Background: The hepatitis C virus (HCV) is hyper-endemic in injecting drug users. There is also excess HCV among non-injection drug users who smoke, snort, or sniff heroin, cocaine, crack, or methamphetamine. Methods: To summarize the research literature on HCV in drug users and identify gaps in knowledge, we conducted a synthesis of the relevant research carried out between 1989 and 2006. Using rigorous search methods, we identified and extracted data from published and unpublished reports of HCV among drug users. We designed a quality assurance system to ensure accuracy and consistency in all phases of the project. We also created a set of items to assess study design quality in each of the reports we included. Results: We identified 629 reports containing HCV prevalence rates, incidence rates and/or genotype distribution among injecting or non-injecting drug user populations published between January 1989 and December 2006. The majority of reports were from Western Europe (41%), North America (26%), Asia (11%) and Australia/New Zealand (10%). We also identified reports from Eastern Europe, South America, the Middle East, and the Caribbean. The number of publications reporting HCV rates in drug users increased dramatically between 1989 and 2006 to 27-52 reports per year after 1998. Conclusions: The data collection and quality assurance phases of the HCV Synthesis Project have been completed. Recommendations for future research on HCV in drug users have come out of our data collection phase. Future research reports can enhance their contributions to our understanding of HCV etiology by clearly defining their drug user participants with respect to type of drug and route of administration. Further, the use of standard reporting methods for risk factors would enable data to be combined across a larger set of studies; this is especially important for HCV seroconversion studies which suffer from small sample sizes and low power to examine risk factors.

Background: Maintaining data quality and integrity is important for research studies involving prospective data collection. Data must be entered, erroneous or missing data must be identified and corrected if possible, and an audit trail created. Methods: Using as an example a large prospective study, the Missouri Lower Respiratory Infection (LRI) Project, we present an approach to data management predominantly using SAS software. The Missouri LRI Project was a prospective cohort study of nursing home residents who developed an LRI. Subjects were enrolled, data collected, and follow-ups occurred for over three years. Data were collected on twenty different forms. Forms were inspected visually and sent off-site for data entry. SAS software was used to read the entered data files, check for potential errors, apply corrections to data sets, and combine batches into analytic data sets. The data management procedures are described. Results: Study data collection resulted in over 20,000 completed forms. Data management was successful, resulting in clean, internally consistent data sets for analysis. The amount of time required for data management was substantially underestimated. Conclusions: Data management for prospective studies should be planned well in advance of data collection. An ongoing process with data entered and checked as they become available allows timely recovery of errors and missing data.

Background: Studies have shown that industry-sponsored meta-analyses of drugs lack scientific rigour and have biased conclusions. However, these studies have been restricted to certain medical specialities. We compared all industry-supported meta-analyses of drug-drug comparisons with those without industry support. Methods: We searched PubMed for all meta-analyses that compared different drugs or classes of drugs published in 2004. Two authors assessed the meta-analyses and independently extracted data. We used a validated scale for judging the methodological quality and a binary scale for judging conclusions. We divided the meta-analyses according to the type of support in 3 categories: industry-supported, non-profit support or no support, and undeclared support. Results: We included 39 meta-analyses. Ten had industry support, 18 non-profit or no support, and 11 undeclared support. On a 0-7 scale, the median quality score was 6 for meta-analyses with non-profit or no support and 2.5 for the industry-supported meta-analyses (P < 0.01). Compared with industry-supported meta-analyses, more meta-analyses with non-profit or no support avoided bias in the selection of studies (P = 0.01), more often stated the search methods used to find studies (P = 0.02), searched comprehensively (P < 0.01), reported criteria for assessing the validity of the studies (P = 0.02), used appropriate criteria (P = 0.04), described methods of allocation concealment (P = 0.05), described methods of blinding (P = 0.05), and described excluded patients (P = 0.08) and studies (P = 0.15). Forty percent of the industry-supported meta-analyses recommended the experimental drug without reservations, compared with 22% of the meta-analyses with non-profit or no support (P = 0.57). In a sensitivity analysis, we contacted the authors of the meta-analyses with undeclared support. Eight who replied that they had not received industry funding were added to those with non-profit or no support, and 3 who did not reply were added to those with industry support. This analysis did not change the results much. Conclusions: Transparency is essential for readers to make their own judgment about medical interventions guided by the results of meta-analyses. We found that industry-supported meta-analyses are less transparent than meta-analyses with non-profit support or no support.

