The protocol was approved by the Ethics Committee of the Faculty of Medicine, Pharmacy and Odontostomatology (FMPOS) of the University of Bamako, and Malian Ministry of Health through the National Directorate of Health and the Directorate of Pharmacy and Drugs. Once the study sites had been identified, the aim of the study and details of the procedures to be undertaken were presented through the different levels of administration to the level of the individual villages where meetings were held to provide community understanding of the protocol. Once community agreement had been obtained, a census of each village was undertaken to identify households, families and individuals, each of whom received a unique number. Subsequently, potential subjects in the study provided their own written consent to information provided on audiotape in Bambara (the local language). Parents or guardians provided written consent for all children participating in the study.

Mali was chosen because trachoma and LF were co-endemic and the baseline prevalence for both diseases from prior surveys was relatively high. In addition, the infrastructure existed to allow a study of this nature to be conducted.

In this study, the treatments were allocated to entire villages (clusters). The decision to use a cluster design was taken for logistical reasons. With the number of subjects involved, it would be very difficult to implement individual randomisation within a village. Furthermore, the study needed to assess the effects of mass treatment in the community as that is the target setting for the treatments under study. The villages chosen for the study needed to be accessible, for logistical reasons, with access to a health facility, but sufficiently far apart to avoid “contamination” between them. They had to be within 5 km of a serviceable road and within 15 km of a Community Health Centre (CSCOM) at the very peripheral level of the health system organization in Mali or a district Reference Health Center (CSREF) that is responsible for providing care to complicated cases referred from the different CSCOMs of the district. Villages also needed to be separated by at least 15 km and be on an East-West axis since the endemicity/epidemiology of LF is more homogeneous than on a North-South axis. They also needed to have populations of about 1200 people of local ethnic groups with similar socio-cultural habits. The anticipated exclusion rate was 25% so village populations needed to be large enough to allow for 750 subjects to be recruited to the study. Two districts, Bougouni and Kolondièba, in the Sikasso region of southern Mali (Figure S1), were selected for the study. Prior to study initiation, these districts had both received four annual standard treatments with albendazole and ivermectin for LF and two mass treatments with azithromycin for trachoma. It is likely that these interventions had considerably reduced the trachoma infection rate and filarial parasite load compared to historical data for the area. For logistical reasons, the infection rates were not re-evaluated immediately prior to the study

Trachoma is highly endemic in the Sikasso region and preliminary results of surveys conducted by ITI in June 2008 (ITI unpublished data, 2008), overall prevalence of trachoma was 14.65% in Bougouni and 24.4% in Kolondièba, while the prevalence in 1 to 9 year old children of the main stages of trachoma, Trachoma Inflammation-Follicular (TF) and

Trachomatous Inflammation-Intense (TI) in Bougouni and Kolondièba were 5.8% (84/1456) and 7.5% (109/1453) respectively, a substantial reduction from earlier surveys [19]. The prevalence of LF in 2004 at before MDA was started was 19.8% in Bougouni and 22.4% in Kolondièba (National LF Elimination Programme, unpublished report). LF prevalence was not assessed in the study villages prior to the study initiation but data from neighbouring villages reported 14% and 24.4% respectively in the districts of Bougouni (Mena, Tienkoungoba) and Kolondièba (Bougoula, Kebila) (unpublished data, National LF Elimination Program, Mali).

Based on the selection criteria above, two villages in each district were chosen, Tienkoungoba (longitude −6.581690, latitude 11.456120) [standard treatment] and Ména (longitude −6.812560, latitude 11.527540) [co-administered treatment] in the District of Bougouni and the villages of Bougoula (longitude −7.094090, latitude 11.253810) [standard treatment] and Kebila (longitude −7.041090, latitude 11.276370) [co-administered treatment] in the District of Kolondièba. Allocation of treatment in each district was decided by the toss of a coin by the investigators following the investigators' training session in Bamako before the treatment phase field trip.

Prior to initiation of the field based phases of the study (census and treatment/follow-up), the whole team was recruited and underwent intensive training to ensure that every member knew and understood their role. Since the treatment phase of the study was to be conducted simultaneously in the 4 villages, the data collection teams were all trained together on study methodology and especially the use of the predetermined adverse event questionaires. The details of the processes and the materials used may be found in the Supplimentary file ‘Study Protocol’ Section 9 pages 32–35, and Appendices D through K.

