Four arrondissements (sub districts) were selected, two in the South: Kessounou, in Oueme department and Allada, in Atlantic department, and two in the North: Kandi1 and Malanville, both in Alibori department (Figure  1). Residents of Kessounou, located on the Oueme River, have ready access to water for washing nets. In contrast, residents of Allada must carry water for washing to their homes. Similar criteria (easy access to washing site versus more difficult access) were applied to the selection of tracking sites in the North. Malanville is located near the Niger River, where water for washing nets is easily accessible. In contrast, Kandi, like Allada in the south, has a more remote water source.

To test the hypothesis that conditions on the ground affect indicators of LLIN useful life, tracking sites were intentionally selected because of ‘differences’ that would, most likely, change loss rates associated with the three indicators: (1) proximity to water for washing LLINs (expected to increase loss associated with durability and bio-efficacy by increasing the frequency of net washing); and (2) mosquito biting density/nuisance level (expected to increase loss associated with durability by increasing the frequency of LLIN use, and in turn LLIN wear and tear). Table  1 summarizes these differences.

This tracking study was planned under the Ministry of Health and approved by the National Ethics Committee for Health Research of the Ministry of Health of Benin. Community leaders were informed before the study and all gave consent before initiation. Written consent was also obtained from all participating households.

In July 2011, around four millions of Olyset® nets, a polyethylene 150D LLIN (PE-150D) impregnated with permethrin (2%), were distributed throughout the country. The monitoring tool was used at the four sites to monitor the durability of the LLIN product distributed in order to provide information for future procurement according to WHO guidelines [20].

Before the distribution campaign, a census of all households was carried out throughout the country including our study sites. The census and the distribution process was already described by the Ministry of Health [21]. The census recorded the name of the village, the name of the head of household, the household identification number, the number of adults and children living in the house and distributed a coupon, used to obtain the LLINs. The information was recorded in a master list. Net allocation to the household was based on the household size and the ratio one net for two persons (the national policy for universal coverage). This distribution campaign covered an average of 90% of the households in the country.

LLINs distributed have labels sewn into them at factory that helps to distinguish them from those distributed during other campaigns or received from other sources.

WHO suggested estimating sample size on the basis of the attrition of the LLIN product. But in Benin, PE-150D product attrition rate is not known. Sample size was then estimated following the WHO guidelines, which reported that a sample of 250 LLINs will allow detection of a 10% point difference if the best-performing product has an attrition rate of 10% and a 12 point difference with an attrition rate of 20% [20]. This sample size was doubled and 500 LLINs were selected by study sites to provide more precision in point difference detection. A subsample of 50 LLINs representing 10% of the total sample was randomly selected to measure insecticidal activity.

One LLIN per household was selected at each study site. Household selection at each of the four sites took into account the number of villages to ensure a representative sampling of the study site. For example, Kessounou is divided in five villages: Kodonou, Kessounou, Glehoue, Hetin-Sota and Glahounsa. Because the assessment was based on information from 500 LLINs, approximately 100 households (LLINs) were selected at random in each of the five villages. Table  2 summarizes the selection process. Households included in each village were randomly selected from the registration list of the distribution.

Two teams composed of two technicians and a local village health worker (VHW) visited each selected household. The head of household or an adult person acting on behalf of the head was interviewed. In certain occasions where no appropriate respondent was found in a particular household, the visit of the next home was scheduled. Approximately 500 households (500–501), where campaign net(s) had been hung (and were in use), were selected at each tracking site. The teams identified the campaign LLIN (on the basis of the label sewn) in each selected household (or randomly chose one net, if multiple nets were present). They marked the LLIN for tracking assessment using a double marking system, an additional label with a unique study code plus an indelible ink mark. The team also recorded the GPS coordinates of each selected household for follow up visits. Google Earth 6.1 was used to map the study households. Written consent, to return at 6-monthly intervals, was obtained from the head of household or an adult living in the household.

Households included for follow-ups were located by the name of the head of household and by GPS coordinates. Households that were not opened for inspection after two visits at the 6 months assessment visit were visited at the following assessment visits until they are recorded as unresponsive for three assessment visits and replaced.

A questionnaire was used to collect data from the head of household or an adult living in the household. The collected information included the status of each LLIN, the pattern of LLIN use and handling, observations on fabric integrity and the condition of the LLIN.

GPS coordinates and net codes were entered into the database. In the field, data were recorded on PDAs (Samsung Galaxy Tablets).