Background: Studies have shown that industry-sponsored meta-analyses of drugs lack scientific rigour and have biased conclusions. However, these studies have been restricted to certain medical specialities. We compared all industry-supported meta-analyses of drug-drug comparisons with those without industry support. Methods: We searched PubMed for all meta-analyses that compared different drugs or classes of drugs published in 2004. Two authors assessed the meta-analyses and independently extracted data. We used a validated scale for judging the methodological quality and a binary scale for judging conclusions. We divided the meta-analyses according to the type of support in 3 categories: industry-supported, non-profit support or no support, and undeclared support. Results: We included 39 meta-analyses. Ten had industry support, 18 non-profit or no support, and 11 undeclared support. On a 0-7 scale, the median quality score was 6 for meta-analyses with non-profit or no support and 2.5 for the industry-supported meta-analyses (P < 0.01). Compared with industry-supported meta-analyses, more meta-analyses with non-profit or no support avoided bias in the selection of studies (P = 0.01), more often stated the search methods used to find studies (P = 0.02), searched comprehensively (P < 0.01), reported criteria for assessing the validity of the studies (P = 0.02), used appropriate criteria (P = 0.04), described methods of allocation concealment (P = 0.05), described methods of blinding (P = 0.05), and described excluded patients (P = 0.08) and studies (P = 0.15). Forty percent of the industry-supported meta-analyses recommended the experimental drug without reservations, compared with 22% of the meta-analyses with non-profit or no support (P = 0.57). In a sensitivity analysis, we contacted the authors of the meta-analyses with undeclared support. Eight who replied that they had not received industry funding were added to those with non-profit or no support, and 3 who did not reply were added to those with industry support. This analysis did not change the results much. Conclusions: Transparency is essential for readers to make their own judgment about medical interventions guided by the results of meta-analyses. We found that industry-supported meta-analyses are less transparent than meta-analyses with non-profit support or no support.

Background: Body mass index (BMI) data usually have skewed distributions, for which common statistical modeling approaches such as simple linear or logistic regression have limitations. Methods: Different regression approaches to predict childhood BMI by goodness-of-fit measures and means of interpretation were compared including generalized linear models (GLMs), quantile regression and Generalized Additive Models for Location, Scale and Shape (GAMLSS). We analyzed data of 4967 children participating in the school entry health examination in Bavaria, Germany, from 2001 to 2002. TV watching, meal frequency, breastfeeding, smoking in pregnancy, maternal obesity, parental social class and weight gain in the first 2 years of life were considered as risk factors for obesity. Results: GAMLSS showed a much better fit regarding the estimation of risk factors effects on transformed and untransformed BMI data than common GLMs with respect to the generalized Akaike information criterion. In comparison with GAMLSS, quantile regression allowed for additional interpretation of prespecified distribution quantiles, such as quantiles referring to overweight or obesity. The variables TV watching, maternal BMI and weight gain in the first 2 years were directly, and meal frequency was inversely significantly associated with body composition in any model type examined. In contrast, smoking in pregnancy was not directly, and breastfeeding and parental social class were not inversely significantly associated with body composition in GLM models, but in GAMLSS and partly in quantile regression models. Risk factor specific BMI percentile curves could be estimated from GAMLSS and quantile regression models. Conclusions: GAMLSS and quantile regression seem to be more appropriate than common GLMs for risk factor modeling of BMI data.

