Tumour necrosis factor allele variants and their association with the occurrence and severity of malaria in African children: a longitudinal study

Study area and population

Data from a randomized trial that aimed to assess the effect of supplementation with
zinc and other micronutrients on malaria rates was used. The trial was conducted between
February 2008 and March 2009 in four rural villages in Handeni District, northeastern
Tanzania. Residents primarily belong to the Wazigua and Wabondei Bantu tribes, but
settlement of migrant plantation workers has resulted in a mixture of tribes with
different origins and much intermarriage. The area is mainly populated by poor farmer
families involved in subsistence farming. Malaria transmission is intense and perennial,
with nearly all infections being due to Plasmodium falciparum20]. Apart from several local traditional healers, the research dispensary was the only
health facility in the area. The study was approved by the Ethical Review Committee
of Wageningen University, The Netherlands (Approval number: NL: 04/07) and the National
Health Research Ethics Review sub-Committee, Dar es Salaam, Tanzania (Approval number:
NIMR/HQ/R.8a/Vol. IX/540). Informed consent was obtained from community leaders, local
government officials and parents or guardians. Further details of this trial are described
elsewhere 21].

Recruitment

All resident children were eligible for randomization when aged 6–59 months and with
a height-for-age z-score in the range ?3 SD to 1.5 SD. Children with haemoglobin concentration
70 g/L, signs of chronic illness, and those unlikely to remain permanently resident
or comply with the supplementation for the duration of the trial, or whose parents
or guardians declined consent, were excluded from the study. Venous blood samples
were collected in EDTA tubes and centrifuged immediately. An aliquot of 90 ?L erythrocyte
sediment with the buffy coat was mixed with 90 ?L phosphate-buffered saline and 180 ?L
of DNA stabilizing buffer (AS1; Qiagen, Hilden, Germany) and stored at 4°C for subsequent
genotyping. Plasma samples were stored in liquid nitrogen in the field and at ?80°C
during transport and subsequent storage until biochemical analysis in The Netherlands.
Haemoglobin concentration was measured in an aliquot of whole blood by a haematology
analyser (Sysmex KX21, Kobe, Japan). Plasmodium infection was detected in fresh blood by rapid dipstick test (CareStart, Access Bio,
Monmouth Jct, USA). Children with a positive test result were treated immediately
with artemether-lumefantrine. The location of the child’s homestead was determined
using a global positioning system. Further details about recruitment procedures are
reported elsewhere 21].

Experimental intervention

Children were randomized within six strata defined by Plasmodium infection (binary) and age class (6–17 months, 18–35 months and 36–60 months) and
randomly permuted blocks with size randomly selected of four or eight. They then received
daily supplements with either zinc alone (10 mg as gluconate), multi-nutrients without
zinc, zinc combined with multi-nutrients or placebo. Supplements, in the form of powder
in colour-coded capsules, were contained in blister packs, and administered orally
after suspending capsule contents in clean water or breastmilk. Supplementation was
performed by local community volunteers, who reported adherence daily to field staff
at the research dispensary.

Follow-up and case detection

A clinical officer was on duty at the research clinic day and night. At recruitment,
parents or guardians were requested to bring participating children to the dispensary
immediately when detecting fever or any other illness during the intervention period.
In samples collected at baseline and from sick children, the presence of parasite-specific
lactate dehydrogenase (P. falciparum and other Plasmodium species) was detected by rapid tests (CareStart, G0121, Access Bio, Monmouth Jct,
NJ). Axillary temperature was measured using an electronic thermometer and dipstick
tests were administered for children with guardian-reported fever; for those with
positive test results, plasma samples were collected and measured whole-blood C-reactive
protein concentrations using a point-of-care test. Plasma was stored as described
for the recruitment procedure. Artemether-lumefantrine (Novartis Pharma, Basel, Switzerland)
was administered to any child with current Plasmodium infection upon enrolment, or with reported fever and a positive dipstick test result
during the follow-up period.

