- Email: [email protected]
The value of a pilot study in breast-feeding research
ARTICLE IN PRESS Midwifery (2004) 20, 188–193
www.elsevier.com/locate/midw
The value of a pilot study in breast-feeding research Sue Carfoot, BA (Hons), RM, (Midwife)a,*, Paula R Williamson, PhD (Reader in Medical Statistics)b, Rumona Dickson, MSc (Lecturer)c a
Women’s Health Directorate, Croft Wing, North Cheshire Hospitals NHS Trust, Lovely Lane, Warrington, Cheshire WA1 5QG, UK b School of Health Sciences, Centre for Medical Statistics and Health Evaluation, University of Liverpool, Liverpool L69 3BX, UK c Department of Nursing, University of Liverpool, Liverpool L69 3BX, UK Received 13 January 2003; received in revised form 13 May 2003, 28 July 2003; accepted 19 November 2003
Summary Objective: to test the integrity of a protocol for a randomised controlled trial (RCT) to examine the effectiveness of skin-to-skin care compared to routine care on the initiation and duration of breast feeding and to provide data to be used in the power calculation for a proposed trial. Design: randomised pilot study. Setting: Warrington Hospital, Cheshire, UK. Participants: women at 36 weeks’ gestation with healthy singleton pregnancies, who intended to breast feed, who had ‘booked’ for care at Warrington Hospital and had given informed consent to participate. Twenty-eight women were randomised in the pilot study. Method: women were randomly allocated to receive either routine or skin-to-skin care following birth. The first breast feed was assessed using the Breast-feeding Assessment Tool (BAT). Mothers were followed up at discharge from hospital and again at four months to provide details of duration of breast feeding. Findings: 66 women were approached to participate in the trial and 44 consented (67% consent rate). Twenty-eight women were randomised in the study and 26 breast feeds were observed (93%). The pilot study identified procedural changes that were required in the design of the main study, provided an estimate of recruitment rates and confirmed the previously calculated sample size. Conclusion: the pilot study demonstrated that a large RCT of skin-to-skin versus routine care was feasible. This is an example of how a pilot study has the ability to identify unforeseen challenges in the conduct of the trial as well as allowing necessary changes to be made to the design that will increase the quality of the subsequent research. & 2004 Elsevier Ltd. All rights reserved.
*Corresponding author. Tel.: þ 44-7970-7572-42; fax: þ 44-1925-6623-84. E-mail address: [email protected] (S. Carfoot). 0266-6138/$ - see front matter & 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.midw.2003.11.002
ARTICLE IN PRESS The value of a pilot study in breast-feeding research
Introduction The UNICEF UK Baby Friendly Hospital Initiative (WHO, 1989) includes a programme to encourage best practice in breast feeding in the form of the Ten Steps to Successful Breast Feeding. However, one of these steps, encouraging prolonged and uninterrupted skin-to-skin contact for all healthy newborn babies, has not been shown in existing reviews to have an effect on initiation or duration of breast feeding (Bernard-Bonnin et al., 1989; Perez-Escamilla et al., 1994; Fairbank et al., 2000) and is recommended on the basis of limited evidence (WHO, 1998). The authors previously carried out a systematic review (Carfoot et al., 2002), however no conclusions could be made regarding the effectiveness of skin-to-skin care due to the lack of sufficient high quality randomised controlled trials (RCTs). Furthermore, implementation of this practice may require significant changes in practice and many maternity units have found it difficult to successfully integrate skin-to-skin care into routine practice (Rapley, 2002). Given the limited evidence of effectiveness of skin-to-skin contact, the authors decided to conduct an RCT to assess the effect of skin-to-skin contact versus routine care on the first breastfeeding experience and duration of breast feeding. In planning this trial there were a number of uncertainties: the feasibility of the study, the sample size required, the consent rate and the number of research assistants required to provide 24-h cover. The authors decided to undertake a pilot study to test as many elements of the research proposal as possible. The literature on the methods for pilot studies in general (Lancaster et al., 2004) and in midwifery research in particular (Hundley and van Teijlingen, 2002) is limited. This pilot study is presented as an illustration of its ability to inform the design of the main study.
Methods North Cheshire Research Ethics Committee approved the study. Access to undertake the study was provided by the obstetricians and senior managers within the Trust.
