Searching non-Anglophone databases I: developing a search strategy for PSYNDEX
Authors: Margareta Langer, Dorothea Neumann, Clive Adams
Background: Searching non-Anglophone databases identifies trials from journals not in sources such as MEDLINE. PSYNDEX, for example, is a German language psychological literature database containing over 100000 reports. It has not only idiosyncratic indexing but the added complication of text and indexing in more than one language. Two German medical informatics students (ML, DN) spent 3 months working with the Cochrane Schizophrenia Group searching PSYNDEX.
Objective: To identify all randomized controlled trials (RCTs) in PSYNDEX.
Methods: PSYNDEX was acquired, on one month's free trial, from SilverPlatter. English phrases, from published MEDLINE searches, were used along with the German equivalent. The precision of each search phrase was inspected. Terms with poor yield (<1 trial: 50 reports) were discarded. Additional phrases were identified by inspecting reports of RCTs found by the first search. The searches were repeated with these phrases.
Results: 701 reports were identified and independently inspected by ML and DN. Approximately 60% of these reports were RCTs. It was difficult to improve upon the search without a prohibitive loss of precision. 27% of trials were identified by German phrases alone, 9.5% solely by English phrases, and 63% by both languages. A mono-lingual German or English searcher would find 90% or 73% of the trials respectively.
Conclusions: PSYNDEX contains many trials not indexed by other databases. Careful multilingual searching of non-Anglophone databases may counter the effects of publication and language bias within reviews.
Patellofernoral pain syndrome: a critical review of the clinical trials on non-operative therapy
Authors: Bruce Arroll, Elaine Ellis-Pegler, A Edwards, Grant Sutcliffe
Background: Anecdotal evidence from local clinicians suggested that a wide range of therapies are employed in treating patello femoral pain syndrome.
Objectives: The aim of this review was to systematically search and critically appraise the literature on patello femoral pain syndrome.
Methods: The definition of patellofemoral pain syndrome was broad and included patients with cartilage damage. This involved computer based and hand searches as well as writing to potential authors.
Results: Five randomised controlled trials were found as well as some follow up studies. The prognosis for most new cases of patellofemoral pain syndrome is good although a proportion will have persisting symptoms. Quadriceps exercises were effective while knee braces were not. Both orthotics and intramuscular glycosaminoglycan polysulfate had encouraging results for patients with cartilage damage and confirmation of these results is needed.
Conclusion: There were many studies of biomechanics with an assumption that alteration of abnormal biomechanics will result in clinical benefit. More emphasis needs to be placed on therapeutic benefit. There is limited evidence on which to base therapy and there needs to be more high quality research. Studies need to be longer, account for predisposing factors and have more standardised means of assessing outcomes.
Calcium antagonists in patients with subarachnoid hemorrhage (SAH): a meta-analysis
Authors: Valery Feigin, Gabriel Rinkel, Ale Algra, Marinus Vermeulen, Jan van Gijn. University Hospital of Utrecht, University of Amsterdam, The Netherlands
Background: It has been reported that calcium antagonist drugs reduce the frequency of delayed cerebral ischaemia (DCI) and improve outcome after aneurysmal SAH, but definitive evidence is lacking.
Objectives: To determine whether calcium antagonists improve outcome and reduce the incidence of DCI in patients with aneurysmal SAH.
Methods: Systematic overview of unconfounded, randomized trials that were completed by January 1996 comparing calcium antagonists with control started within 10 days of SAH onset.
Results: Nine trials met all the inclusion criteria. Overall, calcium antagonists produced a highly significant reduction in the odds of ischemic related poor outcome (OR 0.57; 95% CI 0.44-0.74) and mortality (OR 0.53; 95% CI 0.37-0.76) and DCI (OR 0.58; 95% CI 0.50-0.67) within three months after SAH. In the analysis for only nimodipine, treatment was associated with a highly significant 45% reduction in the odds of poor outcome from all causes (OR 0.57; 95% CI 0.43-0.76). A strong, but statistically insignificant, tendency towards reducing the incidence of rebleeding episodes among patients allocated to calcium antagonist treatment was found (OR 0.77; 95% CI 0.56-1.05). Unlike nicardipine and AT877, nimodipine did not significantly reduce the incidence of angiographically detected cerebral vasospasm (OR 0.60; 95% CI 0.32-1.12).
Conclusions: Calcium antagonist drugs reduce ischemic related poor outcome at three months after aneurysmal SAH. Nimodipine improves overall outcome within three months of SAH. The evidence for a reduction of poor outcome from all causes by nicardipine and AT877 is inconclusive.
