Study area description and population selection

The study included data collected between July 2004 and December 2008. The ASMQ intervention was carried out between July 2006 and December 2008 in three municipalities in the Juruá valley, within the state of Acre in Brazil: Cruzeiro do Sul, Mancio Lima and Rodrigues Alves (Figure 1). The municipalities were selected because they had recorded more than 20 P. falciparum malaria cases per month between 2003 and 2004, and had a stable population (defined by < 15 % proportion of imported cases), as well as supportive and cooperative local health authorities.

Acre state, situated in the most occidental region of Legal Amazon (Figure 1) has a population of 680,073 inhabitants; 15.2 % (103,371) of the population live in Cruzeiro do Sul, Mâncio Lima and Rodrigues Alves, and account for 86 % of Acre’s malaria cases. The state of Acre has a tropical climate, with temperatures ranging between 22–34 °C (72–93 °F), relative humidity of 60–85 %, and a rainy season that lasts from October through to April.

Malaria epidemiology in Acre

Anti-malarial drug resistance has not been investigated in this region for almost two decades [14, 15]. Acre borders Peru and Bolivia, where until recently first-line treatment regimens for P. falciparum malaria differed from those used in Brazil [13, 16, 17]. As a result, sustained population influx from these countries may affect the local patterns of drug-resistance [14, 15].

Malaria incidence in some areas of the Amazon Basin rose between 2003 and 2005, which was attributed to several environmental and socio-economic factors, including climatic changes, new urban growth, cattle ranching and agricultural practices associated with deforestation) [16, 18]. Public health strategies promoted by the PNCM were not fully implemented. The Ministry of Health initiated a collaborative task force of health managers in the Amazon region to coordinate population movements and to prioritize malaria surveillance, prevention and control[12, 19].

In the Juruá Valley of Acre, Brazil, P. vivax predominates in urban populations, but in 2009, P. falciparum still accounted for more than half of the malaria burden in isolated river communities [16].

The most important malaria vector in Brazilian endemic areas is Anopheles darlingi, and the most vulnerable group consists of low-income workers (fishermen and gold miners) [20–22]. Anopheles darlingi is the main species captured outside and inside the homes of inhabitants of the Juruá valley, in a proportion of 4:1, respectively. The mean rate of An. darlingi per person/hour was 3.2 indoors and 11.5 outdoors. Both inside and outside feeding preferences are from ~8:00 p.m. until ~0:00 a.m. (Izanelda Magalhães, personal communication).

Malaria and vector control programme

The control strategy of the Brazilian National Malaria Control Programme (Programa Nacional de Controle da Malária, PNCM) is based on early diagnosis and treatment, selective vector control interventions, early epidemic detection and involvement of local government and concerned populations. Malaria is diagnosed by rapid tests based on the immunochromatographic method or Giemsa-stained thick blood smears, according to PNCM training recommendations and quality control guidelines. Malaria treatment in Brazil is free of charge, as per government policy.

In 2000, malaria control programmes were decentralized to individual states of Brazil. In Acre state, as 90 % of total cases of malaria occur in the Juruá valley, a regional coordination body for malaria control was created. This organization includes support teams (diagnosis, epidemiology, vector control, health education and social mobilization), information systems as well as administrative and pharmaceutical assistance. In November 2005, intensified malaria control activities included active systematic screening of the population as routine practice, with weekly screenings performed by healthcare workers. Health surveillance workers in each locality collected thick blood smears, administered therapy and provided health education. These agents also prevented prescription mistakes and sought to improve treatment compliance. Local health facilities supported diagnosis and drug distribution. Drug supply was planned according to the number of cases reported for each type of malaria infection (P. falciparum vs. P. vivax). The Brazilian Ministry of Health regulated drug distribution and provided epidemiological data by mapping drugs received, distributed, and consumed.

By 2006–2007, the studied municipalities had a network of 73 diagnostic laboratories, 475 health surveillance workers, and 229 hospital beds. The private sector was also involved. Control strategy results were monitored by the Epidemiological Surveillance Information System Malaria module online database (SIVEP-malaria). Epidemiological and service indicators were evaluated on a weekly basis by the management team and all supervisors, ensuring that all information was disseminated widely.

