The study was approved by the Ethical Committee of the Ministry of Health (MINSANP/CE ref.-nr. 051) and was registered at Controlled-trials.com (ISRCTN 11415557). The study was conducted according to the principles expressed in the Declaration of Helsinki. Informed written consent was obtained from all participants in Malagasy and in the case of minors from the parents or legal guardians. At the end of the study, all participants received a pair of closed solid shoes. In each participant, any remaining viable sand fleas were removed under sterile conditions. The flow chart of the study is depicted in Figure 1.

The study was conducted in the villages Tanambe II (507 inhabitants) and Tanambaovao (486 inhabitants) located in Andasibe community, Moramanga district, central Madagascar. The villages are traditional communities, where people live from subsistence farming. Similar to other communities in the hinterland of Madagascar, tungiasis is common. Local people consider sand fleas disease to be inevitable and remember that it always existed.

In the villages roads are not paved. The houses are made from wood and many are constructed on stilts (Figure 2A, 2B). The houses possess a front court used for cooking, washing and drying clothes. Chicken, pigs, dogs and cats are kept as domestic animals and live inside the compound. Water for cooking is derived from public pumps, water for cleaning the body and washing clothes is taken from a river close to the villages. Rice, cassava and pineapples are the main crops. Children usually do not possess shoes. Adolescents and adults have shoes, but do not wear them regularly. Most sand flea lesions are manipulated. Extraction of embedded sand fleas is attempted as soon as the lesions become painful. This is done using non-sterile instruments, such as fixing pins or needles. Neither the instrument nor the skin is disinfected. Often, the extraction is performed by elder members of the household and not by the patient himself (M. Thielecke unpublished observation 2011).

In Madagascar, the occurrence of tungiasis follows a characteristic seasonal pattern: the incidence starts to rise with the beginning of the hot and dry season (May), peaks around September and starts to decline with the beginning of the rainy season (November).

A randomized controlled field study with three arms was carried out between May 31, 2011 (baseline investigation) and August 24, 2011 (final follow up). This period was chosen to coincide with the beginning of the dry season and the high transmission of T. penetrans. The study was conceived as a comparison of individuals who were provided with and encouraged to wear closed shoes and those who got a twice-daily application of the repellent to untreated individuals.

Individuals with tungiasis were identified with the assistance of community health workers. Inclusion criteria were: age ≥5 years and at least one person in the household with ≥7 lesions at both feet. The latter prerequisite was based on the rationale that, in this case, the average attack rate in household members should be high. Exclusion criteria were the presence of ≥70 lesions (total number of viable or manipulated sand flea lesions), or the presence of severe acute morbidity necessitating immediate surgical or antibiotic treatment. Eligible individuals were only enrolled if they were permanent residents in the villages, intended to stay in the community for the next four months and provided an informed written consent. In total, 240 participants were recruited and 219 were randomized.

The major outcome measures were the intensity of infestation, i.e. the number of sand flea lesions present on both feet at the time of examination, the number of newly penetrated sand fleas since the last examination (considered as a proxy of the attack rate [25], [27]), and the severity of acute and chronic pathology as measured by the severity score for acute tungiasis (SSAT) and the severity score for chronic tungiasis (SSCT) [28]. Outcome measures were assessed every two weeks by the same investigator (M. T. and V. R.). No participant was missing on more than one consecutive follow-up examination.

To avoid that members of one and the same household would participate in different treatment arms, we randomized by household. Randomization into three groups was performed using a computer-generated random number table. Participants of group I received a pair of closed solid shoes. To the feet of the participants of group II the repellent was applied twice daily. Group III did not receive any intervention. 21 households belonged to the shoe group, 20 to the repellent group and 21 households to the control group. Households with study-participants were homogenously distributed across the village. The median number of household members was 3 (interquartile range [IQR] 2–5) in the shoe, 4 (IQR 2–5) in the repellent and 3 (IQR 2–5) in the control group.

The intensity of infestation, the attack rate and the degree of tungiasis-associated morbidity were assessed as described previously [25], [28]. Staging of lesions was performed according to the Fortaleza Classification [29]. The following findings were considered diagnostic for tungiasis:

At each examination the number of viable (stage I to III) and dead (stage IV) fleas, the total number of embedded fleas ( = intensity of infestation), and the number of parasites having penetrated since the previous examination ( = attack rate) were documented. Manipulated lesions (such as partially or totally removed parasites leaving a characteristic crater-like sore in the skin), and suppurative lesions caused by the use of non-sterile perforating instruments were noted as well. At each examination data of the two feet were combined.

The exact topographic localization of each lesion, its stage, and appearance were documented on a visual record sheet and lesions were photographed, using a digital camera equipped with a macro objective (EOS 450 D, Canon, Tokyo, Japan).

Clinical pathology was assessed in a semi-quantitative manner using the severity score for acute tungiasis (SSAT), and the severity score for chronic tungiasis (SSCT) [28]. The SSAT score comprises the following signs and symptoms: erythema, edema, pain upon pressure or spontaneously, itching, sleep disturbance due to itching or pain, difficulty walking as indicated by an altered gait; abscess, and suppuration as indicators of superinfection; fissures, perilesional desquamation and ulcers as characteristic chronic skin defects. The score can take a value from 0–35 points. The SSCT ranges from 0 to 30 points and comprises the presence of nail deformation, nail loss, deformation of toes, hypertrophic nail rim, and hyperkeratosis; all of those characteristics are indicators of repeated episodes of tungiasis experienced in the past [28].

