Comparison between Quantitative Polymerase Chain Reaction and Sodium Nitrate Flotation Microscopy in Diagnosing Soil-Transmitted Helminth Infections

ABSTRACT.

There is evolving interest in alternate microscopy techniques and quantitative polymerase chain reaction (qPCR) to evaluate soil-transmitted helminth (STH) burden. Using data from a cross-sectional survey of 540 schoolchildren across six primary schools in three municipalities of Timor-Leste, we compared the performance of microscopy using sodium nitrate flotation (SNF) and qPCR in determining STH prevalence and infection intensity. Prevalence by qPCR was higher than SNF for Ascaris lumbricoides (17.5% versus 11.2%), hookworm (8.3% versus 1.2%), and Trichuris trichiura (4.7% versus 1.6%). Agreement between SNF and qPCR was fair for hookworm (κ = 0.21) and moderate for A. lumbricoides (κ = 0.59) and T. trichiura (κ = 0.44). Moderate or heavy intensity infections were identified in 15.9% of infections detected by SNF, whereas qPCR identified 36.1% as moderate or heavy infections using cycle threshold to eggs per gram conversion formulas. Quantitative PCR is a promising diagnostic technique, though further studies validating infection intensity correlates are required.

INTRODUCTION

Soil-transmitted helminth (STH) infections remain a major global health issue, affecting almost one billion people globally¹ and contributing the highest burden of all the neglected tropical diseases.² Integral to STH control is preventive chemotherapy targeting school-age children in settings where STH prevalence is 20% or higher, according to World Health Organization (WHO) recommendations.³ To evaluate STH burden (prevalence and infection intensity), accurate diagnostic techniques suitable in resource-limited settings are necessary. Interest is evolving in alternate diagnostic techniques to Kato-Katz, which is the WHO standard method for STH diagnosis and quantification,⁴ including sodium nitrate flotation (SNF), which demonstrates superior sensitivity than Kato-Katz in detecting hookworm eggs and can be done in preserved stool samples, thereby simplifying fieldwork procedures.⁵ However, there is concern for the poor sensitivity of microscopy in low-prevalence settings and detecting light intensity infections and the inability to differentiate between hookworm species.^5,6 Quantitative polymerase chain reaction (qPCR) is a molecular method gaining recognition in STH diagnostics, demonstrating higher sensitivity than microscopy across a range of settings and capacity to specify hookworm species.^5,7–11 This is particularly advantageous following STH control programs where prevalence and infection intensity is expected to be lower¹² and to evaluate the burden of zoonotic hookworm species to inform broader control strategies involving animal reservoirs.¹³ Furthermore, qPCR can be conducted on preserved stool samples. However, although qPCR quantifies helminth DNA, correlation with the traditional estimates of infection intensity obtained by microscopy (eggs per gram [epg]) is not well established, and studies have demonstrated linear correlations between qPCR cycle threshold (Ct) values and microscopy epg (having either one or both values log-transformed) using fecal seeding experiments and field samples.^7,8,14,15 However, direct comparisons between SNF and qPCR infection intensity using Ct to epg conversion formulas are yet to be published. This analysis is the first to compare SNF and qPCR in both determining STH prevalence and infection intensity (in epg) using samples collected in a cross-sectional study in Timor-Leste.

MATERIALS AND METHODS

A cross-sectional survey of school-age children across six primary schools in three municipalities of Timor-Leste was conducted in April–May 2019. The study received Human Research Ethics Committee approval from the University of New South Wales (HC190140) and Timor-Leste Ministry of Health (1545MS-INS/DE/X/2019).