Background: Body mass index (BMI) data usually have skewed distributions, for which common statistical modeling approaches such as simple linear or logistic regression have limitations. Methods: Different regression approaches to predict childhood BMI by goodness-of-fit measures and means of interpretation were compared including generalized linear models (GLMs), quantile regression and Generalized Additive Models for Location, Scale and Shape (GAMLSS). We analyzed data of 4967 children participating in the school entry health examination in Bavaria, Germany, from 2001 to 2002. TV watching, meal frequency, breastfeeding, smoking in pregnancy, maternal obesity, parental social class and weight gain in the first 2 years of life were considered as risk factors for obesity. Results: GAMLSS showed a much better fit regarding the estimation of risk factors effects on transformed and untransformed BMI data than common GLMs with respect to the generalized Akaike information criterion. In comparison with GAMLSS, quantile regression allowed for additional interpretation of prespecified distribution quantiles, such as quantiles referring to overweight or obesity. The variables TV watching, maternal BMI and weight gain in the first 2 years were directly, and meal frequency was inversely significantly associated with body composition in any model type examined. In contrast, smoking in pregnancy was not directly, and breastfeeding and parental social class were not inversely significantly associated with body composition in GLM models, but in GAMLSS and partly in quantile regression models. Risk factor specific BMI percentile curves could be estimated from GAMLSS and quantile regression models. Conclusions: GAMLSS and quantile regression seem to be more appropriate than common GLMs for risk factor modeling of BMI data.

Background: The beta-binomial model is one of the methods that can be used to validly combine event rates from overdispersed binomial data. Our objective is to provide a full description of this method and to update and broaden its applications in clinical and public health research. Methods: We describe the statistical theories behind the beta-binomial model and the associated estimation methods. We supply information about statistical software that can provide beta-binomial estimations. Using a published example, we illustrate the application of the beta-binomial model when pooling overdispersed binomial data. Results: In an example regarding the safety of oral antifungal treatments, we had 41 treatment arms with event rates varying from 0% to 13.89%. Using the beta-binomial model, we obtained a summary event rate of 3.44% with a standard error of 0.59%. The parameters of the beta-binomial model took the values of 1.24 for alpha and 34.73 for beta. Conclusions: The beta-binomial model can provide a robust estimate for the summary event rate by pooling overdispersed binomial data from different studies. The explanation of the method and the demonstration of its applications should help researchers incorporate the beta-binomial method as they aggregate probabilities of events from heterogeneous studies.

Background: Some populations targeted in survey research can be hard to reach, either because of lack of contact information, or non-existent databases to inform sampling. Here, we present a methodological "case-report" of the yield of a multi-step survey study assessing views on health care among American emigres to Canada, a hard-to-reach population. Methods: To sample this hard-to-reach population, we held a live media conference, supplemented by a nation-wide media release announcing the study. We prepared an 'op-ed' piece describing the study and how to participate. We paid for advertisements in 6 newspapers. We sent the survey information to targeted organizations. And lastly, we asked those who completed the web survey to send the information to others. We use descriptive statistics to document the method's yield. Results: The combined media strategies led to 4 television news interviews, 10 newspaper stories, 1 editorial and 2 radio interviews. 458 unique individuals accessed the on-line survey, among whom 310 eligible subjects provided responses to the key study questions. Fifty-six percent reported that they became aware of the survey via media outlets, 26 percent by word of mouth, and 9 percent through both the media and word of mouth. Conclusions: Our multi-step communication method yielded a sufficient sample of Americans living in Canada. This combination of paid and unpaid media exposure can be considered by others as a unique methodological approach to identifying and sampling hard-to-reach populations.