Following a complete census of the participating villages, all residents were invited to attend for screening. Male and female consenting/assenting subjects resident in the village for at least the previous 3 months and aged between 5 and 65 years were screened against inclusion criteria with a history and medical examination, which included documentation of signs/symptoms of trachoma (according to the WHO classification, and identified by health technicians specialised in ophthalmology) and LF (elephantiasis, lymphoedema, hydrocoele), prior to formal enrolment. Enrolment continued until the target number of subjects, approximately 40% of the total popualtion in each village, was reached. Blood tests for microfilaraemia or antigenaemia were not taken. All children over 5 years of age but under 90 cm in height and all pregnant and lactating women were excluded. In addition, all those who had significant acute or chronic illnesses or had a history of allergy to the drugs to be given were excluded. Subjects excluded after consent and non-enroled village residents were offered treatment after the study using the standard regimens for trachoma and LF according to national guidelines.

Following consent and screening, all enrolled subjects within a village received the drugs appropriate to the randomization. Patients were identified after randomization since the whole village was alocated to the same treatment. Subjects were aware of the treatment they would receive during the consent process. Appropriate doses for azithromycin and ivermectin were determined using a height pole marked with the doses on either side of the pole (Table 1). In the standard treatment villages, all subjects were treated on day 0, receiving ivermectin based on height (up to 4×3 mg tablets) and albendazole irrespective of height (1×400 mg tablet). They then received azithromycin based on height (up to 4×250 mg tablets) 7 days later. In the coadministration villages, all subjects received ivermectin based on height (up to 4×3 mg tablets), albendazole irrespective of height (1×400 mg tablet) and azithromycin based on height (up to 4×250 mg tablets) on day 1. No treatment was given 7 days later. All treatment administration was witnessed by study staff to ensure the medications were taken. Study staff were not blinded to the treatment administration.

Subjects were seen and assessed on Days 1, 8 and 15. On Day 0, in addition to a general examination, they were asked about any complaints from a pre-specified list (e.g. abdominal pain, headaches) or others (recorded verbatim) that existed prior to treatment. On Days 8 and 15, the subjects underwent a further clinical examination and were asked whether they had experienced any AE since the administration of the drug, or whether any complaints present prior to treatment had changed in any way. Subjects were also able to present themselves at the health centres, which were manned throughout the study, if they were concerned about any events between the planned assessment visits. All AEs were classified either as ‘minor’ (not interfering with daily function) or ‘major’ (interfering with daily function). If AEs required treatment, this was based on the presenting signs and symptoms.

Study investigators were located in village health centers or at other appropriate sites recommended by villagers. Subjects went to the agreed site to be assessed for eligibility for the study. Households were grouped by sector and assigned investigators with the aid of volunteers chosen by the villagers. These volunteers also helped check that all households had been surveyed and assisted in finding those who did not return for study assessments on Days 8 and 15. Any subjects who had not returned for the Day 8 assessment by Day 14 were excluded from the study. Similarly, any subjects not returning for the Day 15 assessment by Day 17 were excluded.

The study was designed to show that there was no difference in the overall incidence of AE between the standard and coadministered treatment. Previous data on azithromycin suggested an AE rate of 5% [20], [21]. The incidence of AE with coadministered treatment was estimated as 8%. With power of 80% and a two-sided 95% confidence interval for the difference between rates, the sample size chosen was 1125 subjects per treatment. The sample size per group was increased to 1328 to allow for an 18% refusal rate. This was in turn rounded up to 1500 per group to increase the power. This required approximately 750 subjects per village in order to produce clusters of equal size. The original sample size calculations did not take into account that the study was based on clusters (i.e., villages), where all subjects within the same cluster received the same treatment and was thus grossly underpowered. Analysis was therefore conducted at an individual, rather than a cluster level.

The objective of the study was to assess safety of the three drugs when coadministered. The primary outcome was therefore the overall incidence of any AE. Any complaint, either one itemized in a pre-prepared list in the case record form (CRF), or any other untoward event occurring after treatment administration, or an existing complaint which worsened after drug administration, was considered as an AE. Each event was classified as minor (not interfering with daily function), moderate (interfering to some extent with daily function) or major (significant effect on daily function). The timing of each event was derived as the first day on which the event was recorded, calculated from the day of first drug administration. The duration was also calculated as the number of days from first appearance to last appearance (inclusive).

The incidence of each event itemized on the list in the CRF was calculated by counting the number of subjects who reported it at any time during the study. The incidence of any AE was derived in the same way, by counting the number of subjects reporting at least one event, whether on the list or not, at some stage during the study. Subjects reporting several events were only counted once. The time to first onset for each subject was calculated as the day on which the first AE of any kind occurred. The maximal severity for each subject was calculated as the maximal severity of any event reported during the study.

