Introduction

The World Health Organisation estimates that 1.5 million babies could be prevented from dying each year if women breast fed their infants (exclusively for about 6 months and until infants were 2 years old).¹ Where a mother uses an alternative to breast milk to feed her baby, it is important that she makes an informed decision and that she has not been pressured by commercial promotions to use a substitute. The international code of marketing of breast milk substitutes² was adopted by the World Health Assembly in 1981 to encourage breast feeding and to protect mothers from pressure to use substitutes for breast milk. At that time one member state (the United States) voted against the code and three abstained (Argentina, Japan, and Korea); by the 1996 World Health Assembly meeting all 191 member states had affirmed their support for the code, its implementation, and the implementation of relevant resolutions. By 1997, 17 countries had adopted all or substantially all of the code's provisions as legal requirements.³ Adoption of the code represents the development of an international consensus.

Anecdotal evidence of violations of the code has been presented but no previous studies have used formal sampling techniques.⁴ This study was designed to measure the prevalence of violations of the code using randomly sampled groups of women, health workers, and health facilities in four cities.

The study was overseen and supported financially or by other means by 27 churches, academic institutions, and international non-governmental organisations. Experts in infant feeding were contacted for advice. Accommodation was provided by Unicef United Kingdom. The organisations and people who chose to participate wanted to obtain unbiased information on violations of the marketing code. No funding, support, or advice was received from the manufacturers of breast milk substitutes or organisations forming part of the International Baby Food Action Network (Penang, Malaysia).

	Subjects and methods

Sampling procedure
Since no studies of this kind had been done before a 10% prevalence of reported violations was assumed to be an important amount of violation. A sample of 800 women would give a 95% power to observe at least one reported violation if the true prevalence was 2%. If the prevalence was 10% the sample size would generate estimates of population prevalence with a standard error of 1%. Sample sizes for health workers and health facilities were constrained by practical considerations.

Selection of countries
Countries were grouped into three categories that reflected the status of requirements for compliance with the marketing code. In 63 countries compliance with the code was a legal requirement; in 36 countries compliance with the code was voluntary; and in 96 countries the code had another status, including no code, code awaiting government approval, or the status was unknown.³ A small subset of countries where there were agencies already working in partnership with the Interagency Group on Breastfeeding Monitoring were selected in each category to participate in preliminary discussions on the feasibility of the study (3 of the 63 countries in which compliance was legally mandatory, 2 of the 36 in which compliance was voluntary, and 5 of the 96 classed as giving the code any other status). These countries had contrasting geographical and economic conditions. After discussions with partner agencies six countries were excluded from the study because government permission or personnel support could not be secured.

The final four countries selected were Bangladesh (where compliance with the code is a legal requirement), Poland (which has no code), South Africa (where compliance is voluntary), and Thailand (where compliance is voluntary).

Cities The capital city was chosen for the study in each country; however, in South Africa, government authorities suggested that the study should be done in Durban because the University of Natal would be able to manage the study.

Districts Seventeen districts in Warsaw and 23 in Bangkok in which there were at least four health facilities that served pregnant women and mothers of infants were identified and numbered serially; 10 districts were selected using random numbers. In Dhaka 17 districts which had a minimum of four health facilities and in which 20% of the people were living slums were selected to reflect the overall proportion of people living in slums in the city. Districts were numbered and 10 were selected using random numbers. In Durban there were 93 districts but there were not enough facilities in each district for the same selection procedure to be followed. To ensure a large enough sample of health facilities 23 districts were randomly selected.

Health facilities To be considered eligible for sampling health facilities had to be large enough to see daily at least 10 pregnant women or mothers of infants who were 6 months old. Altogether 132 facilities met the inclusion criteria in Warsaw, and 159 met the criteria in Bangkok. All facilities that met the criteria in the 10 districts were numbered, and four were selected from each district using random numbers; thus, 40 health facilities were studied in each of these cities. In Dhaka 40 facilities met the inclusion criteria and in Durban 46 met the criteria; all of these facilities were included in the study. The coordinators in each country and the interviewers had had no prior contact with staff or mothers at the health facilities studied.

Definitions
All products marketed for infants younger than 6 months and all follow on formulas were considered to be breast milk substitutes. Infant formulas are those substitutes designed to be used from birth up to 4 to 6 months of age. Follow on formulas are designed to replace infant formulas at the age of 4 to 6 months. These definitions were based on World Health Assembly resolution 47.5 (1994) which clarified the code and recommended "fostering appropriate complementary feeding practices from the age of about six months." World Health Assembly resolutions passed after the code was adopted in 1981 have the same status as the code, which was adopted as part of World Health Assembly resolution 34.22. Article 2 of the code specifies the products covered by the code.

Subjects
Over two days 20 pregnant women or mothers of infants aged 6 months were selected from the appointment register at each facility and interviewed. Where no appointment register was available women were systematically sampled according to their position in the waiting room (for example, every fifth consecutive woman was selected). During the same two days, three health workers at each facility were interviewed. The selection of health workers depended on their availability for interview. Interviews with staff at different levels of seniority were sought. Wherever possible confidential interviews were carried out in a quiet designated area. A target sample size of 800 women and 120 health workers was set for each city, except Dhaka where there were problems with transportation because of the rainy season.

Interview procedure
The research coordinator (AT) trained 10 people in each country to interview pregnant women, mothers, and health workers using structured questionnaires; they were also trained to use a checklist to assess materials produced by companies that manufactured breast milk substitutes. (A copy of the checklist can be found on our website at www.bmj.com.) All data were collected within each country during one month between August and October 1996. 