From Bush Tracks to the Super Highway: Can Cochrane Make the Connection?
Author: Justine Mayer (Royal Australian College of General Practitioners Training Program, Darwin, Australia)
Background: 75% of Australian Aborigines in the Northern Territory live in rural and remote areas. While their needs are considerable, access to even basic health care is limited. New initiatives have been introduced to encourage general practitioners to move to these areas. Developments in information technology and services like the Cochrane Collaboration have the potential to provide considerable support to isolated rural doctors.
Objective: To examine the availability of evidence relevant to the management of 5 common medical problems in a rural Aboriginal community. The information gathered from the databases available at the major referral hospital will be compared with the areas currently under study at the Cochrane Collaboration.
Methods: The 5 medical questions selected all met the following criteria: they were common rural Aboriginal health treatment strategies, they involved a considerable use of local resources, and the evidence of their effectiveness was not known to the author. Electronic searches of Medline, Australian Medical Index, the Rural Health Register and the Cochrane Database of Systematic Reviews were performed to identify relevant randomised controlled trials (RCT). Where RCT's were not available alternative evidence was considered.
Results: On the whole, relevant level 1 evidence, (RCT's), was scarce. The databases were most helpful for 2 questions also applicable to developed countries. The other 3 questions are currently research priorities in the developing world, and access to well-designed studies was limited. At present, the Cochrane Collaboration appears to be addressing 2 of the 5 questions.
Conclusions: Recent developments in information technology and improved access to medical databases are changing the nature of medical practice in some fields. For general practitioners working with remote and disadvantaged groups, the relevant evidence remains difficult to access. It is likely that health divide between rich and poor extends to medical research. Overcoming this obstacle is a major challenge for the Cochrane Collaboration.
40 Years of Randomised Trials in the New Zealand Medical Journal
Authors: B Neal, A Rodgers, M-J Mackie and S MacMahon. Clinical Trials Research Unit, University of Auckland, New Zealand
Background: It is well recognised that a register of randomised trials would be incomplete if it relied solely on trials identified by current electronic searching methods. Aim: To identify all randomised trials published in the New Zealand Medical Journal, to document the basic characteristics of these trials and to count the number that were detectable on Medline.
Methods: All issues of the New Zealand Medical Journal between 1943 and 1995 were systematically hand-searched and identified trials were characterised. The proportion of these trials detectable by a maximally sensitive Medline (1995 release) search strategy was calculated.
Results: The hand search identified 152 randomised controlled trials. Only 18% of studies reported on the method of randomisation and only 13% provided evidence that final analyses were conducted on an intention to treat basis. Since 1966, when Medline became available, 143 trials were published of which 89 (62%) were identified by the Medline search.
Conclusions: Important information on design and analysis was frequently absent from the trial reports. In addition over one third of randomised trials were not identified by the optimal Medline search strategy. These findings are consistent with previous studies.
How to make the medical standard information database to use for all.
Author: Zenon Piasecki, Kazimierz Roszkowski, Tadeusz Olakowski.
Background: Now there is a possibility of exchanging information concerning RCT (Randomized Controlled Trials) both the previous ones and present carring out. That can be done due to fruitful works of Cochrane Collaboration Center and its director Dr Iain Chalmers. it seems to be obvious that exchange of information about assumptions and results as well as persons interested in problems undertaken, can be adopted by other science fields quite easily. But there is a lot of work to do on the medical field.
The basic data describing health conditions of society are analysed practically in every country in the World-Rates like normortality, hospital morbidity,incidence (especially for some of infection diseases), or any other rates similar to them, can be use as a crude parameters or characteristic for selected diseases. The analysis of these parameters in connection with data concerning pollution of natural environment is one of the most important factor for proper financial funds allocation in medical service.
But possibility of preparing comparative analysis depends on accessibility to standards of creating appropriate databases for general medical data (epidemiological data),as well as for demographic,or environmental and socioeconomic data. Possibility of using these information could be helpful in verifying health services (protection) programmes carrinq out in many countries at presenter will contribute to optimalization these programmes and improve its results.
Conclusions: The reports presents some solutions concerned:
Strategies for the investigation of hyperandrogenism and the principles of evidence based medicine
Author: George Phillipov. Endocrine & Diabetes Service, The Queen Elizabeth Hospital, Woodville, South Australia 5011
Background: Evidence based medicine is defined as the integration of individual clinical expertise with the best external clinical evidence. Best external clinical evidence results from a process where diagnostic tests and treatments are evaluated to select only those which are both more powerful and safer. In hirsutism and menstrual disturbances diagnostic tests are important to exclude any serious underlying causes such as congenital adrenal hyperplasia (CAH) and Cushing's syndrome (CS). Most review articles routinely advocate strategies based on a battery of plasma assays and dynamic adrenal testing.