Background: Results from clinical trials are usually summarized in the form of sampling distributions. When full information (mean, SEM) about these distributions is given, performing meta-analysis is straightforward. However, when some of the sampling distributions only have mean values, a challenging issue is to decide how to use such distributions in meta-analysis. Currently, the most common approaches are either ignoring such trials or for each trial with a missing SEM, finding a similar trial and taking its SEM value as the missing SEM. Both approaches have drawbacks. As an alternative, this paper develops and tests two new methods, the first being the prognostic method and the second being the interval method, to estimate any missing SEMs from a set of sampling distributions with full information. A merging method is also proposed to handle clinical trials with partial information to simulate meta-analysis. Methods: Both of our methods use the assumption that the samples for which the sampling distributions will be merged are randomly selected from the same population. In the prognostic method, we predict the missing SEMs from the given SEMs. In the interval method, we define intervals that we believe will contain the missing SEMs and then we use these intervals in the merging process. Results: Two sets of clinical trials are used to verify our methods. One family of trials is on comparing different drugs for reduction of low density lipprotein cholesterol (LDL) for Type-2 diabetes, and the other is about the effectiveness of drugs for lowering intraocular pressure (IOP). Both methods are shown to be useful for approximating the conventional meta-analysis including trials with incomplete information. For example, the meta-analysis result of Latanoprost versus Timolol on IOP reduction for six months provided in [1] was 5:05+/-1.15 (Mean+/-SEM) with full information. If the last trial in this study is assumed to be with partial information, the traditional analysis method for dealing with incomplete information that ignores this trial would give 6:49 +/- 1:36 while our prognostic method gives 5:02+/-1.15, and our interval method provides two intervals as Mean [4:25; 5:63] and SEM [1:01; 1:24]. Conclusions: Both the prognostic and the interval methods are useful alternatives for dealing with missing data in meta-analysis. We recommend clinicians to use the prognostic method to predict the missing SEMs in order to perform meta-analysis and the interval method for obtaining a more cautious result.

Background: The paper examines the key issues experienced in recruiting and retaining practice involvement in a large complex intervention trial in Australian General Practice Methods: Reflective notes made by research staff and telephone interviews with staff from general practices which expressed interest, took part or withdrew from a trial of a complex general practice intervention. Results: Recruitment and retention difficulties were due to factors inherent in the demands and context of general practice, the degree of engagement of primary care organisations (Divisions of General Practice), perceived benefits by practices, the design of the trial and the timing and complexity of data collection, Conclusions: There needs to be clearer articulation to practices of the benefits of the research to participants and streamlining of the design and processes of data collection and intervention to fit in with their work practices. Ultimately deeper engagement may require additional funding and ongoing participation through practice research networks. Current Controlled Trials ACTRN12605000788673

Background: Measurement of socioeconomic position (SEP) over the life course in population health surveillance systems is important for examining differences in health and illness between different population groups and for monitoring the impact of policies and interventions aimed at reducing health inequities and intergenerational disadvantage over time. While face-to-face surveys are considered the gold standard of interviewing techniques, computer-assisted telephone interviewing is often preferred for cost and convenience. This study compared recall of parents' highest level of education in telephone and face-to-face surveys. Methods: Questions about father's and mother's highest education level were included in two representative population health surveys of South Australians aged 18 years and over in Spring 2004. A random sample selected from the electronic white pages (EWP) responded to a computer-assisted telephone interview (n=2999), and a multistage clustered area sample responded to a face-to-face interview (n=2893). A subsample of respondents in the face-to-face sample who owned a telephone that was listed in the EWP (n=2206) was also compared to the telephone interview sample. Results: The proportion of respondents who provided information about their father's and mother's highest education level was significantly higher in the face-to-face interview (86.3% and 87.8%, respectively) than in the telephone interview (80.4% and 79.9%, respectively). Recall was also significantly higher in the subsample of respondents in the face-to-face interview who had a telephone that was listed in the EWP. Those with missing data for parents' education were more likely to be socioeconomically disadvantaged regardless of the survey mode. Conclusions: While face-to-face interviewing obtained higher item response rates for questions about parents' education, survey mode did not appear to influence the factors associated with having missing data on father's or mother's highest education level.