Definitions of care Routine care (control group) At birth babies were quickly dried and wrapped in a towel before being handed to their mother or
189
father. Mother/baby contact was interrupted for weighing, dressing and measuring the baby, or for suturing the mother’s perineum after delivery. The midwife offered assistance with the first feed when both mother and baby were ready. Skin-to-skin care (intervention group) Mothers were given their babies to hold with skinto-skin contact as soon as possible following birth. Where babies were delivered by caesarean section, prolonged skin-to-skin contact began in the recovery room. Babies were placed prone, between the mother’s breasts and covered with a warm blanket or towel. Both mother and baby were covered with a blanket. Mother/baby pairs were left undisturbed until the baby showed signs of readiness to feed and the mother was ready to feed her baby. The midwife offered assistance with the first feed. The specific objectives of the pilot study were: to test the procedure for identifying women eligible to participate; to confirm enrolment rates; to test the randomisation procedure; to inform sample size calculations; to test data collection; and to confirm the number of research assistants needed to provide 24 h on-call cover. Testing the procedure for identifying women eligible to participate Women were identified through the existing computer system at 36 weeks gestation. At the time that this study was undertaken this approach complied with the requirements of the Data Protection Act. All women who fulfilled the inclusion criteria were contacted by telephone by the clinical co-ordinator to provide information and asked if they were willing to be part of the trial. Women were eligible to participate if they were healthy, more than 36 weeks pregnant, there was no known abnormality in the fetus, intended to breast feed and were ‘booked’ for maternity care at Warrington Hospital. They were unable to participate if they requested either skin-to-skin or no skin-to-skin contact after birth or were carrying a multiple pregnancy. Eligible women were provided with verbal information about the trial and asked for their consent to participate. They were also informed that the clinical co-ordinator would attend the delivery and observe the first breast feed. Women who agreed to participate in the trial were sent a written information leaflet about the trial, two consent forms to sign (one of which was returned in a stamped addressed envelope) and a sticker to put on the front of their hand-held case notes. The clinical co-ordinator put a sticker on their hospital case notes.
ARTICLE IN PRESS 190
Confirmation of enrolment rates The hospital had approximately 3200 deliveries per year and approximately 52% of women intended to breast feed. It was anticipated that 50% of mothers would fulfil the inclusion criteria and 50% would give consent to participate in the study. Testing the randomisation procedure When the study participant was close to delivery, i.e. second stage of labour had been confirmed, a call was made to the clinical co-ordinator to attend the delivery in order to observe the first breast feed. A sealed envelope was opened to reveal the treatment group for the trial participant, and the midwife responsible for care was also informed. The sealed envelopes were initially held in a box in labour ward. Early in the pilot study, two envelopes were opened in the labour ward but the clinical coordinator was unable to attend the deliveries and observe the first breast feeds. To better control randomisation the clinical co-ordinator subsequently took responsibility for the sealed envelopes. When she had confirmed her ability to attend, she opened the sealed envelope to reveal the treatment group for the trial participant, and informed the midwife responsible for care. Informing sample size requirements The baby’s first breast feed was assessed using the Breast-feeding Assessment Tool (BAT). The purpose of the tool was to systematically assess and score feeding at the breast. A validated tool to assess the effectiveness of this first feed had not been identified. The tool used in the pilot study (available from the first author on request was adapted by the clinical co-ordinator from the Infant Breast feeding Assessment Tool (IBFAT). This IBFAT was developed and tested by Matthews (1988) and is a short assessment of baby breast-feeding competence measuring readiness to feed, rooting and suckling. The IBFAT also measures the mother’s perception of and satisfaction with the feeding. An assessment is completed by the mother after each breast feed and at least two breast feeds are observed by a researcher, to score the IBFAT for inter-rater reliability and to validate the mothers’ observations. However, Matthews (1988) found that inter-rater reliability between mothers and researchers’ assessments was 91%. The clinical co-ordinator adapted the IBFAT to include a ‘latching-on’ component since a correct latch is crucial to effective breast feeding. The Breast-feeding Assessment Tool (BAT) therefore included specific criteria that form the key components of breast feeding: readiness to feed, rooting