Objectives: To examine whether the prevailing clinical consensus on the diagnostic evaluation of hyperandrogenism is consistent with the notion of best external clinical evidence.
Results: Typically the following assays are recommended -plasma testosterone (ovarian tumour), DHEAS (hyperadrenal activity), 17_-hydroxyprogesterone (17_-OHP) (21-CAH) and urinary free cortisol following overnight dexamethasone therapy (CS). A further extension is ACTH stimulation of the adrenal gland and measurement of 17_-OHP to assess any latent existence of 21-CAH. However abnormal results from such assays usually do not allow a definite diagnosis of the underlying condition and the protocol provides no information on a variety of rare disorders, particularly for CAH. In contrast specific measurement of urinary steroids based on a single 24h urine specimen can uniquely define all forms of CAH, identify and partially differentiate CS and provide information on hypercortisolism, hyperandrogenism, thyroid disorders and various enzymatic activities.
Conclusions: The most powerful and least invasive diagnostic assay for the definition of major endocrine disorders is currently overlooked in clinical reviews detailing the investigation of hyperandrogenism. This is a result of misconceptions entrenched in the literature and represents a failure to seek the best external evidence.
Delphi's twin: A method to improve the quality of standards in medicine
Authors: Franz Porzsolt (Dept Medicine III) and Wilhelm Gaus (Dept Biometry and Medical Documentation, University of Ulm, Germany)
Background: Despite of general availability of the scientific information, the treatment standards differ sometimes considerably among major centers. These differences in standards can be explained either by differences in the selection of published results or in the interpretation of published results or in both of it.
Objective: To examine to which extent differences in the interpretation of published results influence the definition of standards.
Method: 42 German Cancer Centers were invited to submit their standards for three tumors. The submitted standards were re-distributed to another center to be checked for necessary changes. Next, the literature should be submitted which supports the requested changes. The requested changes plus the supportive literature was again re-distributed to a third center in order to assess the general scientific quality of the submitted literature on an scale of 0-100 and to decide how well the the enclosed literature supported the requested changes. Finally, a structured instrument (38 yes/no questions) was designed to describe the "strength of evidence of clinical trials (SECT)".
Results: Of the 15 submitted and re-distributed standards 4 were returned without suggested changes, 7 with suggested changes supported by references and 2 not supported by references. More than 50% of the colleagues felt that the requested changes were only partially or not at all supported by the references. The conventional assessment of the scientific quality of the references was quantified and varied extemely. The structured assessment (SECT) resulted in much more homogenous results.
Conclusions: The assessment of the quality of reported trials should be standardized. The results will improve in three dimensions: in the quality of standards, in the quality of the scientific support of standards, and in the agreement on these standards.
Meta-analysis of the efficacy of NSAIDs for low-back pain
Authors: Rob Scholten, Bart Koes, Jan Mens, Lex Bouter
Abstract: Objectives: To assess the efficacy of nonsteroidal antiinflammatory drugs (NSAIDs) for low-back pain.
Methods: Computer-aided search of published randomized clinical trials (RCTs) comparing NSAIDs with placebo. Assignment of a methods score (MS) for quality (maximum = 100 points) to all RCTs by 2 independent researchers. Attempt at statistical pooling of different RCTs regarding similar patient groups and outcome measures by Peto's method.
Results: Ten RCTs were identified. The methods scores (MS) ranged from 30 to 73 points (median 51.5). Two studies reported relief of pain assessed by the patients by means of a visual analogue scale (VAS), 1 study applied a sequential analysis and reported p-values with no further details, 1 study presented only results of subgroups, 1 study presented no clear outcome measure, and 1 study was a cross-over trial pertaining to change of pain in patients with chronic low-back pain. There were insufficient details regarding the VAS scores to enable statistical pooling and no reaction of the original authors was obtained upon our request for detailed information. Thus, the results of only 4 out of 10 RCTs could be pooled.
Four of the 10 studies addressed the effect of 3 different NSAIDs in patients with acute low-back pain without sciatica. The pooled OR for no improvement after 7-10 days for 2 RCTs with sufficient information was 0.46 (95%-CI 0.24-0.90). Five RCTs (regarding 3 different NSAIDs) pertained to patients with acute low-back pain with sciatica or prolapsed discs. The pooled OR for 2 RCTs with sufficient information was 0.66 (95%-CI 0.30-1.45). The OR for all 4 poolable RCTs combined was 0.53 (95%-CI 0.32-0.89).