Background: This paper focuses on measuring the efficiency and effectiveness of two diagramming methods employed in key informant interviews with clinicians and health care administrators. The two methods are 'participatory diagramming', where the respondent creates a diagram that assists in their communication of answers, and 'graphic elicitation', where a researcher-prepared diagram is used to stimulate data collection. Methods: These two diagramming methods were applied in key informant interviews and their value in efficiently and effectively gathering data was assessed based on quantitative measures and qualitative observations. Results: Assessment of the two diagramming methods suggests that participatory diagramming is an efficient method for collecting data in graphic form, but may not generate the depth of verbal response that many qualitative researchers seek. In contrast, graphic elicitation was more intuitive, better understood and preferred by most respondents, and often provided more contemplative verbal responses, however this was achieved at the expense of more interview time. Conclusions: Diagramming methods are important for eliciting interview data that are often difficult to obtain through traditional verbal exchanges. Subject to the methodological limitations of the study, our findings suggest that while participatory diagramming and graphic elicitation have specific strengths and weaknesses, their combined use can provide complementary information that would not likely occur with the application of only one diagramming method. The methodological insights gained by examining the efficiency and effectiveness of these diagramming methods in our study should be helpful to other researchers considering their incorporation into qualitative research designs.

Background: Regression to the mean (RTM) occurs in situations of repeated measurements when extreme values are followed by measurements in the same subjects that are closer to the mean of the basic population. In uncontrolled studies such changes are likely to be interpreted as a real treatment effect. Methods: Several statistical approaches have been developed to analyse such situations, including the algorithm of Mee and Chua which assumes a known population mean mu. We extend this approach to a situation where mu is unknown and suggest to vary it systematically over a range of reasonable values. Using differential calculus we provide formulas to estimate the range of mu where treatment effects are likely to occur when RTM is present. Results: We successfully applied our method to three real world examples denoting situations when (a) no treatment effect can be confirmed regardless which mu is true, (b) when a treatment effect must be assumed independent from the true mu and (c) in the appraisal of results of uncontrolled studies. Conclusions: Our method can be used to separate the wheat from the chaff in situations, when one has to interpret the results of uncontrolled studies. In meta-analysis, health-technology reports or systematic reviews this approach may be helpful to clarify the evidence given from uncontrolled observational studies.

Background: Preventive drugs require long-term trials to show their effectiveness or harms and often a lot of changes occur during post-marketing studies. The purpose of this article is to describe the research process in a long-term randomized controlled trial and discuss the impact and consequences of changes in the research environment. Methods: The Estonian Postmenopausal Hormone Therapy trial (EPHT), originally planned to continue for five years, was planned in co-operation with the Women's International Study of Long-Duration Oestrogen after Menopause (WISDOM) in the UK. In addition to health outcomes, EPHT was specifically designed to study the impact of postmenopausal hormone therapy (HT) on health services utilization. Results: After EPHT recruited in 1999-2001 the Women's Health Initiative (WHI) in the USA decided to stop the estrogen-progestin trial after a mean of 5.2 years in July 2002 because of increased risk of breast cancer and later in 2004 the estrogen-only trial because HT increased the risk of stroke, decreased the risk of hip fracture, and did not affect coronary heart disease incidence. WISDOM was halted in autumn 2002. These decisions had a major influence on EPHT. Conclusions: Changes in Estonian society challenged EPHT to find a balance between the needs of achieving responses to the trial aims with a limited budget and simultaneously maintaining the safety of trial participants. Flexibility was the main key for success. Rapid changes are not limited only to transiting societies but are true also in developed countries and the risk must be included in planning all long-term trials. The role of ethical and data monitoring committees in situations with emerging new data from other studies needs specification. Longer funding for preventive trials and more flexibility in budgeting are mandatory. Who should prove the effectiveness of an (old) drug for a new preventive indication? In preventive drug trials companies may donate drugs but they take a financial risk, especially with licensed drugs. Public funding is crucial to avoid commercial biases. Legislation to share the costs of large post-marketing trials as well as regulation of manufacturer's participation is needed. [ISRCTN35338757]

