is thanked for the continued support and the permission to publish this paper.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
DSI, MML, MA and ICB conceived of and designed the study; DSI, MDS, BPM and MML conducted the field work and supervised the laboratory analyses. BPM and DSI participated in data management and analysis. DSI, MA, MML and ICB wrote the manuscript. All authors read and approved the manuscript.