Re-examination of old truths: replication of a study to measure the incidence of lactational mastitis in breastfeeding women

In order to build knowledge and theories, the scientific method requires that we either integrate earlier findings with new or alternatively that we challenge previous knowledge if we find that new findings contradict the old. There is a tacit understanding in science that “once is not enough” which implies that experiments should be replicated in order to verify findings. Popper wrote that we do not take our observations seriously or treat them as scientific until we have reproduced and tested them [1]. The reason why description of the methods used in research papers is so important is that it should be possible for other researchers to set up a similar experiment to check the viability of previous results. Despite this, it is rare that replication studies are carried out.

As long ago as 1995, researchers drew attention to the fact that studies regarding incidence and prevalence of mastitis were lacking, in part due to the problem of collecting trustworthy data [2]. These problems remained unsolved ten years later [3]. Despite this, the incidence of lactational mastitis is cited in almost all studies published on the subject.

This was a retrospective questionnaire study carried out at a meeting of lactation specialists in Denmark, 2011. Mastitis was defined as any combination of breast tension, erythema, pain and fever in association with breastfeeding.

The respondents of the questionnaire were not in any state of dependence on the researcher and the president of the Danish Association of Certified Lactation Consultants did not consider there to be any ethical problems for the use of the questionnaire and gave permission for it to be used (personal communication, 27 September, 2011). Since the respondents placed their questionnaires in an envelope, it was possible for those who did not wish to complete the questionnaire to return it blank, without the researcher being aware of this and therefore a completed questionnaire was assumed to have informed consent.

A total of 91 women attended the meeting and 90 (99%) returned a completed questionnaire. Ages ranged between 33 and 60 years with a median age of 48 years. The total number of babies born to these women was 225 and of these 219 (97%) had ever been breastfed. The range of months of breastfeeding of the last child which the respondents had breastfed was 1–54 months, median 11.5 months.

Thirty women (33%) reported that they had ever experienced mastitis during breastfeeding. The respondents reported episodes of mastitis with 40 of the 219 (18%) breastfed babies born to them. Six women had mastitis with more than one of their babies. A total of 12 women (13%) reported that they had suffered mastitis with their latest breastfed child.

A physician examined 35% (14 of the 40 cases) of the respondents and 93% (n = 13) of these were prescribed antibiotics. In 98% (39 of the 40 cases) the respondents reported that they carried on breastfeeding throughout the illness.

Of the 25 who answered the question about which breast was affected, 56% (n = 14) had experienced mastitis in their right breast and 44% (n = 11) in their left breast (Figure 1). One respondent could not remember which breast segment had been affected. Of the 24 who could remember which segment of the breast had been affected, 83% (n = 20) answered that the outside segments (nearest the arms) were affected, upper and lower segments equally. The remaining 17% (n = 4) reported affliction of the inside segments.

Surprisingly, the percentage of women in the present study who reported that they had ever had mastitis was exactly the same as the percentage reported in the original Riordan and Nichols study that is, 33% [11]. The greatest problem in interpretation of the results of both the present study and the original study is the fact that no time limit was stated in which the “population at risk” could be identified. The study population cannot be seen as the population at risk for mastitis. Therefore even though in Riordan and Nichols study, 33% of the women stated that they had experienced mastitis during a single lactation period and in the present study 33% said that they had ever had mastitis, these figures cannot be seen as true incidence rates. These were not new cases of mastitis but could have occurred anytime during a period of five years in the older study and during an unlimited period in the new study. We do not know what the population at risk was during a five-year period or how the size of the populations at risk varied over time.

Since it has been demonstrated earlier that none of the symptoms of mastitis correlate with increasing amounts of potential pathogens in the milk [12] and it has been suggested that some women with mastitis may not be febrile [13], the definition of mastitis used in this study allowed for any combination of the common symptoms of mastitis. This could be interpreted as a loose definition. Neither Riordan and Nichol’s study nor the present study required that symptoms should have been experienced for 24 hours or more, which may mean that some of the respondents had episodes of engorgement rather than mastitis. However, the study population in both studies was limited and selective; lactation consultants are likely to be more aware of differences between engorgement and mastitis than other groups of women. The populations in the two studies differed somewhat; the median age of the participants in the present study was 48 years, which was higher than in Riordan and Nichol’s study (median 36 years).

The predisposing factors suggested by the respondents as primarily responsible for symptoms of mastitis are interesting. These women work as breastfeeding consultants and as such have a great deal of clinical experiences on differing breastfeeding problems. Few of the respondents felt that bacteria were the main cause of mastitis despite the fact that the scientific literature often cites S. aureus as the main causative organism [14–16]. They answered that irregular and incomplete emptying of the breast was the main cause of mastitis. This would infer that they believe an overfull breast causes high temperature, red, tense and painful breasts. In order for this to be the case, the breast milk would have to overflow or permeate from the milk ducts into the surrounding connective tissue. Researchers have proposed that blocked ducts are responsible for this overflow into the connective tissue [16–18] but a diagnosis of blocked mammary ducts is not easily made. A “chicken or the egg” situation arises if we consider that an inflammatory process in the connective tissue with ensuing swelling might well exert pressure on the milk ducts making drainage difficult. Whether blocked milk ducts are a cause of or a product of lactational mastitis has yet to be scientifically confirmed.

It was notable that 33% reported being treated with antibiotics. This figure is relatively low in comparison to other international studies, which show that 77% - 97% of women are prescribed antibiotics [2, 9, 19]. Of the women in this study who were put in contact with a physician a great majority of them were prescribed antibiotics. The present study was carried out in Denmark and it has been shown that in neighbouring Sweden approximately 15% are prescribed antibiotics [12]. These findings might represent a cultural tendency to lessen the use of antibiotics for this group of patients since they are often managed by midwives and lactation consultants who are used to the sight of very red and inflamed breasts. Their experience may have taught them that although the women feel very ill, spontaneous recovery is common.

In an audit of mastitis cases from Australia, it was shown that mastitis occurred most often unilaterally and equally in both left and right breasts [13]. Similarly, results from the present study show a relatively equal distribution between right and left sides. The results shown in Figure 1 indicate that in a large proportion of women, inflammation occurs in the outside quadrants of the breasts.

An American review of the scientific literature has reported a wide range of incidences from 2% to 33% [20] and the 33% incidence reported in the review is also included in a World Health Organisation report on mastitis [21]. Science is dynamic; methods used and conclusions drawn change over time. Retrospective studies have during recent years become less interesting from a scientific view point due to the unsure nature of the results caused by re-call bias. It is therefore relevant to challenge the reported 33% incidence of mastitis. I propose that the term incidence was incorrectly used in Riordan and Nichols paper and for that reason the incidence rate of 33% should not continue to be cited by other researchers.

The population of breastfeeding women is not easily measured because of ever-fluctuating initiation levels and levels of breastfeeding cessation in the communities where breastfeeding is measured. Therefore, by the rule that says that we must be able to identify the population at risk, the incidence of mastitis can only ever be an approximation. Despite this, researchers must strive to be as exact as possible when reporting incidences and should attempt to evaluate the true population at risk. Well-designed studies in many different global locations are needed before any conclusions can been drawn about the range of incidences of mastitis.