Vector control strategies included outdoor thermo-nebulization and indoor chemical residual spraying with cipermetrine, and environmental management of larvae control until 2006. Biolarvicide was used in water tanks from December 2005 to July 2007. As of October 2006, the strategy focused mainly on vector control, with the goal of using indoor chemical spraying in 80 % of the highest risk area homes. In December 2007, an intervention with long lasting impregnated bed nets was initiated, where 7,000 insecticide-treated bed nets were distributed to cover 100 % of the population in 13 localities selected on the basis of epidemiological criteria: high disease burden, high proportion of P. falciparum infections, population age range, continuous access to treatment and diagnosis, good management of information systems, evidence of transmission inside the homes, and acceptability of bed nets to the population. These localities also had the least reduction in malaria incidence in previous years compared with other localities in the Juruá valley. Adequate anti-malarial drug control and supplies were available through the cooperation of health authorities and malaria control programs.

Inclusion and exclusion criteria

All patients aged ≥6 months with an uncomplicated P. falciparum infection diagnosis, confirmed by thick smear enrolled in the public health system of the intervention municipalities were eligible for the study. The exclusion criteria were pregnancy and presence of clinical danger signs and other clinical features of severe malaria [23].

Treatments and study procedures

There were no restrictions to the use of other medications, except for other anti-malarials and drugs with known interaction. Prior to study initiation, malaria health workers were trained in order to prevent uncontrolled use of other anti-malarials in the study areas. Special attention was paid to MQ single tablet therapy, which is no longer recommended and has been withdrawn from Acre state. Standard treatment with quinine, doxycycline or primaquine was available in case patients declined trial participation.

Participating subjects who met the inclusion criteria were treated with FDC ASMQ (manufactured by Farmaguinhos), according to age, using either a low dose (LD) tablet 25 + 50 mg (25 mg of artesunate and 50 mg of mefloquine base, as 55 mg mefloquine hydrochloride) or a high dose (HD) tablet: 100 + 200 mg (100 mg of artesunate and 200 mg of mefloquine base, as 220 mg of mefloquine hydrochloride).

Dosing was daily for 3 days and stratified by age: (i) 6–11 months of age: 1 ASMQ LD tablet, (ii) 1–6 years of age: 2 LD ASMQ tablets, (iii) 7–13 years of age: 1 HD tablet, and iv) 14 years of age and older: 2 HD tablets. Most of the time, treatment was directly observed, but for patients living in remote areas, only the first dose was directly observed and the remainder of the treatment was taken home. At study entry, diagnoses were performed by rapid test or thick smears, and patients were enrolled on the basis of a positive result. During the study period, rapid tests were performed in remote areas, but slides were collected to confirm the result by thick smear.

During the first six months of the study, samples of positive and negative malaria smears from every microscope technician were reread in blinded conditions for quality control, according to PNCM standard operational procedures. From November 2006, every positive only P. falciparum slide was revised, identified by a SIVEP number and filed under the month and notification area. As a routine procedure, 10 % of negative slides were re-examined. When conflicting results occurred, which was less than 1 % of cases, the microscopy technician was advised to attend a training programme.

The procedure to identify double entries and cure-verifying slides used the RecLink software duplicity routine [24, 25] that generates a unique code for every individual. In order to improve software performance, records must be of good quality; therefore, the RecLink routine was alternated with manual verification procedures and repeated about 30 times. Every patient had the earliest positive slide defined as diagnosis date, and any slide between day 7 and day 40 after diagnosis was considered a follow-up slide.

Safety evaluation

Adverse events monitoring was done by spontaneous reporting through the local pharmacovigilance system. In order to record these events, standard national regulatory agency forms were made available at the health facilities, diagnosis posts, and to every health surveillance agent.

Ethical approval

The study was approved by the São Paulo University Ethics Committee. Participation in the study was entirely voluntary. Patients were informed of the risks and benefits relating to the study, and participants could withdraw from the study at any time. Written informed consent was obtained from all literate patients or guardians. Illiterate patients or guardians recorded their consent with a thumbprint. A witness confirmed both types of consent. Before starting the study, a state health team visited the communities involved to describe study objectives and methods. Communities and local healthcare providers also gave approval.