S. Carfoot et al.
and hand movements, latching on to the breast and suckling. The authors considered the amended tool to have face validity. The range of scores for each of the four components is 0–3. A total score can thus range from 0–12. The authors agreed that the baby should be able to lose one point in each category and still be considered to have had a successful feed. A loss of one point in category 1 and category 2 would suggest that the baby is quiet and alert and needs encouragement to ‘root’ for the breast. However, if the baby latches on to the breast correctly after a few attempts and begins to suckle fairly well (thus losing a point from categories 3 and 4), it is possible for the baby to breast-feed successfully. A ‘successful breast feed’ was defined, a priori by the authors as, a BAT tool score in the range 8–12 points. The sample size for the main study was based on the assumption that between 40% and 50% of the routine care group would have a successful first feed. This estimate was based on the experience of the clinical co-ordinator, who as part of her role collected breast-feeding statistics for the hospital. This needed to be confirmed through formal assessment using the tool. Pilot of data collection The clinical co-ordinator collected demographic, labour and birth history data (past and current pregnancies) from the records. Post delivery data collection was carried out prior to hospital discharge, through completion of a questionnaire with the clinical co-ordinator. Post discharge data collection was carried out through telephone contact with the mothers at four months and included information on duration of breast feeding. The decision to follow-up by telephone was based on an assumption that personal contact would ensure a higher response rate. Data were entered in a database specifically designed in conjunction with a database programmer. Confirmation of the number of research assistants necessary to provide 24-h on-call cover The day-to-day activities of the pilot study were managed by the clinical co-ordinator and she provided 24 h on-call cover for all trial participants.
Pilot study sample size The timescale for the proposed RCT allowed 2 months for the pilot study. It was anticipated that approximately 30 mother/baby pairs could be recruited in this time.
ARTICLE IN PRESS The value of a pilot study in breast-feeding research
Data analysis As this was a pilot study, no formal hypothesis testing was undertaken. However, the characteristics of the participants have been described using means and ranges for quantitative variables and proportions for categorical variables. The reasons for non-participation in the study are also listed.
Findings Eligible women were easily identified through the existing computer system and telephone contact was made by the clinical co-ordinator. The consent rate was higher than anticipated: 66 women were asked if they were willing to participate in the pilot study and 44 (67%) women consented to participate. Four of the 22 women who declined to participate wanted to have routine contact with their baby following delivery, two women had planned to have a home birth, three women had changed their feeding intention and wished to artificially feed, and the remaining 13 preferred to have skin-to-skin contact with their babies following birth. These 13 women had attended antenatal classes and early postnatal care, including skin-toskin contact, had been discussed with a midwife. None of the reasons implied that this trial should not proceed. Twelve women who agreed to participate in the pilot were not randomised. In six cases the clinical co-ordinator was not available to attend the delivery, in five because the midwife responsible for care of the trial participant failed to call the clinical co-ordinator and in one case because the time from notification of the admission of the participant was too short for the clinical coordinator to get to the hospital. A further two participants withdrew from the study before randomisation. In one case, the Table 1
191
mother had an emergency caesarean section and did not feel well enough to participate. In the other case, the mother felt that it was inappropriate to call the clinical co-ordinator during the night to attend the delivery. Two more participants were not included because researchers suspended the pilot study during the Christmas period. The remaining 28 women were randomised. A midwife randomised two women, at a time when the clinical co-ordinator was not available. As a result, it was agreed that the clinical co-ordinator should keep hold of the box containing the randomisation envelopes and that participants would only be randomised when it was possible for the first breast feed to be observed. In the control group, all 14 mother/baby pairs received routine care. In two cases the babies were offered a breast feed immediately following birth, but the rest were offered breast feeds when the mothers and midwives were ready. Four of the routine care babies were not offered a breast feed when they were ready (as observed by the clinical co-ordinator) and did not feed well on their first attempt. However, three of these babies appeared to be feeding well before discharge from hospital. In the intervention group, all 14 mother/baby pairs received skin-to-skin care and were offered help with breast feeding when the mothers were ready and the babies showed signs of readiness to feed. In total, 14 (50%) of the babies observed fed for longer than one hour from one or both breasts. The baseline characteristics of the two groups are shown in Table 1. This indicates why formal hypothesis testing in pilot studies should be treated with caution. Pilot studies tend to be conducted with small numbers of participants, which is more likely to lead to chance imbalances in characteristics. The success rate of 62% for routine care (Table 2) was slightly higher than the a priori estimate of (50%) In retrospect it was felt that perhaps the involvement of the clinical co-ordinator may have
Description of participants.
