The CASTLE (Candida and Staphylococcus Transmission: Longitudinal Evaluation) study investigated the microorganisms involved in the development of mastitis and “breast thrush” among breastfeeding women , and confirmed the role of Candida species in the symptoms of “breast thrush” . Three hundred and sixty nulliparous women were recruited from two hospitals in Melbourne, Australia: the Royal Women’s Hospital, (RWH) a public tertiary referral centre, and Frances Perry House (FPH), a private co-located hospital. At recruitment participants completed a questionnaire asking about maternal age, gestational age, intended length of breastfeeding duration, highest level of educational attainment, marital status and previous staphylococcal infections. Following birth, participants were followed-up six times: in hospital and then at home weekly until four weeks postpartum and at eight weeks by telephone.
Eligibility criteria for the study were: between 18–50 years of age; nulliparity; ≥ 36 weeks pregnant at recruitment; singleton pregnancy; breastfeeding intention for at least eight weeks postpartum; sufficient proficiency in English to complete written questionnaires and a telephone interview; residing ≤ 16 km from Melbourne Central Business District (CBD). Criteria for exclusion were: medical conditions which do not allow breastfeeding; breast reduction surgery; dermatitis on nipple during pregnancy; under care of the Women’s Alcohol and Drug Service (WADS); under care of Mental Health Service or social worker.
Self-administered questionnaires completed in hospital after birth and at weeks 1, 2, 3, 4 and 8 postpartum collected information about milk supply, nipple damage and mastitis. At weeks 1, 2, 3, 4 and 8 postpartum women were asked to rate their milk supply since their last interview. They could state that they were “producing enough milk for their baby”, “not producing enough milk”, “over-producing breast milk” or that their “milk supply varied”. They were also asked: “In the last week, have you been expressing breast milk?” The women who had expressed in the last week were asked “on average, how often do you express?” and could select one of the following options: “once this week”, “several times this week”, “once a day”; “several times a day” or “it varies” .
At all time-points postpartum, participants were asked whether they had any nipple damage. If they had, they could define this damage as a small graze/crack (<2 mm in length), a moderate graze/crack (2 to 9 mm in length) or a severe graze/crack (≥10 mm &/or yellow colour present).
Participants were given the opportunity to report any problems they had with breastfeeding. At weeks 1, 2, 3, 4 and 8 postpartum women were asked whether they had any problems with breastfeeding since their last interview. They were supplied with a list of problems, including attachment issues, over or under-supply of breast milk, use of nipple shield to feed, having an unsettled infant, or an infant not interested in feeding. Women were asked about nipple pain and whether they had used any creams or ointments on their nipples, including use of hydrated polymer (hydrogel) dressings (results have been published separately) [30, 31]. They could also state a breastfeeding problem which was not listed.
There is no standard way of diagnosing mastitis. Therefore, we asked about a range of breast symptoms and associated fever or flu-like symptoms, as has been used previously [1, 32]. For this analysis the definition of mastitis was the development of at least two breast symptoms (pain, redness, lump) and at least one systemic symptom (fever or flu-like symptoms) [1, 32] or treatment of mastitis with antibiotics. Women were asked to rate how mastitis had interfered with breastfeeding and with activities of daily living on a scale of 0 to 4 where 0 was “no interference” and 4 was “quite a bit of interference” .
If a participant ceased breastfeeding during the first four weeks postpartum, the subsequent home visits were not conducted but the participant was contacted at eight weeks to complete the final questionnaire by telephone.
Women provided nasal and nipple (both nipples) swabs and breast milk samples (both breasts). Nasal and oral swabs were obtained from infants. In addition, all women were given a 70 ml sterile container (Sarstedt Australia Pty. Ltd.) and asked to collect a sample of milk from the affected breast should they develop mastitis during the first eight weeks postpartum between their scheduled weekly visits. Samples were collected and cultivated for S. aureus as described in the study protocol .
Descriptive analysis included maternal demographic characteristics and details about the birth, breastfeeding duration, breastfeeding difficulties, and episodes of mastitis. Using information from the successive postpartum time-points, we investigated possible predictors of (or risk factors for) mastitis incidence. We defined “time at risk” of mastitis as days between birth and first occurrence of mastitis (for women who developed mastitis) and days between birth and the last study time-point (for women who did not develop mastitis). We used a discrete version of the multivariate proportional hazards regression model  to investigate whether mothers’ reported nipple damage, “over-production of milk”, breastfeeding attachment problems, use of nipple shield to feed, milk expressing, use of hydrogel dressings, and presence of S. aureus in samples collected from swabs and breast milk samples were risk factors for mastitis incidence if they occurred during the time at risk. For each of the risk factors investigated (mothers’ reported nipple damage, “over-production of milk”, etc), we estimated an incidence rate ratio: the incidence rate of mastitis in those WITH the risk factor during the “time at risk” divided by the incidence rate of mastitis in those WITHOUT the risk factor during the “time at risk”. We used the glm command in Stata with link function = cloglog, family = binomial to estimate the rate ratios. This method uses time at risk and time-varying risk factors. Comparisons are presented using incident rate ratios, 95 % confidence intervals (CIs), and p-values. All analyses were carried out using Stata 13 software .
The study received approval from La Trobe University Human Ethics Committee (LTU UHEC No. 06-078), the RWH Human Research Ethics Committee (RWH HREC No. 06-41) and the Medical Advisory Committee at FPH. All participants provided written informed consent.