Statistical analysis

In order to address the impact of the introduction of ASMQ in the Juruá valley region, three outcome variables were analysed: monthly incidence rates of P. falciparum malaria, the ratio between monthly incidences of P. falciparum and P. vivax malaria, and the monthly hospital admission rates due to malaria (ICD-10: B58 to B54). The period of analysis ranged from January 2004 to December 2008 (before and after introduction of FDC ASMQ).

Because the time series spanned five years, and the outcome variables were based on monthly incidence of the disease, the model had to take into account the effects of trend and seasonality.

The effect of ASMQ on the mean level of each outcome time series was evaluated in terms of main effects and interactions. Three main effects were evaluated, years, months and intervention, whilst the interaction involved intervention and months. The model, therefore, contemplated the time trend based on indicator variables for years and seasonality, with indicator variables for months, and the intervention impact with an indicator variable for the whole period after the introduction of ASMQ.

Where $\left\{Y_{\mathit{ij}}:i=1,\dots ,12;j=2004,\dots ,2008\right\}$is the outcome time series,$\left\{M_i:i\ne 7\right\}$are month indicator variables, $\left\{A_j:j>2004\right\}$are year indicator variables, $I_{\left\{ASMQ\right\}}$ is the intervention indicator variable, β₀ is a common intercept or the baseline (here July 2004), $\left\{{\beta }_i:i\ne 7\right\}$ are the monthly effects, $\left\{{\gamma }_i:i\ne 7\right\}$ are the yearly effects, α is the intervention effect, and $\left\{{\delta }_i:i\ne 7\right\}$ are the interaction effects between months and intervention. The offset was either the logarithm of the at risk population for rates, or the incidence of P. vivax malaria for the P. falciparum/P. vivax ratio.

The above is the expression for the second modelling approach. To obtain the expression for the first approach the interaction terms should be dropped from the formula.

As July 2006 was the first month after the start of the intervention, the month of July was chosen as the reference for the monthly pattern. 2004 was selected as the baseline year, as it represents pre-epidemic and pre-intervention levels of the time series. The intervention was implemented in June 2006, so the intervention period in the statistical analysis was regarded as beginning in July 2006 and ending in December 2008. The impact of the ASMQ intervention was assessed in comparison with the baseline, adjusting for the effects of other years and months. Thus, the coefficients of the remaining months and years represent the variation on the log-incidence rates in comparison to July 2004. The monthly counts were analysed and the model parameters estimated according to a quasi-Poisson estimation procedure and took into account an offset variable. Further, residual diagnostics were performed for each of the six adjusted models.

The two modelling approaches are presented as they address different questions;

- the main effect model establishes that the intervention had an effect on the mean level of the time series,

- whereas the main effect + interaction model establishes if the intervention had an effect on the monthly pattern (seasonality) of the time series, hence altering the mean level of the time series, albeit indirectly.

Data were analysed using the following software: Tableau 3.5, Microsoft Office Excel 2003, RecLink, and the R (R Development Core Team 2011) version 2.11.

Description of population

Plasmodium falciparum incidence rates and efficacy evaluation

The P. falciparum malaria cumulative incidence rates (per 10,000 inhabitants) over one year were stratified by age range in the Juruá valley (Table 2). Following the introduction of ASMQ treatment in 2006, a large decrease in the P. falciparum yearly incidence was observed across all age groups of subjects living in the study area. The P. falciparum malaria incidence rate had sharply increased in the Juruá valley during the second half of 2005, with an epidemic peak at the beginning of 2006 before the introduction of ASMQ treatment (Figure 2a).