Upon collection of each stool sample, two aliquots (3 g each) were placed into separate centrifuge tubes and preserved with 3 mL of 5% potassium dichromate, kept on ice in the field, and refrigerated upon arrival at the National Health Laboratory, Dili, Timor-Leste. A modified SNF analysis of samples (using two coverslip readings, after being atop the centrifuge tube for 5 minutes each) was conducted at the National Health Laboratory (May 2019) following a 2-day training by a senior parasitologist (RT).⁵ For qPCR analysis, samples were couriered to the University of Melbourne, Australia, and underwent two quadraplex real-time qPCR assays. The first assay was performed to enumerate Ascaris lumbricoides and Trichuris trichiura and detect Strongyloides stercoralis using Equine Herpesvirus-4 (as an internal qPCR control), and the second was performed to enumerate Necator americanus, Ancylostoma duodenale, and Ancylostoma ceylanicum using human 16S mitochondrial rRNA as an internal qPCR and DNA extraction control.¹⁵ Nuclease-free water was used as negative control. The sequences of primers and probes and PCR conditions were based on published information^8,14–20 and unpublished information from P. A. Zendejas-Heredia et al. (Supplemental Table 1). Quantitative PCR assays were performed from December 2019 to January 2020. Strongyloides stercoralis is not included in this analysis because it is not detected by SNF.

The prevalence and 95% confidence intervals for A. lumbricoides, hookworm, T. trichiura, and infection with at least one STH was compared using SNF and qPCR. Identifying specific hookworm species is not possible by microscopy; therefore, an individual was deemed positive for hookworm by SNF if one or more hookworm species was detected. Diagnostic agreement between SNF and qPCR was assessed using Cohen’s Kappa agreement statistics (very good, κ > 0.8; good, 0.6 < κ ≤ 0.8; moderate, 0.4 < κ ≤ 0.6; fair, 0.2 < κ ≤ 0.4; poor, κ ≤ 0.2). In the absence of a true gold standard, a “true positive” was considered an infection detected by either method,⁷ and the sensitivities of SNF and qPCR were estimated by dividing the positives for each method by the “true positive.” Specificity was assumed to be 100%. Infection intensity by SNF was determined by epg and categorized using WHO criteria.²¹ Infection intensity by qPCR was determined by correlating Ct values with epg through fecal seeding experiments^8,14 and subsequently categorized using WHO criteria.²¹ Samples seeded with known quantities of STH eggs were analyzed by qPCR to develop the following linear regression models: A. lumbricoides Ct = −3.489log₁₀(epg) + 36.97, R² = 0.9881; N. americanus Ct = −3.641log₁₀(epg) + 35.02, R² = 0.982; and T. trichiura Ct = −3.288log₁₀(epg) + 36.73, R² = 0.9812.¹⁵ The Spearman’s rank correlation coefficient was calculated to assess associations between SNF and qPCR in determining infection intensity (very strong, ρ ≥ 0.8; strong, 0.6 ≤ ρ < 0.8; moderate, 0.4 ≤ ρ < 0.6; weak, 0.2 ≤ ρ < 0.4; very weak, ρ < 0.2). A P value < 0.05 was considered statistically significant. For participants with multiple SNF or qPCR results (e.g., multiple samples collected, multiple analyses on same sample, or sample mislabeling), if the results were concordant then one result was included; if the results were discordant then the SNF or qPCR results were excluded. Among the 16 participants with multiple SNF results, 10 had concordant results that had one result included. Among the 10 participants with multiple qPCR results, eight had concordant results that had one result included. All analyses were conducted using SAS version 9.4 (SAS, Cary, NC).

RESULTS

Samples were collected from 548 participants. After removing duplicate samples, samples from 540 participants were included in this analysis. Table 1 compares the prevalence of A. lumbricoides, hookworm, T. trichiura, and infection with any STH by SNF and qPCR. Prevalence was lower for SNF compared with qPCR for all species: A. lumbricoides, 11.2% by SNF versus 17.5% by qPCR; hookworm, 1.2% by SNF versus 8.3% by qPCR; T. trichiura, 1.6% by SNF versus 4.7% by qPCR; and any STH, 13.6% by SNF versus 27.5% by qPCR. The qPCR allowed the identification of hookworm species, with N. americanus being most prevalent (7.5%; 95% confidence interval: 5.3–9.8).

Table 1.

