In coendemic areas, loiasis is a major obstacle in the control of onchocerciasis and lymphatic filariasis, two parasitic NTDs. Indeed, the occurrence of severe adverse events was recorded after IVM or DEC administration in populations with high Loa loa mocrofilaremia. Moreover, recent studies have shown an increased risk of mortality in individuals with filaremia greater than 30000 mf/mL, highlighting its interest in public health research [9, 27]. During a national campaign for loiasis cartography in Gabon, TBS-50 was used as the microfilaremic loiasis diagnostic. However, to the best of our knowledge, the performance characteristics of this technique were not clearly described in the literature, as well as those of blood direct examination of 10 µL and leukoconcentration analysis of 5 mL of blood – two other commonly used diagnostic techniques. The present study aims to fill this research gap.
In our study, the percentage of carriers of L. loa microfilariae identified with the leukoconcentration technique was 12.6% in the classical analysis and 9.7% in the BLCA. Previous studies performed in the same laboratory with an identical diagnostic tool found prevalences of 8.2% and 39.5% [1, 25]. The place of residence of the populations in these studies may partly explain the differences in prevalence observed. Indeed, people who live in endemic areas, mainly in rural settlements, are more exposed to Chrysops (loiasis vector) bites than their counterparts living in urban areas [25, 28, 29]. In the survey performed by Bouyou-Akotet et al, for example, two-thirds of the population came from rural areas, which could explain the higher prevalence observed in that study [1]. Microfilaremia levels identified via direct blood examination were slightly higher than those obtained via the TBS-50 technique, but the differences were not statistically significant, with a p value of 0.97. Blood direct examination is performed with the fresh blood of patients and allows observation of living Loa loa microfilariae mobility. On the other hand, the microfilariae can cross the slide and may be counted more than once in contrast to TBS-50, which fixes worms on the preparation. In the current study, three TBS-50 slides could only be partially analyzed, as during the staining process, the biological material was removed. This may happen with venous blood collection in anticoagulant tubes, contrary to capillary collection, where better adherence of the blood on the slide is observed. In addition, a recent study conducted in Lambaréné and surrounding villages, located in Gabon, found a microfilaremia detection risk 1.24 higher from capillary blood than from venous blood [30]. This may influence the diagnostic accuracy of the TBS-50 and could be considered one of the limitations of the evaluation of the thick smear method in the current survey.
In our study, we first assessed the performance of TBS-50 and direct examination compared to leukoconcentration, assuming the latter is a perfect reference technique. Although TBS-50 requires 5 times more blood than direct blood examination, there were no discordant results between the two techniques. It follows that using leukoconcentration as a perfect reference technique, the diagnostic accuracies of the TBS-50 and blood direct examination were identical, with a sensitivity of 75.0% and a specificity of 100.0%. The positive predictive value was 100.0%, and the negative predictive value was 96.5%. In the literature, only one other study had a similar design [22]. In this study, the researchers used two slides of 20 µL each (40 µL in total) in comparison with saponin hemolysis, which requires a volume of 5 mL of blood, similar to the leukoconcentration technique [22]. Whereas the specificity and the positive predictive value were the same as in our study (100.0%), the sensitivity and negative predictive value were slightly lower in that article (67.0% and 93.0%, respectively). It is possible that the additional collection of 10 µL for TBS-50 in the current study led to increasing sensitivity and negative predictive value. On the other hand, it is well known that a high prevalence is associated with a decrease in the negative predictive value [31]. The prevalence found by Apembe and Noireau was 1.5/2-fold higher than that in our study (19.4% versus 12.6/9.7%).
A perfect diagnostic test can ideally discriminate subjects with and without a disease. However, such tests with performance characteristics at 100.0% do not exist [32]. With a blood volume less than 5 mL in a study performed in Franceville in Gabon, the leukoconcentration technique showed false negative results in L. loa amicrofilaremic individuals when compared with the results of a nested polymerase chain reaction (PCR) assay [33]. Another study showed a difference of almost 50.0% between a concentration technique and PCR for the detection of cases [34]. This imperfection translates into biased estimates when using the classical formula. To overcome this limitation, we applied a Bayesian latent class analysis approach that can be used in the presence of imperfect reference tests [23, 35, 36]. This statistical method has been increasingly used in recent years; however, it has not yet been applied in Gabon in the context of diagnostic testing. Bayesian analysis allows us to estimate both diagnostic performance characteristics of the reference imperfect test and the index test, as well as to estimate the proportion of L. loa microfilariae carriers [37, 38].
In the main model including prior information from the literature, the BLCA analysis showed a higher, nearly perfect sensitivity for the leukoconcentration technique compared to TBS-50 and blood direct examination. However, the leukoconcentration had a false positive rate of 8.6% (= 100.0-91.4 for the specificity). This is a low but nonzero rate, with less than one in ten cases that are falsely declared positive. Indeed, the preparation observation in monochrome white color settings could easily lead to confusion of immobile microfilariae with nonparasitic elements. A potential option to reduce the rate of false positive results is the addition of a staining step to the leucocyte pellet. This will decrease errors by microscopists. However, in the literature, the possibility of an additional staining step was not mentioned. In Congo settings, the observed rate of false positives was higher (33.3%) with two TBS-20 [22].
