Available via license: CC BY 4.0
Content may be subject to copyright.
Plasmodium falciparum parasites with histidine-rich protein 2 (pfhrp2) and pfhrp3 gene
deletions in two endemic regions of Kenya
Khalid B Beshir*1, Nuno Sepúlveda1,2, Jameel Bharmal1, Ailie Robinson1, Julian Mwanguzi1,
Annette Obukosia Busula3, Jetske Gudrun de Boer4, Colin Sutherland1, Jane Cunningham5 and
Heidi Hopkins1
1. London School of Hygiene and Tropical Medicine, London, UK
2. Centre for Statistics and Applications of University of Lisbon, Lisbon, Portugal
3. International Centre of Insect Physiology and Ecology, Nairobi, Kenya. Current affiliation: Kaimosi
Friends University College, Kaimosi, Kenya.
4. At the time of the work reported here: Laboratory of Entomology, Wageningen University,
Droevendaalsesteeg 1, 6708 PB Wageningen, The Netherlands. Current affiliation: Netherlands
Institute of Ecology, Droevendaalsesteeg 10, 6708 PB Wageningen, The Netherlands
5. Global Malaria Programme, World Health Organization (WHO-GMP), Geneva, Switzerland
*Correspondence to Khalid.Beshir@lshtm.ac.uk
Key words: hrp2 deletion, pfhrp2, pfhrp3, malaria, Plasmodium falciparum, rapid diagnostic
test, RDT, HRP2, HRP3, histidine rich protein
Supplementary material
Supplementary Table S1. Data generated using RDT, pfhrp2, pfhrp3 and
parasitaemia (>5 parasites per microliter, n=91).
Analysis of the probability of RDT positivity as function of parasitaemia (in log scale),
presence of Plasmodium falciparum histidine rich 2 and 3 (pfhrp2 and pfhrp3)
respectively) genes using different generalized linear models for binomial responses
(logit, probit and complementary log-log). The general formulation of these models was
the following:
𝑔(𝑥)= 𝑎 + 𝑏 × hrp2 + 𝑐 × hrp3 + 𝑑 × log(𝑃𝑎𝑟𝑎𝑠𝑖𝑡𝑎𝑒𝑚𝑖𝑎),
where hrp2 and hrp3 are binary covariates indicating the presence of pfhrp2 and pfhrp3,
respectively, 𝑔(𝑥)= 𝑙𝑜𝑔 𝑝 (1 − 𝑝)⁄ (logit), 𝑔(𝑥)= Φ(𝑝) (probit), and 𝑔(𝑥)=
log(− log(1 − 𝑝))(complementary log-log). The coefficients a, b, c and d were estimated
by maximum likelihood method using the glm function for R. Models were compared with
each other via Akaike’s information criterion (AIC) and Bayesian Information Criterion
(BIC), where the best model for the data is the one with the lowest estimate for each
measure.
Model
Model comparison
Parameter estimates (Standard Error)
AIC
BIC
Intercept
(a)
hrp2
(b)
hrp3
(c)
Parasitaemia
(d)
Logit
34.22
44.22
-2.45
(1.47)
3.33
(1.21)
-0.03
(1.36)
0.72 (0.34)
Probit
34.36
44.36
-1.55
(0.83)
1.65
(0.55)
0.45
(0.75)
0.36 (0.17)
Complementary
log-log
32.74
42.74
-2.50
(1.08)
1.57
(0.55)
1.09
(0.73)
0.35 (0.16)