Unusual Localization of Blood-Borne Loa loa Microfilariae in the Skin Depends on Microfilarial Density in the Blood: Implications for Onchocerciasis Diagnosis in Coendemic Areas

Abstract Background The diagnostic gold standard for onchocerciasis relies on identification and enumeration of (skin-dwelling) Onchocerca volvulus microfilariae (mf) using the skin snip technique (SST). In a recent study, blood-borne Loa loa mf were found by SST in individuals heavily infected with L. loa, and microscopically misidentified as O. volvulus due to their superficially similar morphology. This study investigates the relationship between L. loa microfilarial density (Loa MFD) and the probability of testing SST positive. Methods A total of 1053 participants from the (onchocerciasis and loiasis coendemic) East Region in Cameroon were tested for (1) Loa MFD in blood samples, (2) O. volvulus presence by SST, and (3) Immunoglobulin (Ig) G4 antibody positivity to Ov16 by rapid diagnostic test (RDT). A Classification and Regression Tree (CART) model was used to perform a supervised classification of SST status and identify a Loa MFD threshold above which it is highly likely to find L. loa mf in skin snips. Results Of 1011 Ov16-negative individuals, 28 (2.8%) tested SST positive and 150 (14.8%) were L. loa positive. The range of Loa MFD was 0–85 200 mf/mL. The CART model subdivided the sample into 2 Loa MFD classes with a discrimination threshold of 4080 (95% CI, 2180–12 240) mf/mL. The probability of being SST positive exceeded 27% when Loa MFD was >4080 mf/mL. Conclusions The probability of finding L. loa mf by SST increases significantly with Loa MFD. Skin-snip polymerase chain reaction would be useful when monitoring onchocerciasis prevalence by SST in onchocerciasis–loiasis coendemic areas.

Onchocerciasis is a parasitic infection caused by the filarial nematode Onchocerca volvulus and transmitted by Simulium blackfly vectors. Sessile adult female worms live inside nodules or worm bundles, where they mate with the (mobile) males and periodically produce thousands of embryos or microfilariae [1]. Microfilariae (mf) leave the nodules to populate the skin under the epidermis and the eyes and may live for a maximum of 2.5 years [1]. Most (99%) onchocerciasis cases occur in sub-Saharan Africa. The infection is associated with skin pathology, ocular lesions that progressively lead to blindness, excess mortality, and epilepsy [2][3][4][5]. The severity of these manifestations depends on the burden of infection, particularly on (past and/or present) O. volvulus microfilarial density (MFD) [6][7][8][9].
In endemic countries, regional control and elimination initiatives have been implemented, based on vector control and/or mass drug administration (MDA) with ivermectin. Elimination of transmission (EOT) has been achieved in formerly endemic Latin American countries (Mexico, Guatemala, Colombia, Ecuador) under the auspices of the Onchocerciasis Elimination Program for the Americas (1993-present) [10], with the exception of the Amazonian focus straddling Venezuela and Brazil [11]. In Africa, the Onchocerciasis Control Programme in West Africa (OCP;1974-2002 and the African Programme for Onchocerciasis Control (APOC; 1995-2015) led to substantial reductions in disease burden [12] and to EOT in some foci of West and East Africa [13,14]. Levels of precontrol endemicity and progress towards morbidity and EOT targets have been measured via assessment of prevalence and intensity of infection: in the OCP by detection and enumeration of skin mf by the skin snip technique (SST) [15] and in APOC by nodule prevalence at the beginning of the program and by SST for phase 1A (monitoring prevalence decline) and phase 1B (stop-MDA) evaluations [16]. The SST consists of taking 2 or more bloodless skin snips (typically with a Holth-type corneoscleral punch), incubating the snips in a suitable medium (eg, saline) for 24 hours, and detecting the presence (for prevalence) and number (for MFD) of emerged mf in the medium under a microscope [15].
However, the microfilarial morphology of Loa loa (another Chrysops-transmitted, filarial infection endemic in West and Central Africa) is superficially similar to that of O. volvulus (albeit L. loa mf are sheathed while those of O. volvulus are not) when preparations are not stained, making it difficult to distinguish between these 2 species using bright-field microscopy (L. loa mf measure 250-300 μm by 6-8 μm and those of O. volvulus measure 220-360 µm by 5-9 µm). Since L. loa mf have been considered to be solely blood-dwelling, the presence of L. loa mf in skin snips has seldom been investigated. A recent study demonstrated the occurrence of L. loa mf when using SST and discussed its implications for potential reporting of O. volvulus false positives [17]. Not only would this impact individual diagnosis but also epidemiological evaluations of interventions. Therefore, this study aimed to analyze the relationship between L. loa MFD in the blood (mf/mL) and the probability of detecting L. loa mf in skin snips during onchocerciasis diagnosis in an area where infections by O. volvulus and L. loa are coendemic.