At the enrollment visit (T0), none of the LLINs had holes (loss of fabric integrity was 0%). After 6 months, LLIN fabric integrity was assessed by a visual examination, without removal of coded nets from tracking households. LLIN physical integrity was assessed by verifying if the coded LLINs had holes or not and recorded the major holes category found on the LLINs as follow:

smaller than a thumb (0.5–2 cm),

larger than a thumb but smaller than a fist (2–10 cm) and

larger than a fist but smaller than a head (10–25 cm)

Holes less than 0.5 cm were ignored. The 6 months follow up assessment only recorded the major categories (no holes size-4 were counted) of holes found in the LLINs but did not count them. Their natures, locations, evidence of repairs and the type of repair were also not recorded, and represented an important limitation of the 6 months assessment study.

WHO recommends the use of the cone bio-assay method [20] for monitoring bio-efficacy. However, problems related to rearing the large numbers of colonized, pyrethroid-susceptible vector females, needed to support the application of this method to a statistically meaningful number of LLINs, hinder its correct use. An alternative method, the colorimetric test was developed for use with deltamethrin-treated LLINs [22,23]. Colorimetric assessment of LLIN bio-efficacy involves a two-step process. In step one, a magnetic sampling device (MSD) is used to sample the insecticide level on the LLIN surface (without removing or damaging the LLIN). The amount of insecticide on the MSD sample, a filter paper disc, is proportional to the amount of insecticide on the LLIN. Because sampling is standardized, results for different LLINs in the sampling frame are comparable. The second colorimetric assessment step uses colorimetric chemistry to estimate the amount of deltamethrin in the sample. Colorimetric results have been validated, by comparing WHO cone test bio-assay results and colorimetric results for a series of LLINs with different surface levels of deltamethrin. Using the standard curve from this study, it is possible to interpret the colorimetric result in terms of whether or not an LLIN meets the minimum WHO ‘threshold’ for bio-efficacy (nets causing >80% mortality in a cone bio-assay test).

A conceptually similar ‘chemical test’ approach was used for tracking the bio-efficacy of the PE-150D LLINs in this assessment. The approach, based on gas chromatography (GC), was developed in order to circumvent the problems previously described that are associated with the WHO cone bio-assay method for bio-efficacy assessment. With PE-150D technology, the insecticide molecules migrate to the surface continuously replacing lost insecticide at the surface, where vectors are exposed to its effects. Based on this theory, the LLIN surface insecticide level sampling tool was modified to enable sampling of PE-150D LLINs without removal or replacement; samples were collected on position B (Figure  2). This was done as a precautionary measure since the owners may put blankets on top of position D of the net which may rub some of the insecticide off. Also, placement on Position A under the mattress may result in the removal of the insecticide. Therefore, the best position to collect the sample is position B in the middle of the net.

The net to be tested is hung and position B is identified and fixed on an embroidery hoop (4 inches diameter). A portion of lens paper is applied to the surface of a cap attached to a 50-mL plastic tube. The lens paper is rubbed along the inside diameter of the hoop for 10 rotations. The lens paper (25 mm diameter) is removed and inserted into a 1-mL syringe and compressed with the plunger. The reverse side of the net is repeated in the same manner (Figure  3). 100 μL of acetone containing 0.05 mg/mL of triphenyl phosphate as an internal standard is added to the syringe containing the 2 compressed portions of the lens paper. The syringe outlet is plugged to allow the paper to soak for 5 minutes. After removal of the plug, the acetone is eluted via the syringe plunger into a small tube followed by another 100 μL of acetone. 1 μL of the eluant is injected into the GC (SRI 8610C GAS CHROMATOGRAPH) and the permethrin and internal standard is detected using flame ionization detection with hydrogen as the carrier gas. Sample response is compared with the response from a calibration standard and the amount of permethrin adhering to the lens paper is determined.

To estimate a bio-efficacy threshold for the GC method, 37 PED-150 LLINs (11 new nets and 26 nets that were in use in field for 3 to 4 months) were submitted to Cone test bioassay following WHO guidelines [24]. Interpretation of results (how GC test data are used to predict whether or not an LLIN would meet the minimum WHO bio-efficacy threshold) was done based on the comparison approach, outlined for the colorimetric method. A cut off value, established by comparing GC results with WHO cone test bio-assay results for the same LLINs, some of which met or exceeded the WHO minimum threshold for bio-efficacy and some which did not, was used to interpret the GC results. Receiver Operating Characteristics (ROC) analysis, applied to the two data sets, GC and WHO cone bio-assay, was used to define the cut off value for nets that would be expected to meet the WHO cone test threshold. A minimum GC permethrin concentration was identified and used as a cut off value to estimate the percentage of nets at T0 and T6 that would not meet WHO minimum standards for bio-efficacy.