Background Vignette studies of medical choice and judgement have gained popularity in the medical literature. Originally developed in mathematical psychology they can be used to evaluate physicians' behaviour in the setting of diagnostic testing or treatment decisions. Objectives To provide an overview of the use, objectives and methodology of these studies in the medical field Design Systematic review. Data sources Electronic databases; reference lists of included studies. Review methods We included studies that examined medical decisions of physicians, nurses or medical students using cue weightings from answers to structured vignettes. Two reviewers scrutinized abstracts and examined full text copies of potentially eligible studies. The aim of the included studies, the type of clinical decision, the number of participants, some technical aspects, and the type of statistical analysis were extracted in duplicate and discrepancies were resolved by consensus. Results 30 reports published between 1983 and 2005 fulfilled the inclusion criteria. 22 studies (73%) reported on treatment decisions and 27 (90%) explored the variation of decisions among experts. Nine studies (30%) described differences in decisions between groups of caregivers and ten studies (33%) described the decision behaviour of only one group. Only six studies (20%) compared decision behaviour against an empirical reference of a correct decision. The median number of considered attributes was 6.5 (IQR 4-9), the median number of vignettes was 27 (IQR 16-40). In 17 studies, decision makers had to rate the relative importance of a given vignette; in six studies they had to assign a probability to each vignette. Only ten studies (33%) applied a statistical procedure to account for correlated data. Conclusion Various studies of medical choice and judgement have been performed to depict weightings of the value of clinical information from answers to structured vignettes of caregivers. We found that the design and analysis methods used in current applications vary considerably and could be improved in a large number of cases.

Background: Investigators are actively testing interventions intended to increase lifespan and wish to test whether the interventions increase maximum lifespan. Based on the fact that one cannot be assured of observing population maximum lifespans in finite samples, in previous work, we constructed and validated several tests of difference in the upper parts of lifespan distributions between a treatment group and a control group by testing whether the probabilities that observations are above some threshold defining 'old' or being in the tail of the survival distribution are equal in the two groups. However, a limitation of these tests is that they do not consider how much above the threshold any particular observation is. Methods: In this article we propose new methods which improve upon our previous tests by considering not only whether an observation is above some threshold, but also the magnitudes by which observations exceed the threshold. Results: Simulations show that the new methods control type I error rates quite well and that the power of the new methods is usually higher than that of the tests we previously proposed. In illustrative analyses of two real datasets involving rodents, when setting the threshold equal to 110 (100) weeks for the first (second) datasets, the new methods detected differences in 'maximum lifespan' between groups at nominal alpha levels of 0.01 (0.05) for the first (second) datasets and provided more significant results than competitor tests. Conclusions: The new methods not only have good performance in controlling the type I error rates but also improve the power compared with the tests we previously proposed.

Background: Despite the benefits, it is uncommon to apply the nested case-control design in diagnostic research. We aim to show advantages of this design for diagnostic accuracy studies. Methods: We used data from a full cross-sectional diagnostic study comprising a cohort of 1295 consecutive patients who were selected on their suspicion of having deep vein thrombosis (DVT). We draw nested case-control samples from the full study population with case:control ratios of 1:1, 1:2, 1:3 and 1:4 (per ratio 100 samples were taken). We calculated diagnostic accuracy estimates for two tests that are used to detect DVT in clinical practice. Results: Estimates of diagnostic accuracy in the nested case-control samples were very similar to those in the full study population. For example, for each case:control ratio, the positive predictive value of the D-dimer test was 0.30 in the full study population and 0.30 in the nested case-control samples (median of the 100 samples). As expected, variability of the estimates decreased with increasing sample size. Conclusions: Our findings support the view that the nested case-control study is a valid and efficient design for diagnostic studies and should also be (re)appraised in current guidelines on diagnostic accuracy research.

