Adherence to Hygiene Protocols and Doxycycline Therapy in Ameliorating Lymphatic Filariasis Morbidity in an Endemic Area Post-Interruption of Disease Transmission in Ghana

Abstract

Filarial lymphedema (LE) remains a significant global problem despite the progress made toward elimination of lymphatic filariasis (LF). In Ghana, the main approach to LF is preventive chemotherapy, but this has minimal impact on individuals who have already developed LE. In 2018-2020, a 24-month randomized, double-blind, placebo-controlled trial was conducted to evaluate the efficacy of stringent hygiene measures using the Essential Package of Care with or without additional administration of doxycycline (DOX) to improve filarial leg LE. This study enrolled 356 participants with LE stages 1-3 from two districts in the Upper East Region of Ghana. In addition to regular training on appropriate care for their affected legs, participants were randomized to receive 6 weeks of either 200 mg/day DOX (n = 117), 100 mg/day DOX (n = 120), or matching placebo (n = 119). Participants were seen every 2 months, with clinical measurements done at 6, 12, 18, and 24 months to assess the status of affected legs. There was a trend toward later appearance of acute attacks after DOX, but surprisingly, DOX showed no effect on LE stage progression. In all groups, leg LE improvement was more common (DOX 200 mg: n = 23 [20%]; DOX 100 mg: n = 23 [19.5%]; placebo: n = 32 [27.4%]) than LE worsening (DOX 200 mg: n = 2 [1.7%]; DOX 100 mg: n = 3 [2.5%]; placebo: n = 2 [1.7%]). Overall, these data show a strong benefit from adherence to a strict hygiene protocol, with some added potential benefit for DOX in preventing acute attacks.

Figure 1.

(A) Flowchart – recruitment and randomization. From the 420 eligible participants, 356 were randomized in group A (LE stages 1–3) and 58 in group B (LE stages 4–6). Five participants were incorrectly randomized. One participant belonged to group A but was randomized in group B, and four participants were randomized in group A but belonged to group B. This became apparent only after treatment was carried out. Another participant started with diuretic treatment right after trial treatment. This resulted in improvement from stage 1 to stage 0. It was decided in consultation with the Data Safety and Monitoring Board before unblinding of the study to exclude the data of these six participants from all but the safety analyses. (B) Flowchart (see next page) – treatment allocation. The number of participants in group A were randomized to one of the three treatment groups; their presence/absence during the follow-up visits as well as their belonging to the ITT and/or PP analysis sets for the respective time points and the reason for their absence or exclusion from PP analysis are shown. ITT = intention-to-treat; PP = per-protocol.

Figure 1.

(A) Flowchart – recruitment and randomization. From the 420 eligible participants, 356 were randomized in group A (LE stages 1–3) and 58 in group B (LE stages 4–6). Five participants were incorrectly randomized. One participant belonged to group A but was randomized in group B, and four participants were randomized in group A but belonged to group B. This became apparent only after treatment was carried out. Another participant started with diuretic treatment right after trial treatment. This resulted in improvement from stage 1 to stage 0. It was decided in consultation with the Data Safety and Monitoring Board before unblinding of the study to exclude the data of these six participants from all but the safety analyses. (B) Flowchart (see next page) – treatment allocation. The number of participants in group A were randomized to one of the three treatment groups; their presence/absence during the follow-up visits as well as their belonging to the ITT and/or PP analysis sets for the respective time points and the reason for their absence or exclusion from PP analysis are shown. ITT = intention-to-treat; PP = per-protocol.

Figure 2.

Sankey diagram of stage changes over time. The Sankey diagram shows changes in the LE stages in the ITT collective over the study period of 2 years for all participants together (left graph) and divided by the three different treatments (right graph). BSL = baseline; DOX = doxycycline; ITT = intention-to-treat; LE = lymphedema; LTFU = lost-to-follow-up.

Figure 3.

Stage progression and improvement. (A) The number and percentage of participants (ITT collective) who showed LE stage progression at the different follow-up time points divided by the treatment groups. (B) The number and percentage of participants (ITT collective) for the parameter LE stage improvement. (C) Forrest plot (see next page) – multivariable analysis for stage improvement over time. The Forest plot shows the variables that were included in a multivariable logistic regression model (PROC GENMOD) for the time-dependent outcome variable “improvement.” The following covariables were chosen for the model: sex (male/female), age, weight, years in endemic area, treatment (DOX 200/DOX 100/placebo), and region (Kassena Nankana East and Kassena Nankana West) as well as the following time-dependent covariables (taking changes during the follow-up period into account): LE staging (stage 1, 2 or 3), FTS positivity (active LF infection), hygiene status (limb not clean/limb clean), and ADL attack during the previous 6 months (no/yes). Odds ratios (ORs) with 95% CIs are given for each covariable. ADL = adenolymphangitis; DOX = doxycycline; FTS = filariasis test strip; ITT = intention-to-treat; LE = lymphedema; LF = lymphatic filariasis; PROC GENMOD = Procedure for Generalized Linear Models.

