Current issues of ACP Journal Club are published in Annals of Internal Medicine


Editorials

Beyond ACP Journal Club: how to harness MEDLINE for etiology problems

ACP J Club. 1994 Nov-Dec;121:A10. doi:10.7326/ACPJC-1994-121-3-A10

Related Content in the Archives
• Editorial: Beyond ACP Journal Club: how to harness MEDLINE for prognosis problems
• Editorial: Beyond ACP Journal Club: how to harness MEDLINE for review articles



In previous editorials in this series, we discussed the use of MEDLINE to solve clinical problems and described search approaches to find methodologically sound articles evaluating the therapy (1) and diagnosis (2) of diseases. In this editorial, we deal with etiology, beginning with a clinical scenario.

It is now generally well accepted that breast milk is beneficial to babies during the first year of life, and recently it has been suggested that long-term beneficial effects exist for the mother. One of your patients, a 25-year-old woman with a family history of breast cancer, is expecting her second child. You are encouraging her to consider breast-feeding but she is unwilling because she had a bad experience in trying to breast-feed her first child, experiencing pain and an infection before she got started. You recall reading something recently about the protective effects of breast-feeding in women at risk for breast cancer but cannot recall the journal or details. You want to be sure of the facts before you discuss this further with the patient.

Your medical texts are unrevealing or unconvincing on this issue. For example, Scientific American Medicine (3) reports that “breast-feeding does not appear to lower the risk of breast cancer, although this finding has been disputed in studies of women younger than 50 years who develop breast cancer,” citing a single publication in 1985 as support (4).

You turn to MEDLINE and do a search. You know that breast-feeding or lactation and breast neoplasms will definitely be terms in your strategy but you want to narrow your search to those studies that investigated whether the former reduces the risk for the latter. Just as there are search approaches for retrieving citations to “on-target” studies evaluating therapeutic interventions (1) or diagnostic tests (2), methodologic terms can be included in a search strategy to retrieve studies of etiology or causation.

The following search, run in the current (1991 to September 1994) MEDLINE file, is a straightforward way to address the task:

breast feeding (mh) or exp lactation (mh)

AND

exp breast neoplasms (mh)

AND

risk (tw)

AND

English language (la)

in which “(mh)” means MeSH (Medical Subject Heading); “exp” means explode (to include all types of breast tumors that are indexed under the main headings, lactation and breast neoplasms); “(tw)” means text word, instructing the computer to look for “risk” in the title, abstract, and subject headings of citations in the MEDLINE database; and “(la)” means language.

This search retrieved 23 citations of which 9 were clearly on target, with 7 documenting a protective effect of breast feeding and 2 providing equivocal or negative results. For 1 citation, Newcomb and colleagues (5), the MEDLINE citation notes that a comment appeared in the May/June 1994 issue of ACP Journal Club. (MEDLINE began noting ACP Journal Club commentaries and posting our editorials in January 1994.) Following up this lead, you leaf through the ACP Journal Club issue and make a copy of the abstract and commentary (6) for your patient and you to discuss.

Why did we include “risk (tw)” in the search? In our extensive testing of search terms to retrieve clinically important and methodologically strong studies of etiology, risk (tw) was the best single term we could find, doing better than MEDLINE indexing terms (7). Even so, its sensitivity for sound, pertinent studies was far from perfect, at 61% in 1986 (with a specificity of 89%) and 67% in 1991 (with a specificity of 79%). You can improve the sensitivity or specificity (but not both at the same time!) by using the complex search strategies in the Table in place of risk (tw) (8). As an example of the effect of the more complex searches, using the most specific strategy retrieves 11 citations, all of which were retrieved in the previous search. It missed, however, 2 of the possibly relevant citations from the first search. Both approaches retrieved the citation to the article by Newcomb and colleagues (6).


Table. Most Effective Searching Strategies for Etiology

Time period Strategy Sensitivity Specificity
To maximize sensitivity
1986 exp cohort studies or all risk (tw) or all causation (tw) or all causal: (tw) 72% 79%
1991* exp cohort studies or exp risk or all odds and all ratio: or all relative and all risk or all case and all control: 82% 70%
To maximize specificity
1986 exp risk or all causation (tw) or all causal: (tw) or all relative and all risk (tw) 51% 97%
1991 case-control studies or cohort studies 40% 98%

*If all the terms in this strategy are combined with a Boolean OR, you get a postings storage overflow message (STORPSTG OVFLW). To avoid this, you must combine, using a Boolean AND, each term in the methodologic strategy with your content term, such as breast-feeding, or run the search twice using half of the methodologic strategy in each search.

Running a MEDLINE search is a fairly fast, highly effective, but quite messy way to locate the current literature on a clinical topic. Searches for studies of etiology that are methodologically sound and relevant for clinical practice can be improved by including methodologic filters, such as the ones described above. The number of relevant citations and the information you retrieve will likely be more comprehensive and current than that in the textbook on your shelf or your colleague down the hall. If the topic is one for which you wish to regularly review the latest studies as they are posted on MEDLINE, you can run regular update searches in MEDLINE in the Selective Dissemination of Information (SDIline) file, which contains the most recent month of citations entered into MEDLINE.

The messiness of MEDLINE searches is gradually being overcome by enhancements in indexing and search strategies and by the creation of specialized literature files that are tailored to the needs of the user. For internists, the contents of ACP Journal Club and the full-text articles abstracted will likely be available on CD-ROM within the coming year and an Internet version is being tested.

Cynthia J. Walker-Dilks, MLS
K. Ann McKibbon, MLS
R. Brian Haynes, MD, PhD


References

1. McKibbon KA, Walker CJ.Beyond ACP Journal Club: how to harness MEDLINE for therapy problems [Editorial]. ACP J Club. 1994 Jul-Aug;121:A10.

2. McKibbon KA, Walker CJ.Beyond ACP Journal Club: how to harness MEDLINE for diagnostic problems [Editorial]. ACP J Club. 1994 Sep-Oct;121:A10.

3. Rubenstein E, Federman DD, eds. Scientific American Medicine. New York: Scientific American Inc. Compact Disc, June 1994.

4. Byers T, Graham S, Rzepka T, Marshall J. Lactation and breast cancer: evidence for a negative association in premenopausal women. Am J Epidemiol. 1985;121:664-74.

5. Newcomb PA, Storer BE, Longnecker MP, et al. Lactation and a reduced risk of premenopausal breast cancer. N Engl J Med. 1994;330:81-7.

6. Lactation was associated with a reduced risk for breast cancer in premenopausal women but not in postmenopausal women [Abstract]. ACP J Club. 1994;120:84. 1994;330:81-7.

7. Wilczynski NL, Walker CJ, McKibbon KA, Haynes RB. Assessment of methodologic search filters in MEDLINE. Symp Comp Appl Med Care. 1993;17:810-6.

8. Haynes RB, Wilczynski NL, McKibbon KA, Walker CJ, Sinclair JC. Developing optimal search strategies for detecting clinically sound studies in MEDLINE. J Am Med Inform Assoc. Nov/Dec 1994; [In press].