Characteristics
Primigravida Attended ante-natal classes Male baby Vaginal delivery Pethidine Epidural/spinal Previous breast feeding experience (multips): mean duration, days Age: mean, years
Skin-to-skin n ¼ 14
Routine n ¼ 14
N
%
N
%
9 10 7 10 4 5
64 71 50 77 29 36 217 31
6 12 7 9 9 6
43 86 50 75 64 43 109 30
ARTICLE IN PRESS 192
S. Carfoot et al.
resulted in this increase. The original estimated sample size for the trial was based on an assumption of a 25% increase in breast-feeding success of first feed, from 50% to 75% period. A change in these estimates to using a base success rate of 60% (maintaining a 25% improvement) would have resulted in a smaller sample size for the main trial. Therefore, it was decided to keep the original sample size. The four-month follow-up data were collected for 26/28 (93%) participants (Table 3). Complete follow-up data were only available for 26 (87%) of the 30 randomised women and therefore it was decided to allow for 20% loss to follow-up in the main trial. The planned sample size of 200 for the main trial was therefore confirmed by the pilot outcome data. The pilot study confirmed that it would not be logistically possible for the clinical co-ordinator to provide 24-h on-call cover for the main trial. Assessment of the workload prompted the research team to employ a part-time assistant to help the clinical co-ordinator with recruitment, data collection and data entry during daytime hours. In addition, it was estimated that six research assistants would be required to provide cover for 12 h during the night, in order to cover the entire week. Provision of 24-h on-call cover would ensure that the recruitment and observation rate remained high throughout the study. Given this new strategy, it was estimated that recruitment for the main trial would take six months.
Table 2 8–12).
Success of first breast feed (BAT score
Skin-to-skin Routine
Table 3
Success rate
Failure rate
n
%
n
%
13 8
100 62
0 5
0 38
Discussion It was important to consider the reasons for nonparticipation in the pilot study to ensure that there was no major objection to the study and to ensure that the consent rate for the main trial would be high. Our consent rate of 67% indicates that women were in fact quite eager to participate in the study. The primary reason to decline was a preference for skin-to-skin care. Although the success rate in the skin-to-skin group was 100% (Table 2), the main trial is justified because the pilot study was small and there were imbalances between the groups that could have biased the result. Experiences in the pilot study led to several changes to the design of the main trial. First, all breast feeds in the pilot study were observed by the clinical co-ordinator who was experienced in the practice of skin-to-skin contact and breast feeding. In the main trial, it was planned that there would be research assistants (observers), who would be trained in observing the success of the first breast feed and use of the BAT tool. The addition of research assistants would provide 24-h cover. Due to the subjective nature of the primary outcome assessment, it could be assumed that the involvement of different observers might decrease observer bias. However, the use of multiple observers decreases the chances of inter-rater reliability. A training course for observers in the use of the BAT tool and an assessment of inter-rater reliability was planned prior to the main trial. For the purpose of the pilot study, it was agreed that we would observe the entire breast feed. However, for the purpose of the main trial, it was decided that a period of 45 min would be sufficient time to assess the success of the first breast feed because BAT scores were completed during the early part of the feed. For the main trial it was decided that if babies were feeding for longer then this would be documented on the form but observation discontinued. It was also evident to the clinical coordinator during the pilot study that several babies from the ‘routine care’ group (three out of four
Feeding at 4 months post birth.
Skin-to-skin Routine
Exclusive breast feeding
Mixed feeding
Artificial feeding
N
%
N
%
N
%
5 4
36 33
2 1
14 8
7 7
50 59
N.B. two participants from routine group could not be contacted and are not included in the four-month follow up.
ARTICLE IN PRESS The value of a pilot study in breast-feeding research
who had scored poorly at the first breast-feeding assessment) were feeding successfully within 24 h. It was therefore felt that it was appropriate to complete another breast-feeding assessment before mother and baby were discharged from hospital. However, it was decided that the success of the first feed would remain the primary outcome and the quality of the feeding prior to discharge would be included as a secondary outcome. The breast-feeding rates in the pilot study confirmed that the original sample size calculations were appropriate. Valuable lessons can be learnt from pilot studies. It has been reported from other studies that valuable lessons can be learnt from conducting pilot studies. For instance, in a pilot study conducted in preparation for the Scottish Births Survey, a national survey of maternity services in Scotland by van Teijlingen et al. (2001), a number of issues were identified. These included, the method of distributing the questionnaire, gaining access to participants and reliance on ‘gatekeepers’. The information gained during this pilot study was deemed to be invaluable because it demonstrated the feasibility of a large RCT. It identified several areas where the main trial protocol needed to be amended and enabled the researchers to test the Breast-feeding Assessment Tool and the data collection forms. The pilot study also provided an estimate of the recruitment rate. Most importantly the pilot study identified procedural issues that may have proved costly had they not been changed prior to the commencement of the main trial (i.e. the process of randomisation). Pilot studies should be recommended as routine practice within the protocol of conducting any RCT.