Figure 3.

Stage progression and improvement. (A) The number and percentage of participants (ITT collective) who showed LE stage progression at the different follow-up time points divided by the treatment groups. (B) The number and percentage of participants (ITT collective) for the parameter LE stage improvement. (C) Forrest plot (see next page) – multivariable analysis for stage improvement over time. The Forest plot shows the variables that were included in a multivariable logistic regression model (PROC GENMOD) for the time-dependent outcome variable “improvement.” The following covariables were chosen for the model: sex (male/female), age, weight, years in endemic area, treatment (DOX 200/DOX 100/placebo), and region (Kassena Nankana East and Kassena Nankana West) as well as the following time-dependent covariables (taking changes during the follow-up period into account): LE staging (stage 1, 2 or 3), FTS positivity (active LF infection), hygiene status (limb not clean/limb clean), and ADL attack during the previous 6 months (no/yes). Odds ratios (ORs) with 95% CIs are given for each covariable. ADL = adenolymphangitis; DOX = doxycycline; FTS = filariasis test strip; ITT = intention-to-treat; LE = lymphedema; LF = lymphatic filariasis; PROC GENMOD = Procedure for Generalized Linear Models.

Figure 4.

(A) Hygiene status – limb washed and clean. Hygiene training was carried out at baseline and at 4, 6, 12, 18, and 24 months. At the same point, participants were assessed for the hygiene of the legs. In this graph, the overall assessment “limb was washed and clean” is shown for all time points. The dotted lines indicate that not all participants were present at all time points. (B) Forrest plot – multivariable analysis for hygiene status over time. The Forest plot shows the variables that were included in a multivariable logistic regression model (PROC GENMOD) for the time-dependent outcome variable “limb washed and clean,” also referred to as “hygiene status.” The following covariables were chosen for the model: sex (male/female), age, weight, years in endemic area, treatment (DOX 200/DOX 100/placebo), other leg affected, and region (Kassena Nankana East and Kassena Nankana West), as well as the following time-dependent covariables (taking changes during the follow-up period into account): LE staging (stage 1, 2, or 3), ADL attack during the previous 6 months (no/yes), FTS positivity (active LF infection), and season at time of assessment (rainy or dry season). Odds ratios (ORs) with 95% CIs are given for each covariable. ADL = adenolymphangitis; BSL = baseline; DOX = doxycycline; FTS = filariasis test strip; LE = lymphedema; LF = lymphatic filariasis; PROC GENMOD = Procedure for Generalized Linear Models.

Figure 5.

(A) ADL – time to first attack after treatment start. The Kaplan-Meyer curve shows the time to the occurrence of the first ADL attack after treatment onset divided by the different treatment groups. The time until 50% of the participants experienced a new attack after treatment onset was 14 months for the placebo group, 16 months for the DOX 100 group, and 18 months for the DOX 200 group. (B) ADL – count model – multivariable analysis on ADL attacks over time. A Cox model formulated in terms of increments in the number of events along the time line was used following an approach by Anderson and Gil^, to take not only the first occurrence of an ADL attack into account but also the fact that multiple episodes could have occurred during the study period of 2 years. The Forest plot shows the effects for the following covariables that were chosen for this model as hazard ratios (HRs) with 95% CIs: sex (male/female), age, weight, years in endemic area, treatment (DOX 200/DOX 100/placebo), other leg affected, and region (Kassena Nankana East and Kassena Nankana West), as well as the following time-dependent covariables (taking changes during the follow-up period into account): LE staging (stage 1, 2, or 3), FTS positivity (active LF infection), hygiene status (limb not clean/limb clean), and season at time of assessment (rainy or dry season). ADL = adenolymphangitis; DOX = doxycycline; FTS = filariasis test strip; LE = lymphedema; LF = lymphatic filariasis.

Figure 6.

(A) QoL – WHODAS 2.0. The mean ± SEM of the WHODAS 2.0 score at baseline and 12 and 24 months. The dotted lines indicate that not all participants were present at all time points. (B) QoL – multivariable analysis for WHODAS 2.0 changes over time. To analyze the influence of covariables on the WHODAS 2.0 score, a linear mixed-effects model was used. The Forest plot shows the covariables that were used in the linear mixed-effects model, which were the following baseline covariables: sex (male/female), age, weight, years in endemic area, treatment (DOX 200/DOX 100/placebo), district (Kassena Nankana East and Kassena Nankana West), and if the other leg was also affected, as well as the following time-dependent covariables (taking changes during the follow-up period into account): LE staging (stage 1 or 2/stage 3) as well as LE change (no change/progression/improvement), FTS positivity (active LF infection, hygiene status [limb not clean/limb clean], and ADL attack during the previous 6 months (no/yes). The effects are presented with the estimate (β) and the 95% CIs. ADL = adenolymphangitis; BSL = baseline; FTS = filariasis test strip; LE = lymphedema; LF = lymphatic filariasis; QoL = quality of life; SEM = standard error of the mean; WHODAS = WHO Disability Assessment Schedule.