Conclusions: Only 4 out of 10 placebo controlled trials regarding the efficacy of NSAIDs for low-back pain provided sufficient information to enable statistical pooling. The results suggest that NSAIDs might be effective for short term symptomatic relief in patients with acute low-back pain.
Searching non-Anglophone databases II: developing a search strategy for LILACS
Authors: Karla Soares, Clive Adams
Background: Searching non-Anglophone databases identifies trials from journals not in sources such as MEDLINE. LILACS (Latin American and Caribbean Health Sciences Literature) is a general health care database produced by the Pan American Health Organization. Although the abstracts of reports are mostly in Portuguese or Spanish, the titles are frequently translated into English and the reports are indexed in Portuguese or Spanish and English.
Objective: To identify all randomized controlled trials (RCTs), and controlled clinical trials (CCTs) in LILACS.
Methods: LILACS was searched with frequently used English phrases, their Portuguese/Spanish equivalent and some unique phrases to each language. Large randomly selected samples of the results of these searches were then inspected independently by an English speaker (CA) and by someone fluent in Portuguese, Spanish and English (KS).
Results: Overall 2093 reports were identified by all languages used in combination. 46% (963) of these were RCTs/CCTs. English phrases alone find 70% of these reports and 57% of this sub-sample were RCTs/CCTs. Portuguese/Spanish phrases found 72% of the total and again 57% of these were RCTs/CCTs. Monolingual English speakers searching the printed off reports were 83% sensitive and 90% specific for RCTs/CCTs.
Conclusions: LILACS contains many trials not indexed by other databases. Although, being a multilingual searcher has the advantage of a more comprehensive identification of trials in this database, an English speaker can identify a high percentual of the trials with adequate degrees of sensitivity and specificity.
Development of a consensus criteria list to assess trial quality in systematic reviews
Authors: Arianne Verhagen, Henrica de Vet, Rob de Bie, Fons Kessels, Maarten Boers, Paul Knipschild
Background: Assessing the quality of RCTs included in a review is important. Firstly, because variation in quality can explain differences in the results. Secondly, because when the methodology of the RCTs included in the review is weak, the conclusions of the review cannot be very strong. It is also needed that experts achieve agreement on quality assessment, especially where the quality of individual studies is incorporated in systematic reviews and meta-analyses. Therefore we aimed at developing a universal criteria list as a minimum reference standard to be used in future systematic reviews.
Objectives: To develop a consensus criteria list to assess the methodological quality of trials in systematic reviews.
Methods: The Delphi method is a consensus method using written questionnaires. The participants remain anonymous and receive feedback after each round about the answers and arguments of other participants.
Results: The procedure started with 33 participants representing different points of view on quality assessment. The participants are clinicians, epidemiologists and statisticians. Starting from 206 quality items we want to come to a hierarchy of importance and end up with a short list. At least 3 or 4 Delphi-rounds are necessary to obtain consensus. The resulting Delphi criteria list will be presented.
Conclusions: A consensus definition about 'quality' will be derived from the results. There are differences in opinion between statisticians on the one hand and clinicians/epidemiologists on the other. Practical implications will also be presented at the congress.
The effects of using different criteria lists on the outcomes of systematic reviews or meta-analyses
Authors: Arianne Verhagen, Rob de Bie, Ton Lenssen, Henrica de Vet, Paul Knipschild
Background: Over the years several criteria lists have been developed to be used in quality assessment of RCTs. A quality criteria list can be used in different ways. It can be used in systematic reviews in which the summarization of the results is based on methodological criteria, and to determine strata of methodologic quality upon which you can decide which RCTs have sufficient quality to allow for pooling in a meta-analysis. It has been shown that the use of different lists for systematic reviews changes the conclusion of the review. We are interested in these effects by using three different criteria lists for quality assessment in a systematic review.
Objectives: To assess if the use of different criteria lists changes the outcomes of a systematic review or meta-analysis.
Methods: A data set of 28 articles of a systematic review concerning the efficacy of 904 nm laser therapy in musculosceletal disorders is used. The three list used are: the 'Jadad-list' of 3 items, developed by Alejandro Jadad, the 'Maastricht list", developed at the University of Limburg and the 'preliminary Delphi-list' of 7 items. This list contains the preliminary results of the development of a consensus criteria list by means of the Delphi method. Articles are reviewed by two reviewers independently and scores are determined in a consensus meeting. The difference is ranking, and the possible differences in strata of methodologic quality will be determined.
Results and conclusions: We will present the results of this research at the Colloquium.
This page was last updated on 18 November 1996
Comments for improvement or correction are welcome.
georg.koch@cochrane.de