Since the study design did not fully account for the cluster design, and since there were only two clusters for each treatment, the scope for statistical analysis was limited. The overall incidence of AE in each village was calculated as a simple percentage of the total number of subjects who received treatment. A 95% confidence interval was calculated using exact methods if necessary (Wilson method, [22]). Note that even if there are no occurrences of a specific event in the study, the prevalence can still be estimated with its 95% confidence interval. The time to first onset of events in each village was investigated using survival analysis methods. Data from any subject reporting no AE by the end of the surveillance period was considered to be censored at the time of last assessment. The analysis was repeated for those subjects who experienced any AE and the median time to first onset was estimated with its 95% confidence interval. The severities of events were summarized in a table.

The frequency of occurrence of each event itemized in the CRF was summarized for each treatment within each district in tables and bar charts showing the percentage of the study population within the village who reported the event. The incidence of each type of event for each village was also shown with associated 95% confidence intervals graphically, with a separate plot for each type of event. The time to onset of the most frequently occurring events was also investigated using survival analysis methods.

The age-sex distribution of the study population within each village was compared to the overall village population using graphs.

The intra-class correlation coefficient and associated 95% confidence interval was calculated using the ANOVA estimator [23]. Due to the small number of clusters, the resulting estimate should be treated with caution.

There were 9109 persons in the villages in the two districts of whom 7515 were eligible for inclusion (aged between 5 and 65 years, inclusive). 3011 subjects were chosen at random from this pool to be included in the study population (Table 2), selection continuing until the target number in a village was reached, with 1510 receiving the standard treatment (755 in each district) and 1501 receiving the coadministered treatment (750 in Bougouni, 751 in Kolondièba) (Figure 1). Recruitment of subjects started on February 7, 2010 with the first treatment being administered the same day while the final subject assessment was done on February 26, 2010. The recruitment took 3 to 4 days per village and all the villages began the recruitment on the same day

The male to female ratio was approximately equal in all villages (Table 2). In all villages, the proportion of males included in the study was slightly higher than the overall proportion of males in the village, with the proportion of females being correspondingly lower. The village allocated the coadministered treatment in the Kolondièba district (Kebila) was larger than the rest, hence the study population formed a smaller proportion of the total village population.

The age distribution of the study population in each village was similar (Table 2), with approximately 70% aged under 19 years and 10.8% aged over 50 years which represents the age structure within the villages as a whole. The age ranged from 5 to 65 years in all villages, with a mean of 20 years and a median of 12 years (Bougouni) or 13 years (Kolondièba).

The majority of study subjects were from the treatment villages (Table 2). The proportion of outsiders recruited to the study in Kolondièba was slightly higher (3.9%) than in Bougouni (0.3%), although the majority came from neighbouring villages and had been domiciled in the treatment villages for at least 3 months. None came from non-endemic areas. The height and weight distributions were similar for the four villages.

Slightly less than 5% of the total study population were found to have signs of trachoma. Two of the villages (Bougouni standard and Kolondièba coadministered treatment) had low rates of trachoma (0.9% and 1.9%) while the other two (Bougouni coadministered and Kolondièba standard treatment) had higher rates (6.0% and 8.7%). For these latter two villages the most frequent types of trachoma were follicular and scarring trachoma (eyelid scarring). 12 of the 14 subjects in Kebila (Kolondièba, coadministered treatment) had follicular trachoma. No subject in any of the villages had corneal opacity. Other findings from the eye examination were seen for between 0.8% (Kolondièba, coadministered treatment) and 5.9% (Bougouni, coadministered treatment). The pattern was similar as seen for the overall incidence of trachoma, with a lower level of other findings for Bougouni standard and Kolondièba coadministered treatment (less than 3% of subjects) and higher levels for Bougouni coadministered and Kolondièba standard treatment (above 4%). The most frequently occurring findings were chronic tropical endemic limboconjunctivitis (TELC), cataract, pterygium and conjunctivitis.

The number of abnormal findings from the LF examination was very low: 12 subjects with hydrocoele (6 on each treatment). One subject (Kolondièba, coadministered treatment) had both hydrocoele and lymphoedema: this was the only subject to have lymphoedema.

The number of tablets administered was almost always correct according to height. There were more dosing errors for azithromycin than for ivermectin. There was a slightly higher error rate for the administration of the coadministered drugs than for the standard treatment: 13 subjects (0.9%) and 27 subjects (1.8%) had incorrect numbers of ivermectin and azithromycin tablets respectively in the coadministered groups compared to 3 (0.2%) and 12 (0.8%) for the standard treatments.