The questionnaire form used in the study was created with ODK Collect 1.2.2 which allows easy data collection on Samsung Galaxy Tab 10.1 used as data terminals. At the end of each assessment visit day, data collected were directly uploaded to a cloud server, and retrieved at the end of the assessment for analysis.

Approximately 10% of the households were revisited independently to make sure that the survey was not biased.

For assessment of bio-efficacy, a sub-sample (n = 50/site) of LLINs was randomly selected and tested to determine the amount of insecticide on the surface of the net. The percentage of nets in the sample falling below a cut-off value for LLINs that cause > 80% mortality in a WHO cone bio-assay test [24] was used to quantify net loss associated with bio-efficacy at T6.

Other factors that contributed to the LLINs durability, as measured by fabric integrity and insecticidal activity, were assessed by multivariate regression analysis. The contributing factors included the house environment and behaviour related to net use, handling and washing, was derived from answers to the questionnaire used for the sub-sample of 50 LLINs selected for bio- efficacy in each location (around 200 LLINs in total).

Table  3 summarizes T6 results for survivorship and attrition. One hundred forty seven out of 2002 coded LLINs were no longer present in the households where they were located at T0. That is, after 6 months, survivorship had dropped from 100% to between 90 and 96%, giving attrition of 10%, 9%, 4% and 6% percent respectively in Kessounou, Allada, Kandi and Malanville. Attrition rate-2 (LLINs removal) was the principal reason for LLINs lost; ranging from 4-8%. Attrition rate-1 was the second cause of LLINs lost; ranging from 0-2%. Attrition rate-3 was very low and only observed at Kessounou (Table  3). We did not assess factors associated with higher and lower survivorship loss rates, however, differences in survival between locations in the South and the North were significant (p < 0.05).

At the 6 months assessment visit, LLINs found with holes (Table  4) was 36% at locations with less access to water for washing and 52-64% at locations with ready access to water (p < 0.01). The number of LLINs with size 1 holes was relatively low at all sites (when compared with the number of LLINs observed with size 2 and 3 holes).

A total of 222 nets were subjected to GC analysis to assess bio-efficacy. Results are summarized in Figures  6 and 7. There was significant loss associated with bio-efficacy at all four sites during the assessment period (T0-T6) (p = 0.0001). However, there was no significant difference between locations with more versus less access to water for washing. Measurements were based on gas chromatographic analysis of LLIN surface samples collected at randomly selected households in each study at T6. Results are compared with baseline (T0) measurements on recently distributed nets using the same approach. The red dotted line shows the GC cut-off equivalent to a surface insecticide level high enough to give >80% mortality in the WHO cone bio-assay, a minimum threshold for adequate LLIN bio-efficacy set by WHO.

Bio-efficacy at baseline (T0) was high. The results showed that 98% (49/50) had surface insecticide levels that were above the minimum WHO threshold for bio-efficacy (loss associated with bio-efficacy at T0 was set at 2%). After six months of use, net loss associated with bio-efficacy was 58% in Kessounou, 52% in Kandi, 44% in Allada and 9% in Malanville. At T6 there were 42% LLINs below the GC cut off for WHO minimum bio-efficacy.

Table  5 indicated responses from a questionnaire used to assess the house environment and behaviour related to net use, handling and washing of the sub-sample selected. At Kessounou and Malanville, most residents indicated that they washed their nets between 2–5 times in the preceding 6 months, whereas most residents at Allada and Kandi, stated that washing occurred less frequently <2 times in the preceding 6 months. Modalities with very few responses were aggregated for the multivariate analysis.

In summary (Table  6), the 6 months data showed that frequency of LLINs use, sleeping material, washing frequency and the distance to water for washing were significantly associated with loss of fabric integrity (p < 0.05).

Table  7 showed that only factors like frequency of LLIN use and sleeping material were significantly associated with insecticide decay (p < 0.05). Factors like washing frequency, LLINs maintenance, location of the kitchen and distance to water for washing did not seem to play a key role in the insecticide decay observed.