Ethical Approval
The study received ethical clearance from the Cameroon National Ethics Committee for Research for Human Health (CNERSH; no. 2015/01/545/CE/CNERSH/SP). The survey was approved by and undertaken under the authority of the Ministry of Public Health of Cameroon following the Helsinki Declaration. Participation in this study was entirely voluntary and refusal to participate had no consequence for individuals. The protocol (objectives, methodology, use of collected data, and dissemination of results) was carefully explained to all eligible individuals. Those who agreed to participate signed an informed-consent form before undergoing clinical examination and sample collection. Parents or legal guardians provided their approval upon enrollment of minors (aged <21 years).

Study Area and Design
Data were collected during a study conducted in the East Region of Cameroon in March 2015 for integrated remapping of onchocerciasis, lymphatic filariasis, loiasis, malaria, and soiltransmitted helminthiases, which are coendemic in this region [18,19]. Since 2004, community-directed treatment with ivermectin (CDTI) for onchocerciasis has been implemented in 3 health districts (HDs)-namely, Bertoua, Betare Oya, and Lomie-out of the 14 HDs in the East Region.
For epidemiological monitoring, the East Region has been organized into 5 zones or evaluation units (EUs). A cross-sectional survey was conducted in a community selected in each of these EUs. Individuals of both sexes, aged 5 or more years, and residing in the selected community were eligible for the study and invited to participate. At least 150 individuals in each selected community were sampled.

Parasitological and Serological Assessment
Loa loa MFD (Loa MFD) was quantified using calibrated thick blood smears. A sample of 50 µL of peripheral blood was collected by finger-prick between 10 am and 2 pm (due to diurnal periodicity of L. loa mf [20]). After de-hemoglobinization and Giemsa staining, the slides were examined using bright-field microscopy. All mf present on the slide, including L. loa and Mansonella perstans, were identified, counted, and expressed as mf/mL. Here we only report Loa MFD.
Onchocerciasis was assessed by 2 different methods-namely, the SST to detect mf in the skin and a rapid diagnostic test (RDT)-to detect the presence of immunoglobulin G4 (IgG4) antibodies against the O. volvulus-specific antigen Ov16 [21]. For SST, 2 skin biopsies (one from each posterior iliac crest) were taken using a 2-mm Holth-type corneoscleral punch and incubated for 24 hours in saline; the presence of emerged mf was assessed under an optical microscope. Antibody testing was done using the Standard Diagnostics (SD) Bioline biplex Ov16/ Wb123 RDT as per the manufacturer's instructions, with specificity listed at 100%. Recently published studies using the biplex RDT [22] and studies using earlier prototypes have reported a specificity of 97.5% [23].

Statistical Modeling Analysis
The results for the categorical variables (sex, community, SST status, Ov16 RDT status) are presented as proportions. The results for the numerical variables (age, Loa MFD) are summarized by medians and their interquartile range (IQR). A logistic regression model was used to assess associations between SST false positivity and the following covariates: individuals' age and sex, community of residence, and Loa MFD. The covariate Loa MFD was introduced in the model as a fractional polynomial [24] to capture possible nonlinear associations. The choice of polynomial powers was based on minimizing the deviance information criterion (DIC) of the model. The Akaike information criterion (AIC) [25,26] was also used to select the best-fit model (that with the lowest AIC value). The predicted probability of SST false positivity was estimated using the function "predict" of the R software [27] using the best-fit regression model.
A Classification and Regression Tree (CART) model (based on an iterative algorithm) was used to estimate the covariates associated with SST false positivity. CART is a nonparametric multiple regression statistical model to perform supervised classifications with regard to explanatory variables compared with a categorical variable [28,29]. The modeling is carried out in 3 stages: (1) identification of the covariates associated with the response variable, (2) classification of each covariate to discriminate the response variable into 2 distinct groups, and (3) iterative repetition of the 2 previous steps until it is no longer possible to perform segmentation [30]. The 95% confidence interval (CI) for the identified threshold was obtained by bootstrapping [31], based on the percentile method with 10 000 replicates. In order to assess the difference in risk of being SST false positive between different Loa MFD classes, a logistic regression model was implemented to quantify, through the calculation of odds ratios (ORs) and their 95% CIs, the association between the probability of false positivity and covariates (age, sex, community, Loa MFD).
All analyses were performed with R version 3.6.1 using the packages MFP (multiple fractional polynomial) and ggplot2 [27,31]. The P value used for statistical significance was P < .05.