Year (population at risk)	<1 year	1 to 6 years	7 to 13 years	≥14 years
2004 (96496)	5.58	56.43	55.56	60.22
2005 (106882)	22.64	148.58	161.81	164.33
2006 (109827)	67.95	298.59	327.63	307.84
2007 (112755)	22.88	68.15	79.64	71.44
2008 (103799)	13.38	38.58	38.01	33.61

Monthly incidence of P. falciparum malaria increased steadily over the first half of the study period, with a subsequent steady decline (Figure 2a). The peak P. falciparum malaria incidence occurred a few months before the introduction of FDC ASMQ therapy. The P. falciparum/P. vivax ratio increased in the same period up to the initiation of intervention, when there was a sharp drop (Figure 2b). The peak P. falciparum/P. vivax ratio coincided with the month of introduction of ASMQ. The pattern of malaria hospital admission rates represents a mix of the previous two series, however with a substantial reduction in the mean level at the end of the study period (Figure 2c).

Model parameter estimates transformed to the original scale are presented in Additional file 1, namely rate ratios in the case of incidence and hospital admissions, and ratios in the case of the P. falciparum/P. vivax ratio. Mutually adjusted increases (decreases) on the mean levels of the outcome are shown for a given month, year or intervention period. The different columns represent the adjusted model according to what was described in the methods section, thus six models altogether, two for each outcome. The two first columns correspond to incidence rates, the next two to the P. falciparum/P. vivax ratio, and the last two to hospital admission rates. Under each outcome, the first column shows results based on the first approach (main effects) whilst the second column presents results of the second approach (main effects and interactions). In order to address the difference between both approaches, the number of parameters fitted under each model and its quasi-deviance statistics are shown at the bottom of the table.

There was a significant decrease in the mean level of the time series for the three outcomes after the introduction of ASMQ, namely 0.34 (95%CI 0.20 – 0.58), 0.67 (95%CI 0.50 – 0.89), and 0.53 (95%CI 0.41 – 0.69) for incidence rates of P. falciparum, the P. falciparum/P. vivax ratio, and for malaria hospital admission rates, respectively. These results are adjusted for trend and seasonality (i.e. under the first modelling approach). It is important to notice the yearly effects pattern: for P. falciparum malaria incidence rates, 2005, 2006, and 2007 had significantly higher levels than 2004, whereas in 2008 there was no significant increase or decrease compared to 2004; for the P. falciparum/P. vivax ratio, 2006 had a significantly higher level than 2004; and for malaria hospital admission rates, 2005 and 2006 had significantly higher levels than 2004, but 2008 had significantly lower levels than 2004. In all outcomes, there was no statistically significant effect for months.

In addition, there was a marked change in the seasonal pattern of all three outcomes (i.e., under the second modelling approach). The incidence rates in the months of November (3.87, 95%CI 1.82 – 8.22) and December (4.50, 95%CI 2.14 – 9.46) before the intervention showed increases compared to July, whereas there were decreases of 0.74 (95%CI 0.42 – 1.32) for November and of 0.56 (95%CI 0.30 – 1.05) for December, after the intervention. A similar picture emerged for the P. falciparum/P. vivax ratio regarding the months of October, November, and December. It is important to notice that usually October marks the beginning of the rainy season in the Juruá valley region (Additional file 1).

Figure 3 shows the yearly effects for the three outcomes. Compared to 2004, the annual effect of the incidence rate of P. falciparum is statistically significant for every year except 2008, suggesting a return to pre-epidemic levels (Figure 3a). Regarding the P. falciparum/P. vivax ratio, Figure 3b shows that in 2007 the yearly effect of this outcome had returned to pre-epidemic levels. Finally, Figure 3c shows a marked and significant decrease in the yearly effect of malaria hospital admissions from 2007 onwards, suggesting that the intervention may have had a greater impact for this outcome.

Figure 4 shows the observed and predicted values for the three outcomes, under the different statistical approaches. Clearly under the second approach the models adjust more closely to the observed data, therefore justifying the inclusion of interaction between months and intervention.

Safety evaluation

No serious adverse events, including deaths, relating to the use of fixed-dose ASMQ were reported. A single non-serious adverse report was recorded in 2007 in a seven-year old patient who experienced dizziness and vomiting, which are adverse events associated with MQ. Patient exposures were pooled from the SIVEP-malaria database. No direct adverse event reports were made to the toll-free pharmacovigilance telephone number for Farmanguinhos or to the national regulatory agency, ANVISA.