Since the distributions of the outcome measures were skewed, the median and the interquartile ranges were used to indicate the central tendency and dispersion of data, respectively. Since the randomization was based on households, outcome measures in participants were considered to be correlated (i. e. not to be independent). To compare measurements between groups the method of generalized estimating equations (GEE) was used [30]. GEEs are appropriate to analyze longitudinal and other correlated data, especially when they are in the form of counts [30]. In all GEE models the baseline measure of the dependent variable was used as a covariate. Intention-to treat and per protocol data analyses were performed.

Assuming that the application of the repellent and the availability of shoes would have a similar protective effect, 75 participants were needed in each of the three groups to detect a difference of 50% in the intensity of infection between an intervention group and the control group (α = 5%; power of the test = 86%). A dropout rate of 25% until the end of the study was included in the calculation.

The demographic and clinical characteristics of the study population at baseline are shown in Table 1. There were more female participants in each group than males. This was due to the fact that many of the male inhabitants were not eligible for the study as they worked in town during the week or were absent >8 hours, so that they could not be examined. The median intensity of infestation was identical in the shoe and the control group (22 lesions), but slightly lower in the repellent group (16 lesions).

Table 2 depicts the time course of the four major outcome measures. Onward from week 2, the attack rate became zero and the intensity of infestation and the SSAT decreased significantly in the repellent group, whereas no significant change occurred in the control group. In the shoe group, there was a tendency of decrease in the intensity of infestation and of the SSAT over time.

In order to avoid bias in the comparisons of the follow-up data between the groups, the absolute difference to baseline was calculated for the intensity of infestation, the SSAT and the SSCT. Figure 3 shows the time course of the intensity of infestation. In the repellent group the decrease was already significant at week 4 (p = 0.02 compared to the control group). In contrast, at no point of time the difference between the shoe cohort and the control cohort was significant.

The time course of the attack rate in the three cohorts is depicted in Figure 4. The median attack rate in the repellent group became zero already in week 2 and remained so until the end of the study. In contrast, in the shoe cohort the median attack rate started to decrease slightly in week 6. Only in week 10, it was significantly lower as compared to the control cohort (p = 0.049). In the control group, the attack rate varied over time.

The time course of SSAT is shown in Figure 5. At baseline, this indicator of acute tungiasis-associated morbidity was rather low in the three cohorts: median 4 (IQR 3–7), 3 (IQR 2–7) and 3 (IQR 1–7) in the control, the shoe and the repellent cohort, respectively (Table 2). Whereas in the repellent cohort the absolute difference to the baseline value became already significant in week 2 (p<0.001 compared to the control group), in the shoe cohort the absolute difference of the SSAT to the baseline value remained insignificant during all follow ups.

In all cohorts, the SSCT for chronic tungiasis was very low: median 1 (IQR 0–2), 2 (IQR 0–3.5) and 2 (IQR 0–3) in the control, the shoe and the repellent cohort, respectively (Table 2). In the repellent and the shoe cohort, a slight reduction became obvious with time. However, these differences were not significant.

To compare the efficacy of the application of the repellent with the protective effect of the availability of closed shoes, we compared the four major outcome measures at week 10. Whereas in the repellent group the intensity of infestation had decreased to a median of 3 (IQR 1–13), in the shoe group this outcome measure remained high: median 15 (IQR 6–36; p = 0.001). The attack rate was zero in the repellent group (IQR 0-0), and 2 in the shoe group (IQR 0–4.5; p = 0.03). The SSAT showed a similar tendency: median 1 (IQR 0–1) versus median 3 (IQR 1–4; p<0.0001). No significant difference between the repellent and the shoe group existed with regard to the SSCT (p = 0.17).

Of the 5.042 applications of Zanzarin foreseen in the repellent group, 4.832 (96%) had been applied as foreseen in the protocol. A per protocol analysis was, therefore, not considered to be meaningful in this group. Wearing shoes in the shoe group was assessed during 62 days. Fourteen participants wore shoes never/seldomly, 24 irregularly and 25 regularly; 14 were not classified, since the number of observations was too low to permit a conclusion. The average compliance was 51.6% (range 0%–98%). Stratification of members of the shoe group into individuals who never/seldomly, irregularly, or regularly wore the donated shoes is shown in Table 3. The intensity of infestation decreased continuously in the subgroup of participants wearing the shoes regularly: median 19 (IQR 11.5–29) at baseline versus median 9 (IQR 4–20) at week 10 (p = 0.03). In the subgroup of people wearing shoes irregularly the decrease was less obvious: median at baseline 18.5 (IQR 9.5–28) versus median 11.5 (IQR 5.75–28.5) at week 10 (p = 0.72). In contrast, individuals who never or rarely wore their shoes remained infested to the same degree during the entire observation period. Similarly, the attack rate decreased in participants of the two former groups, but not in individuals who never/rarely wore shoes (Table 3).

At baseline, there was inverse relationship between the intensity of infestation and wearing of shoes later: median = 24.5 lesions (IQR 16–52) in participants who later wore their shoes seldomly, median = 18.5 (IQR 10–28) in participants who later wore their shoes irregularly and median = 16 (13–29) in participants who later wore their shoes regularly.

Figure 6, 7 and 8 show a photo series of typical clinical pathology presentations at the soles of individuals from the control, shoe and repellent cohort. The type and degree of clinical pathology in the individual of the control cohort remained essentially the same during the observation period: several lesions disappeared and new lesions developed, and the soles remained heavily inflamed (Figure 6 A, B, C, D). In the individual of the shoe cohort, a slight amelioration was apparent onwards from week 6. However, at week 10, signs of inflammation still persisted (Figure 7 A, B, C, D). In contrast, in the individual of the repellent group, a reduction of tungiasis-related inflammation became apparent already at week 3. At week 10, the sole of this individual only showed residues of sand flea lesions without signs of inflammation (Figure 8 A, B, C, D).