Validation procedure

Assessment of written information Health workers were asked to show the interviewer the materials that had been provided for professional use by manufacturers. These materials were then assessed by the interviewer using a checklist based on the requirements of relevant articles of the code (4.2 and 7.2) (box).

Data management The country coordinator was responsible for the quality of the data and supervision of the interviewers. In each country data were entered by a trained data clerk using EpiInfo software; data were analysed in London. Confidence intervals for proportions were estimated assuming a  binomial random effects model to take account of possible variation between clinics.⁵

	Results

Health facilities and respondents
Forty health facilities were visited in Dhaka, 46 in Durban, 40 in Bangkok, and 39 in Warsaw (one facility was closed for a holiday). In Dhaka 704 women were approached for an interview and all agreed to participate. In Warsaw 794 women were approached and three declined to participate. In Durban 804 women were approached; five declined to participate. In Bangkok 748 women were approached; 52 declined to participate. In Dhaka 112 health workers were asked to participate; one declined. In Warsaw 119 were approached and four declined. In Durban 123 were approached and one declined. In Bangkok 112 were approached and eight declined.

Free samples
In Bangkok 97 out of 370 (26%; 95% confidence interval 18% to 36%) mothers reported receiving free samples of a breast milk substitute, feeding bottle, or teat (table); many women received more than one sample. This was a greater proportion than that found in the other three cities. In Dhaka only one woman reported receiving a free sample. In all cities mothers were more likely than pregnant women to report having received free samples.

View larger version (130K): [in this window] [in a new window]   View larger version (149K): [in this window] [in a new window]   Calendar given to a health worker in Dhaka, Bangladesh. The code states that no material inducement to promote a product should be offered by manufacturers of breast milk substitutes to health workers

Validation
In Bangkok 97 out of 370 mothers reported receiving free samples. Eighteen of the 19 mothers in the validation subset described the samples they had received. The interviewer found that 12 samples were still available. One type of sample described was no longer available but the researcher was able to confirm that that type of sample had at previously been dispensed by the facility. No additional information could be obtained about the other five samples that mothers reported receiving. In Warsaw 34 mothers reported receiving free samples; some mothers received more than one sample. All seven mothers in the validation subset were contacted. Information was available about only three of the samples. In Dhaka the single reported free sample was not verified. Problems with transportation in Durban prevented us from verifying any of the 13 free samples reported.

Information and gifts provided to health workers
Health workers in 22 out of 39 (56%; 95% confidence interval 41% to 72%) health facilities in Warsaw received information from manufacturers which violated the code. This was higher than in the other three cities where prevalence ranged from 6 out of 40 (15%; 4% to 26%) health facilities in Dhaka to 13 out of 40 (33%; 18% to 47%) in Bangkok (figure).

	Discussion

In this study pregnant women and mothers in four cities reported receiving free samples of infant formula, other breast milk substitutes, feeding bottles, or teats in contravention of articles 5.2 and 7.4 of the international code of marketing of substitutes for breast milk. None of the free samples were to be used for professional research purposes according to reports from health workers interviewed at the health facilities. The greater the number of facilities given samples in each city the greater the proportion of pregnant women and mothers who received samples. Most of the samples were reported to have come from a health facility; this suggests that samples given to facilities were passed on to mothers, whether or not that was the intention of the company donating the samples.

The effect of giving free samples of infant formula, bottles, and teats to women who are breast feeding has been examined in five studies.^6-10 Frank et al reported that women receiving a discharge pack that contained products consistent with the code (such as breast pads and pamphlets on breast feeding) were more likely to breast feed for longer than those receiving a commercial discharge pack (containing formula, bottles, or teats), were more likely to be breast feeding at 4 months post partum, and were more likely to delay feeding solid foods.¹¹ Two meta-analyses of the five studies confirmed that commercial discharge packs have a detrimental effect on breast feeding at one month after birth (odds ratios 1.45 (95% confidence interval 1.07 to 1.96)¹² and 1.4 (1.0 to 2.1)¹³).

Article 7.3 of the code outlines the responsibilities of both health professionals and manufacturers as they pertain to the donation and receipt of gifts, yet widespread violations were detected in the four cities surveyed. Although many of the gifts were of little value financially, health professionals working in underfunded healthcare systems, such as in Bangladesh, may find it difficult to resist accepting these inducements. The presence of brand name items in the health facility may constitute professional endorsement of a particular product to patients seen in the facility.

The code states that all information produced by companies that manufacture or distribute breast milk substitutes must include details on the benefits and superiority of breast feeding and the risks associated with bottle feeding (articles 4.2 and 7.2). Nevertheless, 15% to 56% of health workers interviewed stated that their facilities had received materials that contravened these articles.

It may not be possible to generalise these findings to other areas because conditions in rural districts could be different. Equally, the promotional activities of companies may vary from country to country and city to city. However, both the number and nature of code violations suggest that systematic contravention of the code exists; it would be reasonable to believe that similar violations are occurring at similar rates in other cities and countries. These findings are the consequences of the promotional activities of 21 companies, six of which were trying to sell their products in more than one of the countries studied.

Importance of implementation and monitoring of the code
Bangladesh was the only country studied which had laws governing the marketing of breast milk substitutes; the smallest number of free samples were detected there. Warsaw had the highest number of health facilities in which information that violated the code was available to health professionals. Warsaw also had the highest proportion of health professionals who received free gifts; Poland has no legal or voluntary code governing any aspects of marketing discussed in this paper. South Africa and Thailand have voluntary codes. In Thailand the voluntary code was revised in 1995.