Literature search
Research librarian Ingrid Harboe planned and executed all systematic searches in collaboration with the project group. The strategy included both subject headings (MeSH, Emtree) and text words. Searches were limited to systematic reviews (SR) and controlled studies, both in the time period from 1995 to the date for the search.
The reason for choosing 1995 as the starting point was that erythropoietin was in-troduced about that time (for more information about erythropoietin see Glossary and abbreviations). The searches for SRs and controlled studies were performed separately. The complete search strategies are listed in appendix 1.
We searched the following databases:
The Cochrane Library; CDSR, DARE, Central, HTA, NHS EED
Centre for Reviews and Dissemination (CRD); DARE, HTA, NHS EED
Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations and Ovid MEDLINE(R) 1946 to present
EMBASE (Ovid) 1980 to present
We checked the reference lists in systematic reviews that we reviewed in full text, and performed manual searches in the following websites:
INAHTA (International Network of Agencies for Health Technology Assessment), Clinical Evidence, ISI Web of Knowledge, NHS Evidence, AHRQ (Agency for Healthcare Research and Quality's), SBU (Swedish Council on Health Technolo-gy Assessment), Dacehta, Finohta/ THL (National Institute for Health and Wel-fare), CADTH (Canadian Agency for Drugs and Technologies in Health), AHTA (Adelaide Health Technology Assessment ), NIHR (National Institute for Health Research), and NICE (National Institute for Health and Care Excellence).
Inclusion criteria
Population Patients above 18 years with end-stage renal failure, inde-pendent of comorbidity, who need dialysis treatment, either as life-time treatment or while waiting for kidney
transplantation
26 Clinical evaluation - Methods
Interventions Hemodialysis carried out in hospital
Self-care hemodialysis carried out in hospital Hemodialysis carried out in satellite unit
Hemodialysis carried out in the patient’s home
Peritoneal dialysis (continuous ambulatory peritoneal dialysis (CAPD) and automated peritoneal dialysis (APD)) carried out at home after training in hospital
Comparator: Dependent on available data we intended to compare all interventions with the alternative interventions.
Outcomes: Mortality
Quality of life (QoL)
Complications that require special measures (i.e.
hospitalisation, antibiotic treatment) Study design: Systematic reviews
Randomized controlled trials Controlled observational studies
Languages: No limitations in languages during the search, but we
only included articles in English, articles with English abstract or articles in Scandinavian.
Selection of articles
Two reviewers independently inspected all citations generated by the search in order to identify potentially relevant articles based on title and/or abstract. Full text publi-cations were obtained for articles appearing to meet the inclusion criteria or in cases where sufficient information was not available to make a decision. Two persons in-dependently assessed whether the article was relevant or not according to our list of inclusion criteria. Disagreements were resolved by discussion or by consulting a third reviewer.
Articles meeting the pre-defined inclusion criteria were assessed for risk of bias (11).
All assessments were performed and agreed upon by two researchers.
In addition to assessing risk of bias, we also performed a close inspection of baseline characteristics in the different treatment groups included in the studies. To be certain that the estimates of efficacy or safety reflected the delivered treatments, and not different prognostic features of the patients, we examined the baseline data of patients included in the studies. If differences in comorbidity between groups were reported or detected by our own analysis; or if no description of the patients comorbidity were reported, descriptive information about the study will only be presented in an Appendix. If however the article provided analyses that adjusted for this difference, the study will be included in our analyses and our assessments. This
27 Clinical evaluation - Methods
means that the studies presented in the Appendix had either no comorbidity data reported or had adjustments that did not fulfil our study aim.
Data analysis
One reviewer extracted data from the included articles and another reviewer checked these results for accuracy. We extracted data as they were presented in the included publications. When data were presented in several ways, we chose to report data in our preferred order: hazard ratio (HR), relative risk (RR) and odds ratio (OR) with 95 % confidence intervals (CI). Where both unadjusted and adjusted data were available, we preferred adjusted data if adjustments seemed reasonable. When possible we performed meta-analyses using a random effects model. Forest plots are presented in the results section. Sometimes, when we found it helpful, we also pre-sented outcome data from a single study in forest plots.
In cases where both events and patients at risk were available from the publications for all studies for a specific outcome, RRs were calculated using the Mantel-Haenzel approach in Review Manager. When the data regarding the same outcome were re-ported in different ways in the included publications, we re-calculated to log risk ra-tios and standard error, and the common RR was calculated using inverse variance in Review Manager. Footnotes in the forest plots provide details about the original data.
Grading the quality of evidence
Two persons assessed the overall documentation for each outcome by using GRADE (Grading of Recommendations, Assessment, Development and Evaluation,
www.gradeworkinggroup.org). In the GRADE system outcome documentation from observational studies starts at low quality and outcome documentation from ran-domized controlled trials starts at high. The method involves an evaluation of study type, study quality/risk of bias, consistency between trials, directness (how similar the population, intervention, and outcomes are between the trials and the objectives of this report) and precision of the estimates. For the observational studies it is pos-sible to rate up the quality of evidence. The three primary reasons for rating up are:
1) a large magnitude of effect exists; 2) there is a dose-response gradient, and 3) all plausible confounders or other biases increase our confidence in the estimated ef-fect. Finally the overall quality will be categorized as high, moderate, low or very low.
28 Clinical evaluation - Methods
GRADE gives the following definition of the different classes of evidence:
Grade Definition
High We are very confident that the true effect lies close to that of the estimate of effect Moderate We are moderately confident in the effect estimate: The true effect is likely to be
close to the estimate of effect, but there is a possibility that it is substantially different
Low Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
29 Clinical evaluation - Results