There were two instances of regurgitation of drugs on Day 0, both with the coadministered treatment. There were also two cases on Day 8 (azithromycin only). The drugs were not taken again in only one case. The majority, at least 95%, of subjects in all villages except Kebila (Kolondièba, coadministered treatment) took their drugs after a meal: in Kebila, only 577 subjects (76.8%) took their drugs after a meal. On Day 8, all the subjects with data recorded took their medication (azithromycin only) after a meal; 2 subjects had no data on time of drug intake recorded.

There were no losses to follow-up, all subjects allocated to treatment being evaluated within the timelines established for the protocol.

Table 3 summarizes the prevalence of complaints reported prior to treatment administration, with the most frequently occurring shown in Figure 2. The prevalence ranged from 26.2% (Kolondièba, standard treatment) to 46.9% (Bougouni, standard treatment). The most frequently reported of the complaints named in the CRF were headaches (348 subjects, 162 on standard treatment, 186 on coadministered treatment) with prevalence ranging from 8.3% (Kolondièba, standard treatment) to 15.3% (Kolondièba, coadministered treatment), and abdominal pain (239 subjects, 118 on standard treatment, 121 on coadministered treatment), with prevalences ranging from 4.9% (Bougouni, coadministered treatment) to 11.2% (Kolondièba, coadministered treatment).

However, there were more subjects with pre-existing complaints classified as “Other”: 496 subjects (202 on standard treatment, 294 on coadministered treatment). The frequency of these varied widely between the villages: 57 subjects (7.5%) in Bougoula (Kolondièba, standard treatment), 100 subjects (13.3%) in Ména (Bougouni, coadministered treatment), 145 subjects (19.2%) in Tienkoungoba (Bougouni, standard treatment) and 194 subjects (25.8%) in Kebila (Kolondièba, coadministered treatment). The most commonly reported were cough (all villages except for Bougoula), lower back pain (Kolondièba, both villages), epigastralgia (Bougouni, both villages) and various other types of pain.

Table 4 summarizes the overall incidence of AE. The incidence of events in each village with its 95% confidence interval is also summarized in Figure 3.

The confidence intervals are all narrow due to the large sample size in each village, and there is no overlap between the four. The difference between treatment regimens differs between the districts, with a higher incidence for the coadministered treatment than the standard treatment in Kolondièba (30.4% vs. 11.0%), but a lower incidence in Bougouni (7.1% vs. 20.9%). Due to the clear differences between the incidences of AE in each village, formal statistical comparison between the two treatments is not appropriate. The intra-class correlation coefficient (ICC) was estimated to be 0.069 (95% CI: 0 to 0.174).

No serious adverse events were reported following treatment. Twelve subjects (8 out of 241 with any AE on standard treatment and 4 out of 281 on coadministered treatment) reported events of major severity. Where necessary, subjects were treated symptomatically, most requiring paracetamol for pain or headache.

Most subjects only experienced one type of AE: 66.4% (160/241) of subjects reported an AE with the standard treatment and 70.1% (197/281) with the coadministered treatment. Fewer than 2% of all subjects in each village reported 3 or more events. Of those subjects who did report an AE, 7.6% and 12.0% in the control villages and 9.4% and 5.7% in the coadministration villages reported 3 or more. One subject (Bougouni, standard treatment) had seven types of AE recorded.

Overall, there was no difference in AE rates according to whether there were pre-existing complaints reported.

In most cases, pre-existing complaints had either disappeared or improved by Day 8. Some were ongoing, and a small number still continued up to Day 15. At the Day 8 assessment, a small number of complaints were reported as worse after treatment: these were mainly headaches (7 subjects) and abdominal pain (6 subjects). There was also one case each of fever, joint pain and deafness, which were exacerbated by treatment. At the final (Day 15) assessment, only one subject reported an exacerbation of a pre-existing complaint. This was a case of abdominal pain in the standard treatment (Kolondièba).

All but one of the headaches exacerbated by treatment were reported as AEs. Of the six subjects with abdominal pain, four were reported as an AE, two were not. The case of worsened fever was not reported as an AE, nor was the exacerbated deafness. There was one case of exacerbated joint pain that was reported as an AE.

Table 5 shows the frequency of each individual type of event as listed in the CRF. The data is summarized graphically for the most frequent events in Figure 4. Abdominal pain, headaches, diarrhea and events other than those itemized in the CRF were the most common types of event in all villages. The incidence of each of the main events with 95% confidence intervals is summarized for each village in Figure 5.