RESULTS
A total of 1053 (19.8%) individuals aged 6-85 (median, 28; IQR, 12-48) years participated in the study (out of 5314 individuals residing in the 5 study communities). The number of participants per community ranged between 156 and 318, with a sex ratio generally female biased ( Table 1).
The characteristics of the 28 SST false-positive individuals are shown in Supplementary Table 1. Although not statistically significant in the univariate analysis (P = .070), the number of false positives was approximately 3 times higher among females (21) than males (7). Likewise, the number of SST false positives was nearly double in individuals aged older than 40 years (18) compared with those aged 40 years or younger (10) (P < .001). The median Loa MFD was 5940 (IQR, 450-13 820) mf/mL among the SST positives and 0 (IQR, 0-0) among the SST negatives (P < .001). The proportions of SST false positives were similar between communities (P = .603) ( Table 2). Multivariate analysis (Supplementary Table 2) revealed that SST false positivity was significantly and negatively associated with male sex (P = .0226) and positively with Loa MFD (P < .0001).

Predictors and Predicted Probability of Being False Positive for Onchocerciasis
Several models were run to select the best one according to the AIC and DIC criteria. The logistic model with fractional polynomial (for Loa MFD) was the best-fit model (lowest AIC and DIC) (Supplementary Table 3). Since age and Loa MFD were positively and statistically significantly associated (P < .0001), age was not included in the multivariate analysis. Males were less likely than females to be SST false positive (P = .023). The probability of being SST false positive increased with Loa MFD (P < .001). Figure 1 illustrates the predicted probability of being SST false positive as a (nonlinear) function of Loa MFD according to sex. For both sexes, the probability of being SST false positive increases steeply for low Loa MFD values, with the rate of increase slowing down for high Loa MFD values. For a given Loa MFD, females were consistently more likely to be SST false positive than males. Supplementary Table 4 presents values of predicted probabilities of being SST false positive according to Loa MFD.  Figure 2). The distribution of the sexes among these 2 classes was similar (P = .183). Figure 3 shows the distribution of individual Loa MFD values for the 1011 Ov16-negative individuals according to the threshold of 4080 mf/mL and whether they tested SST negative or positive. Most SST positives were found in class B and exhibited high Loa MFD. Class B individuals were significantly more likely to be false SST positive than class A individuals (OR, 46.7; P < .001). Males were less likely to be false SST positive compared with females, regardless of Loa MFD category (OR, 0.3; P = .013) ( Table 3).