Species-specific soil-transmitted helminth prevalence as determined by sodium nitrate flotation and quantitative polymerase chain reaction

Organism	Prevalence by SNF (N = 493)		Prevalence by qPCR (N = 531)		Prevalence by SNF or qPCR (N = 540)
	n	% (95% CI)	n	% (95% CI)	n	% (95% CI)
Ascaris lumbricoides	55	11.2 (8.4–13.9)	93	17.5 (14.3–20.8)	104	19.3 (15.9–22.6)
Hookworm*	6	1.2 (0.3–2.2)	44	8.3 (5.6–10.6)	45	8.3 (6.0–10.7)
Necator americanus	–		40	7.5 (5.3–9.8)	–
Ancylostoma duodenale	–		1	0.2 (0–0.6)	–
Ancylostoma ceylanicum	–		3	0.6 (0–1.2)	–
Trichuris trichiura	8	1.6 (0.5–2.7)	25	4.7 (2.9–6.5)	26	4.8 (3.0–6.6)
Any STH	67	13.6 (10.6–16.6)	146	27.5 (23.7–31.3)	156	28.9 (25.1–32.7)

CI = confidence interval; qPCR = quantitative polymerase chain reaction; SNF = sodium nitrate flotation.

^*Hookworm species cannot be determined by microscopy but can be distinguished by qPCR; therefore, only hookworm prevalence is reported for SNF, and the prevalence of specific hookworm species are reported for qPCR.

Sensitivity and diagnostic agreement were calculated for the 483 participants with SNF and qPCR results. For A. lumbricoides, SNF sensitivity was 58.1% (54/93), and qPCR was 89.2% (83/93); for hookworm, SNF sensitivity was 12.8% (5/39), and qPCR was 100% (39/39); for T. trichiura, SNF sensitivity was 30.4% (7/23), and qPCR was 95.8% (23/24). The diagnostic agreement between SNF and qPCR was fair for hookworm (κ = 0.21; P < 0.001) and moderate for A. lumbricoides (κ = 0.59, P < 0.001) and T. trichiura (κ = 0.44; P < 0.001) (Table 2).

Table 2.

Diagnostic agreement between sodium nitrate flotation and quantitative polymerase chain reaction

	SNF result	qPCR detected (n)	qPCR not detected (n)	Agreement (%)	Kappa statistic*	P value
Ascaris lumbricoides	Detected	44	10	428 (89.7)	0.59	< 0.001
	Not detected	39	384
Hookworm	Detected	5	0	444 (92.9)	0.21	< 0.001
	Not detected	34	439
Trichuris trichiura	Detected	7	1	461 (96.4)	0.44	< 0.001
	Not detected	16	454

qPCR = quantitative polymerase chain reaction; SNF = sodium nitrate flotation.

^*Kappa agreement classification: < 0.20 = poor; 0.21–0.40 = fair; 0.41–0.60 = moderate; 0.61–0.80 = good; 0.81–1.00 = very good.

Table 3 shows infection intensity and mean epg values by SNF and qPCR (using Ct to epg conversion formulas). For A. lumbrocoides the mean epg by SNF was 2,778 (range: 2–21,654), and by qPCR it was 57,521 (range: 6–1,707,831); for hookworm the mean epg by SNF was 294 (range: 10–1,294). For N. americanus the mean epg by qPCR was 5,898 (range: 3–217,680); for T. trichiura the mean epg by SNF was 81 (range: 2–411) and by qPCR was 282,741 (range: 1–7,065,590). All hookworm and T. trichiura infections were light intensity by SNF, compared with 2/40 (5%) of N. americanus infections and 2/25 (8%) of T. trichiura infections being moderate or heavy intensity by qPCR. Only 11/55 (20%) of A. lumbricoides infections were moderate or heavy intensity by SNF, compared with 53/93 (57.0%) by qPCR. Overall, moderate or heavy infection intensity was identified by SNF in 11/69 (15.9%) cases, whereas 57/158 (36.1%) were moderate or heavy infection intensity by qPCR. The correlation between SNF and qPCR for A. lumbricoides infection intensity was not significant (ρ = 0.20; P = 0.20) and could not be calculated for N. americanus or T. trichiura due to the low number detected by SNF.

Table 3.

