Main findings from systematic review –Campylobacter
Study selection. Only 74 of a total 1467 identified papers contained primary research from published journals, with the vast majority of identified papers presenting commentary on other issues or results presented elsewhere. Primary data was found to be particularly lacking for intervention studies, with only five presenting primary research. For search B, of 456 websites reviewed, only eight documents were finally selected in addition to those identified from search A.
Characteristics of the studies. The review allowed certain trends to be analysed in the areas of country of origin, language of publication and most prolific year for publications. The finding from the review that around 25% of relevant studies had been performed in the USA may be a reflection of the size of the poultry industry in that country. In 2002, the US market share for world broiler production was 27% (USDA 2004). Although approximately 58% of studies were completed in Europe (European Union (EU), Switzerland and Norway), poultry production in the EU for 2002 amounted to only 13% of the world total (USDA 2004).
Considering the fact that scientific journals with the largest impact are written in English, it is therefore unsurprising that 93% of relevant publications were written in English. The review also revealed that there were variations between studies on sampling and testing methods, such that no two studies were the same. Unfortunately, the study sample sizes at the individual category level were too small to investigate whether differences in method were the cause of any conflicting results.
Common foci of studies. It was found that many factors are frequently tested for their association with Campylobacter occurrence, whilst others were rarely mentioned in the papers. The frequency with which vertical transmission has been selected for investigations may reflect the initial assumption that Campylobacter epidemiology was similar to that previously found for Salmonella, where vertical transmission is an important factor (Newell and Fearnley 2003). Or it may be an indication of research into the importance of transfer mechanisms that enable control strategies to be implemented, such as increased biosecurity.
Since vertical transmission has been frequently investigated and not emerged as a major source of Campylobacter transmission, limited funding resources may be better spent investigating the contribution of other sources more rarely investigated, such as the presence of surface waste (puddles) outside broiler houses.
Most relevant factors and sources of Campylobacter infection in GB broilers. It is interesting to note that the top three sources and contributing factors to Campylobacter occurrence are self-supporting, with a paired source and factor concerning (i) movement on farm during depopulation, (ii) between broiler houses and (iii) transfer by staff. In addition, the outcomes of this systematic review are in agreement with those from narrative reviews, for example, Newell and Fearnley (2003).
These three identified variables may be priorities included in further fieldwork. However, those contributing factors and sources with negative scores, such as routines of removing dead birds, feed and litter, have been found here to have the least bearing on Campylobacter occurrence for GB broilers, and therefore should not be the primary focus of any trials.
Analysis of agreement and conflict in the data. Several of the results found by this analysis were not surprising. Both feed and litter were unanimously found to have no effect on Campylobacter presence. Vertical transmission, which was found to have one of the highest disagreement scores, is an acknowledged area of substantial differences in opinion in this field of research (Newell and Fearnley 2003).
Gaps in knowledge. Quantitative information was absent on the level of Campylobacter contamination associated with the various potential sources. An advantage of this review is that it is attached to a national scale survey that has currently completed its pilot phase. In this way, certain data that have been found absent from this systematic review have been requested within the framework of the experimental on-farm testing to be conducted in GB. In addition, the systematic review performed did not include the contacting of authors for raw data or further information on their current research. It has been established from discussions with experts in this field that substantial amounts of raw data are available but have not been published due to the low impact of such a publication. In response to this finding, a separate request for data and information within Europe research groups was initiated. It may be useful to compare any unpublished information that is submitted from this request to that contained in the systematic review to evaluate the amount of useful data that goes unpublished.
Evaluation of a systematic review approach
From conducting a systematic review rather than a traditional literature review, a number of advantages have been proposed and supported by the work conducted here. The approach is a reproducible and transparent method, which can be validated. The criteria for identification, selection, assessment and integration are detailed which is not the case for narrative reviews. In this review, extracts of the database (the selection phase and extraction of data) were reviewed a second time independently by a colleague not previously involved to ensure that these processes were objectively conducted. The analysis of primary studies was completed in a manner to prevent bias. Also outputs included the characteristics of publications (country of study, year of publication, language, location, testing method), most frequently studied parameters, areas of conflict and agreement in the data and the identification of knowledge gaps. An additional output of the approach is the construction of a database, which can be easily used and transferred to other people in the field.
Although the basic principles of conducting a systematic review are the same regardless of the area of research, there are specific procedures that need adapting for veterinary applications. Within veterinary science, the literature base is smaller that that of the biomedical industry where systematic reviews are commonly conducted. Therefore, the inclusion criteria taken for Campylobacter were necessarily broad, including all descriptions of contributing factors and sources of infection on-farm, as many languages as possible, were not dependent on country of origin, all types of observation or experiment design (including nonrandom farm selection). Using this approach, there were few exclusion criteria based on the quality of methods.
Another difference from traditional biomedical reviews is that meta-analysis, or the analysis of the numerical results from each selected study, was not possible due to the great heterogeneity between the papers that were included in the review. This has been a finding from other systematic reviews that have been conducted in the field of veterinary science (Le Fevre et al. 2002; Trotz-Williams and Trees 2003), which may indicate a lack of standardized reporting structure for publications and study design to enable the pooling of appropriate results. Therefore, a relevancy scoring system was developed to enable the combining of selected findings and ranking according to their applicability to GB and the internal strength of findings and level of precision.
The relevancy score approach allows the estimation of the major contributing factors and sources whilst addressing concerns over the precision, applicability and internal strength of findings of the studies included. It has been found that the type of scale used to assess trial quality can dramatically influence the interpretation of combined studies (Juni et al. 1999). A sensitivity analysis was carried out on the score which found that the removal of any one criterion did not considerably alter the baseline rank of contributing factors or sources, which infers that no one criterion dominates the score. However, the use of any type of meta-analysis or quality scoring of results has been criticized as introducing bias (Petitti 1994; Juni et al. 1999).
To investigate the importance of language bias the number of languages that contributed to the database was assessed. It was found that although translation facilities exist, very few papers considered relevant to this study had been fully published in a language other than English. In this review, efforts have been made to reduce the effect of publication and trial bias in a number of ways: (i) a nonpublished format was included – the Internet search. However, only eight documents were included from this search, compared to 151 published papers. (ii) Using the relevancy score, an average score for each factor or source was calculated to remove the potential bias from the number of publications concerning one category. However, despite all efforts to prevent bias, the quality of the results produced by this approach cannot exceed the quality of the reported information from the individual studies it contains (Shapiro 1994).
Although there are major drawbacks to conducting a systematic review, such as the considerable amount of time and labour required, there are proven advantages of conducting a systematic review as opposed to a traditional narrative review. This is the first systematic review of the major risk factors and sources of Campylobacter infection in broilers. The findings of this review are being used to assist the development of practical control measures for Campylobacter in GB broiler flocks.