Regarding the TBS-50 and blood direct examination, there was a 21.1% (= 100 − 78.9% for the sensitivity estimate) rate of false negative results, i.e., More than one-fifth of the screened population who test negative has the disease. This is a high rate and could be considered the main limitation of these diagnostic tests. Nevertheless, the other accuracy parameters are all above 90.0%, with a perfect specificity of 100%.
To assess the robustness of our results, we conducted sensitivity analyses using noninformative priors for prevalence and test accuracies. The results were similar when only priors for test accuracies were assumed to be noninformative or the prevalence prior. At first sight, it seems that the prior associated with the prevalence does not influence the model. However, we could not confirm this hypothesis because when removing the prior associated with the prevalence while keeping those associated with the TBS-50 and leukoconcentration, the model did not converge due to incompatibility between the priors and the data. Sensitivity estimates increased for TBS-50 and direct examination when using noninformative priors (alternative models) compared to the scenario when informative priors were used (main model). However, the results comparison should be conducted with caution, as credible intervals largely overlap between all three models. It has been observed that with the use of noninformative priors, the credible intervals for negative and positive predictive values are very wide (with the point estimates at approximately 50%), which reflects high uncertainty in the estimates of NPVs and PPVs that has not been the case in the model with all priors being informative. One of the reasons for this is that the absence of prior information noticeably decreases the ability of the model to estimate all of the diagnostic characteristics precisely, especially taking into account the relatively small sample size of the study itself. On the other hand, certainty in all of the estimates achieved in the main model, where all priors have been informative, was also expected, as more informative priors are, more certain the estimates drawn from posterior are.
On the other hand, certainty in all of the estimates achieved in the main model, where all priors have been informative, was also expected, as more informative priors are, more certain the estimates drawn from posterior are.
In the current study, we also observed that sensitivities were higher in a subgroup with hypereosinophilia and lower in those with normal eosinophilia, regardless of the diagnostic test considered. Eosinophils are involved in the immune response against helminth pathogens. In Low Income and Developing Countries of Central and West Africa, eosinophilia is often found to be significantly associated with helminth infection and with loiasis, such as in Lambaréné, in Gabon [39]. Eosinophils secrete immunoglobulin E, which is implicated in the allergic reaction: a positive correlation was observed during L. loa infection [24]. Eosinophils can potentially be a good marker for microfilaremic loiasis in Gabon in high prevalence settings. However, Mansonella perstans and intestinal parasites were not researched in the study population, which limited the development of a real picture of loiasis and eosinophil production. They also influence the human immune system toward the T-reg/Th2 polarized response [24, 40].
The deployment of a test depends on both the sensitivity and specificity. The TBS-50 is used in Gabon in mass surveys but generally not for diagnostic testing. One of the qualities recommended for a screening test when a trade-off has to be made is good sensitivity at the detriment of specificity. TBS-50 has a sensitivity of approximately 80.0% and is currently the best tool that can be easily implemented with limited resources. In addition to the previously mentioned criteria, a screening test should be well accepted by the populations [41]. In our study, TBS-50 was not tested in communities but rather in populations coming for medical consultations in the hospital environment. Future research should investigate the acceptability of the TBS-50 among the population in Gabon.
There are other diagnostic tools, such as LoaScope, that have 100.0% sensitivity and 94.0% specificity when compared to a TBS [42]. This test was developed to quickly diagnose (~ 2 minutes) hypermicrofilaremic loiasis in rural areas during the Mass Drug Administration with ivermectin, where no technical platform is available for the realization and reading of the TBS. Patients with low microfilaremia or without microfilaremia can be safely administered ivermectin/DEC without the occurrence of severe adverse events. However, this technology has limited availability [43]. As an alternative, blood collection on a microscope slide could be performed in the field, and staining and reading of the TBS-50 would be carried out later in the laboratory. The diagnostic indicators of LoaScope were obtained in comparison with a thick blood smear of the same volume, namely, 2.59 µL [42]. It would be interesting to compare this point-of care with a technique using a high volume of 10 µL, 50 µL or 5 mL of blood. Finally, the study by D’Ambrosio et al. was conducted in an endemic rural area and recruited a different population than our study, which can explain the difference observed.
The LAMP (loop-mediated isothermal amplification) technique is increasingly being developed and evaluated for the rapid molecular diagnosis of loiasis. This technique has shown a sensitivity of approximately 90.0% when compared to microscopy and PCR-based assays [44]. It can provide information both on the presence and the absence of microfilariae but also on the microfilaremia without the need to use microscopy, with colored reaction [45, 46]. In the future, these molecular tools should be developed for diagnosis in the field to be applicable in rural areas. Molecular tools are known to be more sensitive and would allow the detection of all cases of microfilaremic loiasis [33]. Indeed, patients with loiasis-related clinical symptoms have low microfilaremia, often not detectable with parasitological techniques [1]. As such, it would be interesting to assess the suitcase mobile laboratory for loiasis diagnosis used during the COVID-19 pandemic [45–47]. The present results have to be interpreted in their context, and test accuracies can be extrapolated to rural areas, in which the prevalence of microfilaremic loiasis is much higher. An implementation phase of this technique in these areas is now necessary to allow rural populations to benefit. Future research should assess the acceptability of TBS-50 in rural areas in Gabon.