Laboratory procedures

For children who presented with malarial episodes, whole-blood concentrations of haemoglobin
and C-reactive protein were measured using point-of-care tests (HemoCue, Ängelholm,
Sweden and QuikRead, Orion Diagnostica, Espoo, Finland, respectively). Plasma concentrations
of P. falciparum-specific histidine-rich protein-2 (HRP2) in samples collected during the first malaria
episode were measured using a commercial enzyme-linked immunosorbent assay kit (Malaria
Ag Celisa; Cellabs, Brookvale, NSW, Australia). Plasma concentrations of C-reactive
protein were measured (Meander Medical Centre, Amersfoort, The Netherlands) on a Beckman
Coulter Unicel DxC880i system according to the manufacturer’s instructions. Genotypes
were determined using Illumina’s VeraCode™ GoldenGate Genotyping Assay on a BeadXpress™
platform. TNF variant allele clustering was assessed visually to determine success of genotyping.
Further quality control cut-offs were: a GenCall Score of 0.5 and a call rate of ?0.95
22]. For the variant TNF_?1031,
the reference allele was T and the alternate C. Individuals homozygous for the TNF_?1031
reference allele (T) are hereon described as wild type while those homozygous for
the alternate allele (C) are referred to as homozygote mutant. For the variant TNF_?308
, the reference allele was G and the alternate was A. Individuals homozygous for the
TNF_?308
reference allele (G) are hereon described as wild type while those homozygous for
the alternate allele (A) are referred to as homozygote mutant.

Statistical analyses

Linkage disequilibrium (LD) pattern assessment as well as haplotype analysis was conducted
using HAPLOVIEW 23]. Anthropometric indices were calculated using Epi Info software 24] All analyses were performed using SPSS (v15.0 for Windows, SPSS, Chicago, IL, USA),
CIA (v2.1.2) 25] and STATA (v11; College Station, TX, USA). For each TNF genotype, Fisher’s exact
test was used to assess whether populations were in Hardy–Weinberg equilibrium. The
differences in baseline characteristics were calculated using the homozygote wild-type
group as the reference for each variant using CIA (v2.1.2). Normality of variables
was assessed by visual inspection of histograms. Because variables of interest (age,
haemoglobin concentrations, distance from homestead to dispensary and anthropometric
indices) were normally distributed, means, SDs and 95% CIs were reported.

Association of TNF genotype with malaria rates

The primary outcome, an episode of malaria, was predefined as a positive result for
either a pLDH or a HRP2 dipstick test with any of the following: (a) confirmed fever
(axillary temperature ?37.5°C as measured by electronic thermometer); or, (b) guardian-reported
but unconfirmed 24-h history of fever in the presence of inflammation (whole blood
C-reactive protein concentrations 8 mg/L), separated by at least 14 days from a previous
malaria episode. Incidence per TNF genotype and incidence ratios based on time to first episodes, with wild-type homozygotes
as reference group was calculated. In the primary analysis, group rates were compared
using Cox regression with robust estimates of the standard error to account for multiple
episodes within children. The extent to which supplementation with either zinc or
multi-nutrients including iron modulated the magnitude of the association of TNF genotype on malaria rates was explored. The extent to which adjustment for baseline
factors that were a priori expected to be prognostic for malaria (Plasmodium infection status, distance between homestead and clinic [continuous variable], height-for-age
z-score [continuous variable], mosquito net use [binary variable]) and experimental
intervention modulated the estimated association of genotypes were explored. In this
adjusted analysis, experimental intervention as a binary variable indicating pooled
groups receiving multi-nutrients (with or without zinc) and pooled groups receiving
no multi-nutrients (with or without zinc) was included. Kaplan–Meier analysis with
Peto tests were used to assess associations of TNF genotypes with time-to-first episode of malaria.

TNF genotype modulation of the association between Plasmodium falciparum infection and haemoglobin concentration at baseline

Since the additive genetic model was considered, for each genotype, two dummy variables
were used to indicate the three classes, resulting in two interaction terms per variant.
Regression models were then used to assess the modulation of genotypes on the associations
between P. falciparum and haemoglobin concentration. Main terms for P. falciparum infection and genotype dummies were included, and adjusted as pre planned for age
class, mosquito net use, height-for-age z-score, and distance between the child’s
homestead and the dispensary in regression analyses.

TNF genotype modulation of the association between Plasmodium falciparum infection and haemoglobin concentration at first episode of malaria

The additive genetic model was considered so for each genotype, two dummy variables
were used to indicate the three classes. TNF genotype modulation of the association
between P. falciparum infection and haemoglobin concentration at first episode of malaria was then assessed
by analysis of variance (ANOVA) and adjusted for age class, height-for-age z-score
and distance between the child’s homestead and the dispensary and treatment group.