Comments by Reviewer 1,
This in an interesting and relevant report from a study done under the difficult circumstances posed by the ongoing conflict in Somalia. The researchers -opportunistically- took advantage of a broader nutrition survey to tag on a question on the use of bednets and collection of a finger prick blood sample. The objective was to study effectiveness of bednets against malaria infection in this low transmission setting.
Despite the inevitable limitations this difficult area poses to research, the authors make a reasonable claim that the effectiveness of the bednets is relatively high, followed by a good discussion why this could be the case (in the understudied low transmission zones in Africa), together with a number of cautions as to what this should practically imply in terms of distibution of bednets in this kind of environments.
The paper has a number of suggestions for continued research and should be of interest to both malaria researchers as to those involved in making allocative decisions (like wide-scale distribution of bednets) under these kind of conditions.
Overall the paper is well written, but the description of the used methodology could be improved, along with attention to some minor detail.
We thank the reviewer for his comments and recognition of the difficulties of working in Somalia. Our responses to his more specific questions follow.
1. The sampling process is not fully clear. The sampling process is initially described as a 30x30 cluster design for 11 zones. Clusters are not selected proportionate to size, but apparently later on a weighting has been applied. This could be stated more explicitly together with a statement as to the perceived quality of the used sampling frame. Based on 30 clusters x 30 households x 11 zones, one expects around 300 clusters and around 9,900 households to be included. It is not clear how individuals have been selected and what is meant with the sentence on 'approximately 50 people' that have been examined per cluster. The report reports data from 197 clusters with 10,500 individuals, and the discrepancy with the expected 300 clusters may need more explanation.
Only 7 zones were surveyed and not 11. Lower and Middle Shabelle were combined as one region with two zones (agro-pastoralist and riverine zones). Bay had pastoralist and agro-pastoralist zones; while Gedo had all three zones (pastoralist, agro-pastoralist and riverine). This makes a total of 7 zones and a sample size of 210 clusters. However only 201 could be located in the survey, the remaining 9 clusters (all pastoralists) could not be located during field work due to the nomadic nature of such groups. Of the 201 surveyed clusters 4 were removed from analysis due to un-interpretable RDT and ITN responses leaving us with 197 clusters. In most clusters the minimum sample size of 50 were surveyed, in 57 clusters however more than the minimum of 50 were sampled resulting greater than the expected number of individuals examined for parasitemia. This information is now presented in the methods and results sections.
2. The introduction talks about 'ministries of health'. It is not self-evident what is meant with these 'ministries'. Is this the term used for health authorities at some sub-national level?
Since the fall of the central government in 1989, the country has been divided into three self-declared states. South-Central; Puntland; and Somaliland. All have ministries of health which have to be engaged in any health-related work even though none of these states are internationally recognized. In addition there is currently a Ministry of Health of the transitional federal government recognized internationally but in reality has authority only over South-Central and Puntland through a peace charter not subscribed to by Somaliland. These are the ministries of health we refer to in the paper.
3. In discussion: Why do authors suspect that the majority of used bednets were LLIN ?
As described in the text, a multiple indicator cluster survey (MICS) just a few months before the study showed that 88% of all nets in use were LLIN most of which were provided by UNICEF-supported programmes all of which strictly provide LLINs. Reference number 18 is the relevant source of this information.
4. In discussion: sentence 'The meta-analysis ... only 13%' may need an edit around the word 'variously'. And next sentence misses an 'as' after 'largely'
These changes have been made.
5. The figure would benefit from a small insert locating this part of the country to known entities like Mogadishu / Nairobi.
Position of Mogadishu now marked on Figure 1
Reviewer #2 (Remarks for the Author):
This paper is very well written. The motivation for the study is clear. The experimental design is appropriate and the statistics applied defendable. The results are interesting and a first in a low endemicity Anopheles arabiensis dominated malaria transmission zone. They results are discussed thoroughly and positioned well within the body of existing literature. The questions these results raise for ITN distribution policy in low malaria endemicity environments are considered in a balanced fashion.
We thank the reviewer for these observations.
The only omission is that the following references should be cited. They find a similar paradoxical relationship with prevalence for riparian communities, although in this case largely explained by net usage. How this differed from what was found here should be very outlined briefly.
Thomson M, Connor S, Bennett S, DAlessandro U, Milligan P, et al. (1996) Geographical perspectives on bednet use and malaria transmission in the Gambia, west Africa. Social Science and Medicine 43: 101-112.
Thomson MC, Dalessandro U, Bennett S, Connor SJ, Langerock P, et al. (1994) Malaria prevalence is inversely related to vector density in the Gambia, West Africa. Trans Roy Soc Trop Med Hyg 88: 638-643.
We thank the reviewer for pointing this out. We have now included these references (No. 22 and 23) in the discussion and in the reference list.