Background: To create and find accurate and reliable instruments for the measurement of physical activity has been a challenge in epidemiological studies. We investigated the reliability and validity of two different physical activity questionnaires in 71 adolescents aged 13-18 years; the WHO, Health Behaviour in Schoolchildren (HBSC) questionnaire, and the International Physical Activity Questionnaire (IPAQ, short version). Methods: The questionnaires were administered twice (8-12 days apart) to measure reliability. Validity was assessed by comparing answers from the questionnaires with a cardiorespiratory fitness test (VO2peak) and seven days activity monitoring with the ActiReg, an instrument measuring physical activity level (PAL) and total energy expenditure (TEE). Results: Intraclass correlation coefficients for reliability for the WHO HBSC questionnaire were 0.71 for frequency and 0.73 for duration. For the frequency question, there was a significant difference between genders; 0.87 for girls and 0.59 for boys (p<0.05). The intraclass correlation coefficients the IPAQ varied between 0.10 and 0.62 for the reliability. Spearman correlation coefficients for validity for both the WHO HBSC questionnaire and the IPAQ (recoded into low, moderate and high activity) measured against VO2peak were fair, ranging between 0.29 - 0.39. The WHO HBSC questionnaire measured against VO2peak for girls were acceptable, ranging between 0.30 - 0.55. Both questionnaires, except the walking question in IPAQ, showed a low correlation with PAL and TEE, ranging between 0.01 and 0.29. Conclusion: These data indicate that the WHO HBSC questionnaire had substantial reliability and were acceptable instrument for measuring cardiorespiratory fitness, especially among girls. None of the questionnaires however seemed to be a valid instrument for measuring physical activity compared to TEE and PAL in adolescents.

Background: In trials designed to estimate rates of perinatal mother to child transmission of HIV, HIV assays are scheduled at multiple points in time. Still, infection status for some infants at some time points may be unknown, particularly when interim analyses are conducted. Methods: Logistic regression models are commonly used to estimate covariate-adjusted transmission rates, but their methods for handling missing data may be inadequate. Here we propose using coarsened multinomial regression models to estimate cumulative and conditional rates of HIV transmission. Through simulation, we compare the proposed models to standard logistic models in terms of bias, mean squared error, coverage probability, and power. We consider a range of treatment effect and visit process scenarios, while including imperfect sensitivity of the assay and contamination of the endpoint due to early breastfeeding transmission. We illustrate the approach through analysis of data from a clinical trial designed to prevent perinatal transmission. Results: The proposed cumulative and conditional models performed well when compared to their logistic counterparts. Performance of the proposed cumulative model was particularly strong under scenarios where treatment was assumed to increase the risk of in utero transmission but decrease the risk of intrapartum and overall perinatal transmission and under scenarios designed to represent interim analyses. Power to estimate intrapartum and perinatal transmission was consistently higher for the proposed models. Conclusions: Coarsened multinomial regression models are preferred to standard logistic models for estimation of perinatal mother to child transmission of HIV, particularly when assays are missing or occur off-schedule for some infants.

Background: There is a growing recognition of the value of synthesising qualitative research in the evidence base in order to facilitate effective and appropriate health care. In response to this, methods for undertaking these syntheses are currently being developed. Thematic analysis is a method that is often used to analyse data in primary qualitative research. This paper reports on the use of this type of analysis in systematic reviews to bring together and integrate the findings of multiple qualitative studies. Methods: We describe thematic synthesis, outline several steps for its conduct and illustrate the process and outcome of this approach using a completed review of research on the promotion of healthy eating amongst children. Thematic synthesis has three stages: the coding of text 'line-by-line'; the development of 'descriptive themes'; and the generation of 'analytical themes'. While the development of descriptive themes remains 'close' to the primary studies, the analytical themes represent a stage of interpretation whereby the reviewers 'go beyond' the primary studies and generate new interpretive constructs, explanations or hypotheses. The use of computer software can facilitate this method of synthesis; detailed guidance is given on how this can be achieved. Results: We used thematic synthesis to combine studies of children's views and identified key themes to explore in the intervention studies. Most interventions were based in school and often combined learning about health benefits with 'hands-on' experience. The studies of children's views suggested that fruit and vegetables should be treated in different ways, and that messages should not focus on health warnings. Interventions that were in line with these suggestions tended to be more effective. Thematic synthesis enabled us to stay 'close' to the results of the primary studies, synthesising them in a transparent way, and facilitating the explicit production of new concepts and hypotheses. Conclusion: We compare thematic synthesis to other methods for the synthesis of qualitative research, discussing issues of context and rigour. Thematic synthesis is presented as a tried and tested method that preserves an explicit and transparent link between conclusions and the text of primary studies; as such it preserves principles that have traditionally been important to systematic reviewing.