Infection intensity as determined by sodium nitrate flotation and quantitative polymerase chain reaction*

Organism	Light, n (%)		Moderate, n (%)		Heavy, n (%)		Mean epg (range)
	SNF (N = 493)	qPCR (N = 531)	SNF (N = 493)	qPCR (N = 531)	SNF (N = 493)	qPCR (N = 531)	SNF	qPCR
Ascaris lumbricoides	44 (8.9)	40 (7.5)	11 (2.2)	41 (7.7)	0	12 (2.3)	2,778 (2–21,654)	57,521 (6–1,707,831)
Hookworm	6 (1.2)	–	0	–	0	–	294 (10–1,294)	–
Necator americanus	–	38 (7.2)	–	1 (0.2)	–	1 (0.2)	–	5,898 (3–217,680)
Trichuris trichiura	8 (1.6)	23 (4.3)	0	1 (0.2)	0	1 (0.2)	81 (2–411)	282,741 (1–7,065,590)

epg = eggs per gram; qPCR = quantitative polymerase chain reaction; SNF = sodium nitrate flotation.

^*Using cycle threshold to epg conversion formulas.

CONCLUSION

This analysis adds to the growing evidence for the value of qPCR in STH diagnostics. Our results demonstrating higher sensitivity of qPCR compared with SNF are consistent with previous analyses.^5,7,8,14,15 The lower sensitivity of SNF and diagnostic agreement between SNF and qPCR in our study compared with a previous analysis from Timor-Leste⁷ may reflect issues with egg recovery using SNF and the operator-dependent nature of microscopy, with the prior study using experienced researchers and this study using routine laboratory technicians. The small number of STH infections detected by SNF, particularly for hookworm and T. trichiura, is one of the main limitations of our study and is likely to have contributed to the poorer diagnostic agreement in our analysis. The small number of STH infections detected by SNF but not qPCR may reflect the heterogeneity of egg distribution within stool samples, false-positive microscopy results due to the presence of eggs from other morphologically similar nematodes,^22,23 or the misidentification of fecal material.

Only 15.9% of STH infections detected by SNF were moderate or heavy intensity, whereas 36.1% of infections detected by qPCR were moderate or heavy intensity. Importantly, we did not observe egg embryonation in qPCR samples, which would increase DNA amount beyond that at the time of infection. However, individual sample variations in egg embryonation cannot be excluded if defecation occurred the day prior to sample processing, which would result in falsely elevated qPCR-based egg intensity estimates. Quantitative PCR may also overestimate STH burden by detecting DNA from nonviable eggs and free DNA. There have been difficulties in comparing infection intensity obtained by microscopy, where egg counts are measured directly, and qPCR, where egg counts are estimated based on qPCR Ct values and calibration curves. Previous attempts to derive infection intensity using Ct to epg conversion formulas consistently demonstrate higher infection intensities compared with microscopy, with Hii et al.¹⁴ reporting hookworm egg intensity estimates 4- to 8.5-fold higher using Ct conversion formulas compared with field-based microscopy and increasing discrepancy as egg intensities lighten. Additionally, STH egg-seeding experiments demonstrated egg recovery rates of microscopy-based techniques (Kato Katz and SNF) to be 20% lower compared with quadraplex qPCR assays, with qPCR being better able to predict the accuracy of light or moderate egg intensity cut-offs compared with microscopy-based techniques.¹⁵ Our study is the first to use Ct to epg conversion formulas to directly compare SNF and qPCR in determining infection intensity using WHO epg thresholds. Additional studies are needed to evolve the application of molecular quantification to determine STH infection intensity.

In conclusion, this analysis further demonstrates the increased sensitivity of qPCR in detecting and quantifying STH infection compared with microscopy. The diagnostic performance of qPCR, in conjunction with straightforward sample preservation, lower potential for interoperator variability, and capacity for species-specific identification of hookworm makes for a promising diagnostic. Where accessible, qPCR should be incorporated into assessments of STH burden to refine quantitative correlations with microscopy techniques and establish validated conversion of qPCR Ct values to epg. This will allow a more accurate assessment of STH burden and evaluation of progress toward STH control and potential elimination.

Note: Supplemental table appears at www.ajtmh.org.