Carel Thijs, Maastricht University, dept. of Epidemiology
31 May 2011
The review of breastfeeding promotion trials by Imdad et al. [1] is a timely extension of the systematic review by Chung et al from 2008 [2]. Yet we are concerned about selective reporting. While Chung identified and reviewed 36 trials, mainly from developed countries, Imdad’s review expands this to 39 studies from developed and 14 from developing countries. However, compared to Chung, Imdad missed about 20 trials for unclear reasons. Imdad states that “hand search of bibliographies of relevant reviews was performed” but did not refer to Chung’s review. It is important to know how they selected the trials to exclude the possibility of a systematic bias. For instance, Imdad missed the trial by Kools et al. [3] which meets all of their inclusion and none of their exclusion criteria. When we replicated the search exactly as formulated in Imdad’s review [4], we found 7,647 publications (including Kools). Imdad writes under the heading ‘Selection’: “All studies that were included in our review looked at patterns of breastfeeding at 4-6 weeks and at six months postpartum.” It is not clear whether this was an inclusion criterion too, and at what moment this played a role in their selection process; in their in- and exclusion criteria this timing of endpoints is not mentioned. Because of the big number of initial studies in Imdad’s review (they started with 968 publications whereas we found 7,647), the further selection is likely to be based on the text of the abstracts alone. The caveat is that abstracts may selectively report positive findings for one time point, while additional endpoints may be reported only in detail in the text, either because they do not add information (as in the trial by Kools: results were negative regardless of timing), or because positive results are selectively reported in the abstract and negative results at other time points only in the text (selective reporting). This is an important difference with the review by Chung that had a selection step based on full text assessment. When performing a systematic search, the number of publications can easily be restricted with the help of a methodological search key (e.g. limiting the PubMed to randomised trials would have resulted in only 400 publications). With the resulting lower number of publications, more time can be spent to screen the full text and tables. This could prevent bias by selecting publications based on results in abstracts alone. It would be of utmost importance to know more details of the selection process in Imdad’s review to exclude the possibility of a systematic bias and a potential overestimation of the effectiveness of breastfeeding promotion programmes.
Carel Thijs, MD, PhD, associate professor of Epidemiology Hein de Vries, PhD, professor in Health Communication Maastricht University, department of Epidemiology and department of Health Education and Health Promotion, PO Box 616, 6200 MD Maastricht, the Netherlands
[1] Imdad A, Yakoob MY, Bhutta ZA. Effect of breastfeeding promotion interventions on breastfeeding rates, with special focus on developing countries. BMC Publ H. 2011, Apr 13; 11; Suppl 3: S24. http://www.biomedcentral.com/1471-2458/11/S3/S24 [2] Chung M, Raman G, Trikalinos T, Lau J, Ip S. Interventions in primary care to promote breastfeeding: an evidence review for the U.S. Preventive Services Task Force. Ann Intern Med. 2008 Oct 21;149(8):565-82. [3] Kools EJ, Thijs C, Kester ADM, van den Brandt PA, de Vries H. A breast-feeding promotion and support program- a randomized trial in the Netherlands. Prev Med 2005 Jan; 40(1): 60-70. [4] ("Breast Feeding"[Mesh] OR breastfeed* OR lactation) AND (education OR promotion OR counseling OR intervention OR support OR social support); searched in PubMed, 21 April 2011.
Effect of breastfeeding promotion interventions
31 May 2011
The review of breastfeeding promotion trials by Imdad et al. [1] is a timely extension of the systematic review by Chung et al from 2008 [2]. Yet we are concerned about selective reporting. While Chung identified and reviewed 36 trials, mainly from developed countries, Imdad’s review expands this to 39 studies from developed and 14 from developing countries. However, compared to Chung, Imdad missed about 20 trials for unclear reasons.
Imdad states that “hand search of bibliographies of relevant reviews was performed” but did not refer to Chung’s review. It is important to know how they selected the trials to exclude the possibility of a systematic bias. For instance, Imdad missed the trial by Kools et al. [3] which meets all of their inclusion and none of their exclusion criteria. When we replicated the search exactly as formulated in Imdad’s review [4], we found 7,647 publications (including Kools). Imdad writes under the heading ‘Selection’: “All studies that were included in our review looked at patterns of breastfeeding at 4-6 weeks and at six months postpartum.” It is not clear whether this was an inclusion criterion too, and at what moment this played a role in their selection process; in their in- and exclusion criteria this timing of endpoints is not mentioned. Because of the big number of initial studies in Imdad’s review (they started with 968 publications whereas we found 7,647), the further selection is likely to be based on the text of the abstracts alone. The caveat is that abstracts may selectively report positive findings for one time point, while additional endpoints may be reported only in detail in the text, either because they do not add information (as in the trial by Kools: results were negative regardless of timing), or because positive results are selectively reported in the abstract and negative results at other time points only in the text (selective reporting). This is an important difference with the review by Chung that had a selection step based on full text assessment.
When performing a systematic search, the number of publications can easily be restricted with the help of a methodological search key (e.g. limiting the PubMed to randomised trials would have resulted in only 400 publications). With the resulting lower number of publications, more time can be spent to screen the full text and tables. This could prevent bias by selecting publications based on results in abstracts alone.
It would be of utmost importance to know more details of the selection process in Imdad’s review to exclude the possibility of a systematic bias and a potential overestimation of the effectiveness of breastfeeding promotion programmes.
Carel Thijs, MD, PhD, associate professor of Epidemiology
Hein de Vries, PhD, professor in Health Communication
Maastricht University, department of Epidemiology and department of Health Education and Health Promotion, PO Box 616, 6200 MD Maastricht, the Netherlands
[1] Imdad A, Yakoob MY, Bhutta ZA. Effect of breastfeeding promotion interventions on breastfeeding rates, with special focus on developing countries. BMC Publ H. 2011, Apr 13; 11; Suppl 3: S24. http://www.biomedcentral.com/1471-2458/11/S3/S24
[2] Chung M, Raman G, Trikalinos T, Lau J, Ip S. Interventions in primary care to promote breastfeeding: an evidence review for the U.S. Preventive Services Task Force. Ann Intern Med. 2008 Oct 21;149(8):565-82.
[3] Kools EJ, Thijs C, Kester ADM, van den Brandt PA, de Vries H.
A breast-feeding promotion and support program- a randomized trial in the Netherlands.
Prev Med 2005 Jan; 40(1): 60-70.
[4] ("Breast Feeding"[Mesh] OR breastfeed* OR lactation) AND (education OR promotion OR counseling OR intervention OR support OR social support); searched in PubMed, 21 April 2011.
Competing interests
We declare we have no competing interests