Overall, and in three of the four villages, the most frequently reported event was abdominal pain, with rates ranging from 2.1% (Bougouni, coadministered treatment) to 17.2% (Kolondièba, coadministered treatment) which was considerably higher than in any of the other three villages (Figures 4 and 5). The exception was Tienkoungoba (Bougouni, standard treatment), where more people experienced headaches (55 subjects, 7.3%) or other types of event (56 subjects, 7.4%). The incidence of diarrhea was slightly higher in the coadministered treatment village in Kolondièba than in the other three villages (6.7% compared to rates below 3%). The incidence of joint or muscular pain was similar in all four villages.

The only other events reported by more than 1% of the study population in any of the villages were vomiting (2.8% Kolondièba, coadministered treatment), nausea (2.0% Kolondièba, coadministered treatment) and fever (1.2%, Bougouni, standard treatment). Of these, the only event for which the upper limit of the confidence interval exceeded 4% was vomiting (Kolondièba, coadministered treatment, upper 95% confidence limit = 4.2%).

The upper 95% confidence limit for the prevalence of events specified in the CRF but which were not experienced by any subjects in the study is 0.5%. With the exception of nausea and vomiting, the incidences of all other AE types specified in the CRF are comparable between the villages, with overlapping confidence intervals.

Of the five subjects who had pre-existing complaints exacerbated which were not recorded as AE, two reported worsened abdominal pain, and one each reported heightened fever, headache and exacerbated deafness.

Of the 12 subjects with AE classified as major, 8 received standard treatment (6 in Bougouni district, 2 in Kolondièba) and 4 received coadministered treatment (1 in Bougouni district, 3 in Kolondièba). The distribution of events and their nature is shown in Table 6.

There was no clear pattern to the types of event reported as “Other”. In Tienkoungoba (standard treatment, Bougouni), the most common events were respiratory in nature (18 cases, comprising coughs and rhinitis), and vertigo (15 cases). In Bougoula (standard treatment, Kolondièba) where the event rate was lower, vertigo was the most commonly reported “Other” event (6 cases). In Kebila (coadministered treatment, Kolondièba), the most common events were somnolence (26 cases) and vertigo (9 cases). The most frequent type of “Other” event in Mena (coadministered treatment, Bougouni) was epigastralgia, reported by 3 subjects.

There is a clear difference between the villages in time to onset of events, with most of the events occurring immediately after treatment in Kebila (Kolondièba, coadministered treatment, Figure 6). There is also evidence that the event rate increased on day 8 with the administration of azithromycin in the standard treatment villages.

The median time to the onset of the first event, of any type, calculated from all those with any event, is 8 days for the two standard treatment villages compared to 2 days and 0 days for the coadministered treatment villages in Bougouni and Kolondièba respectively.

For the five most common types of event, namely, headaches, abdominal pain, diarrhea, joint/muscular pain and other non-specified events, a similar pattern in terms of time to onset was seen. This was most marked for abdominal pain: in Kebila (Kolondièba, coadministered treatment), almost all cases of abdominal pain first occurred on the day of drug administration. For diarrhea and events not specified in CRF, however, the time to onset was most rapid in the Kolondièba coadministered treatment village. The median time to onset of events with the standard treatment was always Day 8, coinciding with the administration of azithromycin. It is clear from the graphs that onset was quickest with the coadministered treatment in Kolondièba, with most events being reported as starting on the day of treatment administration. Duration of adverse events (number of days from first to last reporting of an event) is summarized in Figure S2

The design of the study with a small number of large clusters means that comparisons of the overall AE rates for the two treatment regimens are difficult to make since there were only two villages receiving each treatment, and the AE rates differed markedly between villages. Nevertheless, there is no clear evidence to suggest that the coadministered treatment is associated with a higher rate of AE than the standard treatment. The types of event occurring most frequently were similar for all four villages, and the events reported, especially abdominal pain and diarrhoea were temporally accociated with azithomycin administration as expected. The study demonstrated that it is feasible to administer the three drugs together, which would significantly reduce the logistical demands on conducting mass treatment of LF and trachoma in this setting.

However, given the limitations of the current study, further investigation would be desirable. Any new study will need to consider the design issues raised here, and especially whether a cluster randomised approach is appropriate. Furthermore, the studies need to be conducted in areas where there is an appreciable prevalence of LF to determine whether coadministered treatment increases the incidence of AEs associated with microfilaraemia.