Questionnaire and examination of net
The study began 1 year after the 2011 mass distribution, so LLINs selected were already 12 months of age. At enrolment, these LLINs were marked with an identification number with permanent black marker, then followed prospectively every 6 months at 18, 24, and 30 months of age. At the initial and follow-up visits, interviews were conducted with a household adult aged 18 years or older using a standard questionnaire asking about the use, care, and, if the LLIN was not present, reason for attrition. Also at each visit, nets were examined for repairs, burns, holes, and tears. Holes were counted and measured in the field using the thumb, fist, head method described by the WHO [
3]. Hole sizes were approximated as follows: smaller than a thumb (0.5–2 cm diameter), larger than a thumb but smaller than a fist (2–10 cm diameter), larger than a fist but smaller than a head (10–25 cm diameter), and larger than a head (>25 cm diameter). Then, proportional hole index (pHI) was calculated by weighting each hole by its size and summing as described elsewhere [
12]. The weights were derived by taking approximations of hole area for each size category (1.23 cm
2 for holes categorized as “less than a thumb”, 28.28 cm
2 for holes “larger than a thumb but smaller than a fist”, 240.56 cm
2 for holes “larger than a fist but smaller than a head”, and 706.95 cm
2 for holes larger than a head), and dividing these hole areas by the smallest hole area (1.23 cm
2). For the smallest through largest categories of holes, the weights were 1, 23, 196, and 576, respectively. The location of holes was also documented (roof, upper half, lower half).
Net collection and preparation
At the initial visit when nets were 12 months of age and at the follow up visit at 24 months of age, random samples of nets were collected for bioassays and chemical analysis. A total of 74 LLINs were collected; 36 at the initial visit (18 in each province), and 38 at the 24-month visit (18 in one province, 20 in the other). Sections were cut from five areas of the net including the roof and side (top ¼, upper-middle ¼, lower-middle ¼, lower ¼); five 30 × 30 cm sections for bioassays, and five 10 × 10 cm samples for chemical analysis. Samples were individually wrapped in foil, and placed in a black plastic bag for transport and storage.
Chemical analysis
For the initial samples of nets at 12 months, only four specimens (from the roof, top 1/4, upper-middle 1/4, lower-middle 1/4) were analysed, with the sample from the bottommost location on the net being unavailable. For the nets at 24 months, specimens from all five locations of the net were analysed. The specimen set from each net was analysed as a group to yield an average value of insecticide concentration for the net. Chemical analysis was based on methods published by the Collaborative International Pesticides Analytical Council (CIPAC) [
13,
14].
Deltamethrin analysis of PermaNet 2.0 net samples was based on CIPAC method 333 [
14]. Each specimen set was weighed and added to a 125 ml Erlenmeyer flask fitted with a polytetrafluoroethylene-lined screw cap. Precisely 50 ml of an extraction solvent, consisting of 80/20 (v/v) isooctane/1,4-dioxane, was added to the flask. The flask was then sonicated for 15 min followed by 30 min of agitation in a shaker bath at 25 °C and 155 cycles/min. Approximately 1.5 ml of the extract was then transferred to a chromatographic sample vial using a glass syringe fitted with a 0.45 μm reconstituted cellulose syringe filter. High performance liquid chromatography (HPLC) was conducted on the extract using an Agilent 1200 HPLC equipped with a 150 × 4.6 mm (i.d.) Ascentis Si 5 μm column held at 40 °C. The mobile phase was 94/6 (v/v) isooctane/1,4-dioxane pumped at 1.5 ml/min. A UV detector set to 254 nm was used to detect analyte peaks. For each extract, three injections of 20 μl were made, averaged, and the results compared to those of two samples containing known deltamethrin concentrations (i.e. external standards).
Permethrin analysis of Olyset net samples was based on CIPAC method 331 [
13]. Each specimen set was weighed and placed in a 100 ml round-bottom boiling flask, followed by heptane (50 ml) and triphenyl phosphate internal standard (5.0 ml of known concentration in heptane). The flask was fitted with a reflux condenser and heated to boiling for 45 min. After cooling, approximately 1.5 ml of the extract was transferred to a chromatographic sample vial using a glass syringe fitted with a 0.45 μm reconstituted cellulose syringe filter. Gas chromatography (GC) was conducted using an Agilent 6890 N chromatograph fitted with a 30m × 0.25 mm (i.d.) fused silica DB-1 capillary column coated with 0.25 μm cross linked polydimethylsiloxane stationary phase. Ultra high purity nitrogen (1.2 ml/min) was used as the carrier gas. Extract (1 μl) was injected into the column inlet using a split flow rate of 96.1 ml/min. Injector port, column oven, and detector temperatures were 265, 240, and 300 °C, respectively. Flame ionization was used for analyte detection. Two injections were used for each sample and the results averaged. Permethrin concentration was calculated by comparing permethrin/triphenyl phosphate peak area ratios against a calibration curve generated from solutions containing known permethrin/triphenyl phosphate mass ratios.
