Submicroscopic malaria infection is not associated with fever in cross-sectional studies in Malawi

The surveys included all individuals over six months of age from households that were randomly selected in Blantyre, Chikhwawa and Thyolo districts. Briefly, 10 enumeration areas (EA) were randomly selected from each district. The 10 EAs per district were divided into compact segments containing 30 households each. One segment was selected from each EA to be included in the study and the same compact segments were visited for each survey. All households in the segment selected were eligible for the study; however, if one of the households did not meet the criteria for inclusion in the study it was replaced by the nearest household [17]. Same houses were sampled at each visit (survey), but information to aid in tracking survey participants to determine if they had repeated observations was not collected.

The surveys targeted all members of the household who had slept in the house for at least 2 weeks of the month prior to the survey. The analysis included household members aged 6 months old and above. Household visitors and individuals from households where the eldest person present at the time of the survey was a minor were excluded from the survey. Consent to participate in the study was sought from the head of the household and from each adult present. Parents or guardians were asked to provide consent on behalf of the children below the age of 18 years old and assent was sought from children who were 13–17 years old. Once consent and assent were granted, interviewers collected data on demographic and socioeconomic status (SES) variables, malaria preventive measures used, and any anti-malarial medication taken within 2 weeks of the survey. In addition, axillary temperature was measured using digital thermometers, and peripheral blood samples were collected on glass slides (thick and thin films) for malaria microscopy and on Whatman 3MM filter paper (GE Healthcare Ltd., New Jersey, USA) for molecular testing as described previously [22].

The primary outcome of interest was fever. Three fever measures were collected: reported fever in the past two weeks, reported fever in the past 48 h and temperature ≥ 37.5 °C measured at the time of specimen collection. In this study, for the primary analysis, fever was defined as having a fever using at least one of the three fever measures outlined above. A secondary analysis was conducted with objective fever as the outcome, and this was defined as axillary temperature of ≥ 37.5 °C. Submicroscopic P. falciparum infection was defined as a positive result by real-time PCR targeting P. falciparum lactate dehydrogenase gene and a negative malaria microscopy result. Details of the procedures for P. falciparum detection using real-time PCR and microscopy have been previously described [17, 23]. Individuals who tested negative by both microscopy and PCR were included as the comparison group. Participant ages were categorized as follows: 6 months–5 years, 6–15 year and 16 years and above. For SES, ten indicator variables were used to derive a wealth index weighted based on principal component analysis methods as described by Filmer and Pritchett [24]. The survey households were then subdivided into four wealth quartiles based on the weighted indicator score.

Median age in years and proportion of individuals were estimated within strata of sex, SES quartile, season and age group in each district. Proportion of individuals with and without submicroscopic P. falciparum infection and proportion of individuals with and without fever were computed within strata of district, season, and age groups. Mixed effects logistic regression models were used to estimate the association between submicroscopic P. falciparum infection and fever within strata age groups, districts, SES quartile and season, while accounting for clustering within household and EA. In the final model comparing the log odds of fever between submicroscopic P. falciparum infection and no parasitaemia, district, participant age and season were included as covariates. In addition, a term for the interaction between submicroscopic infection and season (submicroscopic infection*season) was included in the final model. To determine if the association between submicroscopic infection and fever would be different when using broad (at least one of these three measures: reported fever in the past 2 weeks, reported fever in the past 48 h and temperature ≥ 37.5 °C measured at the time of the interview and specimen) or narrow (objective fever) definition of fever, we conducted a sensitivity analysis using objective fever (axillary temperature of ≥ 37.5 °C measured at the time of the interview) as the outcome of interest. All data analyses were done using SAS 9.4 (SAS Institute Inc., Cary, North Carolina) statistical software and were conducted at alpha level of 0.05.

A total of 16,650 people participated in the six surveys conducted between 2012 and 2014 in Blantyre, Chikhwawa and Thyolo districts. The median age of survey participants was 13 (IQR: 6–27) in Blantyre, 12 in Chikhwawa (IQR: 6–28), and 13 in Thyolo (IQR: 6–30). The distribution of participants within sex, age groups, and seasonal categories was similar in the three districts (Table 1). Blantyre residents tended to have higher scores on the asset-based SES index. More of the participants in Chikhwawa and Thyolo (34.5% and 31.9% respectively) were in the lowest SES quartile compared to Blantyre (3.3%). In Blantyre District, 59.1% of study participants were in the highest SES quartile compared to only 7.9% in Chikhwawa and 13.3% in Thyolo. The characteristics of the study populations grouped by district are described in Table 1.

Of the 16,650 participants, 1517 (9%) had submicroscopic P. falciparum infections, 1570 (10%) had microscopic infection and 13,563 (81%) had no parasitaemia (Table 2). The limit of detection of PCR and microscopy in this study was ~ 2.7 parasites per microliter and 50–100 parasites per microliter, respectively [23, 25]. When comparing prevalence of infection between districts, Chikhwawa had the highest prevalence of submicroscopic (13%) and microscopic infection (17%). Participants with microscopic parasitaemia had lower median age (10 years) compared to those with submicroscopic infection (14 years) and no parasitaemia (13 years). Children between the age of 6 years to 15 years had high prevalence of submicroscopic (11.6%) and microscopic infection (15.4%) compared to other age groups. Higher prevalence of submicroscopic infections was observed in the rainy season compared to the dry season, and in the lowest to medium wealth quartiles compared to participants in the highest wealth quartile Prevalence of any parasitaemia (microscopic and submicroscopic) was highest in the lowest SES quartile (Table 2).

Overall, 1500 of 16,650 (9%) participants had recent or concurrent fever by objective measurement or self-report. Fever was most common among residents of Chikhwawa District (11.2%), young children aged 6 months to 5 years (12.6%), people in the lowest SES quartile (10.8%), and rainy season participants (10.8%) (Table 2).

The prevalence of fever among individuals with submicroscopic P. falciparum infection was compared to that of individuals with no parasitaemia within strata of district, age group, and season. A total of 14,239 participants were included in these final analyses after excluding participants with P. falciparum parasites seen on microscopy, those with missing PCR results, those who reported taking anti-malarial medications within the previous 2 weeks and those missing results for fever. Of these participants, 1408 (10%) had submicroscopic P. falciparum infections and 1032 (7%) had fever by objective measurement or self-report. Submicroscopic P. falciparum infection was significantly associated with a 49% decrease in odds of fever in the dry season (odds ratio [OR] = 0.51; 95% CI 0.33–0.78), and a non-significant increase in the odds of fever in the rainy season (OR = 1.22; 95% CI 0.94–1.58).

There was no significant association between submicroscopic P. falciparum infection and fever within strata of either district or age groups. Only 1.2% of the population had fever when defined by objective measure of body temperature ≥ 37.5 °C at the time of survey. In the sensitivity analysis limited to objectively measured fever, the association between submicroscopic infection and objective fever was not statistically significant within strata of either season, district or age group, though the directions of the association were generally consistent with those of the primary analysis (Table 3).

The final model was stratified by season and included random errors to account for clustering by household and EA. After adjusting for age and district, those with submicroscopic P. falciparum infections had lower odds of any fever than uninfected people in the dry season (OR = 0.52; 95% CI 0.33–0.82); no statistically significant association was observed in the rainy season (OR = 1.20; 95% CI 0.91–1.59). In the sensitivity analysis limiting the definition of fever to objective measurement of axillary temperature ≥ 37.5 °C at the time of the assessment, the estimated ORs were similar but not statistically significant in either season (Table 4).