DISCUSSION
One of the challenges facing onchocerciasis control and elimination program managers and their partners is the availability of reliable diagnostic tools that can be used to assess infection trends during the treatment phase to be able to make stopping ivermectin MDA decisions and transition to post-treatment surveillance. The tests generally used in these contexts are the SST (to monitor progress during the treatment/intervention phase [16]), the Ov16 serology test (to measure exposure, particularly in children younger than 10 years), and the O. volvulus DNA (O-150 polymerase chain reaction [PCR] pool screen) test in blackflies (the latter 2 to confirm transmission interruption and stop MDA [32][33][34]). A major disadvantage of Ov16 serology is that it cannot differentiate between active and past O. volvulus infection. Also, it has been suggested that approximately 20% of   individuals in endemic communities may be unable to mount an IgG4 antibody response to Ov16 [35], resulting in a falsenegative outcome. In addition, there is a multiplicity of Ov16 assays, including various versions of the enzyme-linked immunosorbent assay and the RDT used here, which have different sensitivity and specificity characteristics, and this may vary between laboratories or between these and field settings [34].
In the case of the biplex RDT used here, we accepted the value of 100% specificity given by the manufacturers [34,36] but are aware that others have reported a lower (97.5%) specificity [22,23]. We assumed that when the RDT test was negative, an individual with an SST+/ve result would be a false positive for O. volvulus. Given that our overall proportion of false positives was 2.7%, our results should be interpreted with caution. The SST has been considered as the "gold standard" benchmark to detect and quantify O. volvulus microfilaridermia. However, its sensitivity decreases during control programs because O. volvulus MFD declines as a result of CDTI, and although SST sensitivity can be improved by increasing the number of snips taken [37], this may not be practical or acceptable. Besides, in areas of coendemicity with skin-dwelling Mansonella streptocerca (endemic to Africa, mf measuring 180-240 µm by 3-5 µm), or blood-dwelling M. perstans (200 µm by 4.5 µm), Mansonella ozzardi (coendemic with O. volvulus in the extant Amazonian focus, mf measuring 160-205 µm by 4 µm) [38], and L. loa [17], mf of these species can occur in skin snips and be misidentified as O. volvulus mf when assessing microfilaridermia by SST. Consequently, there is a risk that in areas where filarial diseases are coendemic, false O. volvulus positives can be detected, with repercussions for individual diagnosis and epidemiological evaluations. The aim of this work was, therefore, to determine the relationship between Loa MFD and the probability of a positive diagnosis by SST while being negative by Ov16.
Results indicate that the relationship between false positivity to SST and Loa MFD is nonlinear and statistically significant. The probability of being false positive to O. volvulus by SST increases sharply at low Loa MFD and less steeply at high Loa MFD values, being consistently higher for females compared with males-that is, for the same MFD, the probability is 2-3 times as high for females as for males (Supplementary Table 4). In the Nana-Djeunga et al [17] study, O. volvulus SST positivity was significantly higher in males (who might therefore be less likely to exhibit false positivity). Sex-specific exposure patterns to O. volvulus and L. loa are likely to be different [33,39].
The discrimination of the sample through a supervised classification model yielded statistical significance for Loa MFD only. Neither age or sex nor community of residence were significant predictors. The CART model produced 2 Loa MFD classes, the optimal discrimination value between the 2 being 4080 mf/mL, representing the threshold above which it is very (nearly 30%) likely to find L. loa mf in skin snips. This relatively low value reflected the distribution of Loa MFD in the individuals examined. Half of the sample had Loa MFD less than 6000 mf/mL, the median was 5940 mf/mL, and the 95% CI for the threshold was 2180-12 240 mf/mL. Above 12 240 mf/ mL, the probability of being SST false positive was greater than 35% for females and greater than 15% for males, reaching more than 60% for females and approximately 35% for males for Loa MFD = 80 000 mf/mL. This result is consistent with the previous study [17], in which the chances of being SST positive increased with Loa MFD.
The mechanisms explaining the presence of L. loa mf in the skin are yet to be investigated, but potential explanations  may include that L. loa mf are in the dermal microcapillaries when the skin biopsy is taken. Although skin snips are supposed to be bloodless, this may not always be achieved when using corneoscleral punches, sampling a large number of individuals, and/or having a number of technicians undertaking the procedure who may vary in their skill to perform the SST. Importantly, the association of this phenomenon with increasing microfilaremia levels is of interest, and similar to what has been described with O. volvulus mf, which may be found in blood when skin MFDs are high [40]. This phenomenon led to the description of Microfilaria bolivarensis in heavily infected Amerindian populations of the Amazonian focus [41], which was later demonstrated to be O. volvulus in blood [40]. Notwithstanding these considerations, it is increasingly recognized that the skin may be an organ of crucial importance in the transmission of (blood-dwelling) vector-borne parasites (eg, Trypanosoma brucei gambiense [42], Leishmania donovani [43], and Plasmodium falciparum [44]) and this phenomenon should be rigorously investigated in loiasis. We also recommend that further studies be carried out in different epidemiological settings and populations to assess the applicability and variability of the Loa MFD threshold identified in this study.

Conclusions
The diagnosis and epidemiological evaluations of onchocerciasis, particularly at baseline and during early phases of control interventions, rely on the detection and enumeration of O. volvulus mf by SST. In areas coendemic with loiasis, L. loa mf may also be found in the skin, emerging from dermal arterioles and venules when snips are taken with corneoscleral punches during SST, potentially confounding its results. In onchocerciasis-loiasis coendemic areas, a test-and-not-treat strategy has been developed to prevent post-ivermectin severe adverse events, enabling safe treatment of onchocerciasis. This strategy relies on LoaScope testing to measure Loa MFD and to exclude individuals with Loa MFD greater than 20 000 mf/mL from ivermectin treatment [45]. In these areas, testing such individuals for onchocerciasis and treating them with anti-Wolbachia therapies (eg, doxycycline) provides a safe and appropriate alternative treatment strategy to help eliminate onchocerciasis [46], as L. loa lacks Wolbachia endosymbionts [47]. This testand-treat with doxycycline strategy [48] would benefit from using methods such as real-time PCR and loop-mediated isothermal amplification (LAMP) assays for the amplification of O. volvulus DNA from skin snips [34]. A next step to investigate the impact of our results on onchocerciasis programs would be the inclusion, in a loiasis transmission dynamics model such as EPILOA [49], of the distribution of L. loa microfilarial loads to ascertain the proportion of the population who would likely test false-positive for O. volvulus in coendemic areas. The Loa MFD threshold (4080 mf/mL) above which the likelihood increases of having an SST false-positive result for onchocerciasis should be further investigated and validated. Should this phenomenon be confirmed, its consequences for disease manifestations and onward transmission would warrant further studies.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.