Statistical analyses
All analyses were done using SAS® v9.3 (SAS Institute, Cary, North Carolina, USA). Baseline characteristics such as net use, presence of holes and repairs, and reasons for LLIN loss were summarized with simple frequencies. Medians and interquartile range (IQR) were calculated for continuous variables that were not normally distributed. Then, functional LLIN survival was examined looking first at attrition. LLINs gone due to destruction, discarding, or use for other purposes were classified as LLINs with “known” reasons for attrition, and these LLINs were used in the numerator to estimate LLIN attrition. LLINs lost for other reasons such as being given away, used in a different location, stolen, no longer being used because a new LLIN was received, or lost to follow up (ex: family moved or not home), were categorized as “outcome unknown” and were not included in the denominator when calculating attrition. The formula for functional survival was:functional survival = (nets in “good” or “damaged” condition as defined by pHI)/(All nets present + nets missing due to known reasons).
Furthermore, to check representativeness of LLINs with known outcomes, baseline characteristics of nets with known outcomes were compared to nets with unknown outcomes using Chi square tests.
The other component of LLIN survival, physical integrity, was examined. The geometric mean (GM) and 95 % confidence intervals (CI) of the pHI were calculated and compared using t-tests for significance for nets at different ages and then stratified by LLIN type. This calculation was done for all LLINs in the study, and then for only the subset of LLINs present at all follow-up visits to better examine changes in median pHI over time in the same net. Then, using pHI values, nets were categorized as good (0–64), damaged (65–642), or too torn (>643) as suggested by WHO [
7] and these proportions were compared by LLIN age and type.
The relationship between pHI and factors that might influence hole formation was then examined. Generalized estimating equation models of log-transformed pHI were done, and independent variables of interest were factors such as LLIN age, net type, burns, washing, and repair.
To examine if a particular part of the net was more prone to hole development, the geometric mean 95 % CIs of pHI values were calculated for different areas of the net—the roof, upper half, and lower half. Generalized estimating equations were also done with log-transformed pHIs to examine hole development by area of the net. Then, proportion of LLINs functionally surviving was calculated for each follow-up time point. Nets present at a particular time point for follow up and nets with a known reason for attrition were included in the denominator. The numerator was the number of LLINs that were present at that time point and classified as “good” or “damaged.” LLINs in the “outcome unknown” category were censored at time of net loss. These proportions and their confidence intervals were plotted by years since distribution, and compared against reference LLIN survival curves provided by the WHO [
7]. The failure endpoint was therefore defined as either an LLIN classified as “too torn” or an LLIN having a known reason for attrition. Then, a Kaplan–Meier survival analysis was done and used to estimate overall median survival time, and to compare median survival time between net brands. Using the same definitions for censoring and failure endpoint, survival analysis using Cox proportional hazards (PROC PHREG in SAS v9.3) was also done to examine factors that might affect survival such as LLIN type, use the night before the survey, having ever been repaired, having ever been washed, and use over a reed mat. For these survival analyses, if a net was missing an observation point between the initial and a subsequent follow-up visit (for example, an LLIN present at enrolment, missing at 18 months of age, present at 24 months of age), the outcome status of the net was assessed at its last available time point.
For the bioassay results, geometric means and 95 % CIs were calculated for KD60 and mortality at 24 h, and compared with significance testing via t-tests. While there is no cutoff for net failure for bioassay results, the WHO Pesticide Evaluation Scheme (WHOPES) criteria for optimal bioefficacy are ≥95 % KD60 or ≥80 % mortality in nets that have had at least 20 washes and 3 years of use [
12]. Proportion of nets with meeting these optimal bioefficacy levels of KD60 and mortality were calculated. A minimal bioefficacy criteria of ≥75 % KD60 or ≥50 % mortality has been used in the field [
15], so proportion of nets meeting this minimal criteria was also calculated. Using Chi squared tests, proportions of nets meeting optimal and minimal bioefficacy results were compared at 12 and 24 months, and further stratified by net type. The bioassay results were stratified by LLIN type, age, and history of washing. The chemical analysis results were summarized as median insecticide content expressed in both mg/m
2 and g/kg. Target insecticide content as recommended by WHO was 55 mg/m
2 (or 1.8 g/kg for 75 denier net, and 1.4 g/kg for 100 denier net) deltamethrin used in PermaNet 2.0 nets, and at least 20 g/kg of permethrin used in Olyset nets [
16,
17]. Results for chemical analysis were compared for nets of different ages and history of washing. Then, to describe the relationship between bioefficacy and chemical content in the study nets, the correlation between bioassay and chemical analysis results was examined by using R (R package version 1.4.8, Stanford, California, USA) to create McFadden probit models to obtain a pseudo R-squared value.
The protocol for this study was approved by investigational review boards at the US Centers for Disease Control and Prevention and the Tropical Disease Research Centre in Zambia. Written consent was obtained, using a consent form that had been translated to the local language, from an adult over the age of 18 years old at participating households.