193
Acknowledgements Funding for this project has been given from the North West Executive Regional R&D Directorate.
References Bernard-Bonnin, A., Stachtenko, S., Girard, G., Rousseau, E., 1989. Hospital practices and breast feeding duration: metaanalysis of controlled trials. Birth 16, 64–66. Carfoot, S., Williamson, P., Dickson, R., 2002. Successful breastfeeding: a systematic review of randomised controlled trials evaluating the effect of mother/baby skin-to-skin care. Midwifery 19, 148–155. Fairbank, L., O’ Meara, S., Renfrew, M., Woolridge, M., Sowden, A., Lister-Sharp, D., 2000. A systematic review to evaluate the effectiveness of interventions to promote the initiation of breast feeding. Health Technology Assessment 4, 25. Hundley, V., van Teijlingen, E., 2002. The role of pilot studies in midwifery research. Midwives 5, 372–374. Lancaster, G., Hinds, S., Williamson, P.R., 2004. Use of pilot studies in medical research: what is a pilot study? Journal of Evaluation in Clinical Practice 10 (2). Matthews, M.K., 1988. Developing an instrument to assess baby breast feeding behaviour in the early postnatal period. Midwifery 4, 154–165. Perez-Escamilla, R., Pollitt, E., Lonnerdal, B., Dewey, K., 1994. Baby feeding policies in maternity wards and their effect on breast-feeding success: an analytical overview. American Journal of Public Health 84 (1), 89–96. Rapley, G., 2002. Keeping mothers and babies together- breastfeeding and bonding. Midwives 5, 332–334. Van Teijlingen, E.R., Rennie, A.M., Hundley, V., Graham, W., 2001. The importance of conducting and reporting pilot studies: the example of the Scottish births survey. Journal of Advanced Nursing 34 (3), 289–295. WHO, 1989. Protecting, promoting and supporting breastfeeding. The special role of maternity services. A joint WHO/UNICEF statement. Geneva. WHO, 1998. Evidence for the ten steps to successful breast feeding. Division of Child Health. World Health Organisation, Geneva.
www.elsevier.com/locate/midw
The value of a pilot study in breast-feeding research Sue Carfoot, BA (Hons), RM, (Midwife)a,*, Paula R Williamson, PhD (Reader in Medical Statistics)b, Rumona Dickson, MSc (Lecturer)c a
Women’s Health Directorate, Croft Wing, North Cheshire Hospitals NHS Trust, Lovely Lane, Warrington, Cheshire WA1 5QG, UK b School of Health Sciences, Centre for Medical Statistics and Health Evaluation, University of Liverpool, Liverpool L69 3BX, UK c Department of Nursing, University of Liverpool, Liverpool L69 3BX, UK Received 13 January 2003; received in revised form 13 May 2003, 28 July 2003; accepted 19 November 2003
Summary Objective: to test the integrity of a protocol for a randomised controlled trial (RCT) to examine the effectiveness of skin-to-skin care compared to routine care on the initiation and duration of breast feeding and to provide data to be used in the power calculation for a proposed trial. Design: randomised pilot study. Setting: Warrington Hospital, Cheshire, UK. Participants: women at 36 weeks’ gestation with healthy singleton pregnancies, who intended to breast feed, who had ‘booked’ for care at Warrington Hospital and had given informed consent to participate. Twenty-eight women were randomised in the pilot study. Method: women were randomly allocated to receive either routine or skin-to-skin care following birth. The first breast feed was assessed using the Breast-feeding Assessment Tool (BAT). Mothers were followed up at discharge from hospital and again at four months to provide details of duration of breast feeding. Findings: 66 women were approached to participate in the trial and 44 consented (67% consent rate). Twenty-eight women were randomised in the study and 26 breast feeds were observed (93%). The pilot study identified procedural changes that were required in the design of the main study, provided an estimate of recruitment rates and confirmed the previously calculated sample size. Conclusion: the pilot study demonstrated that a large RCT of skin-to-skin versus routine care was feasible. This is an example of how a pilot study has the ability to identify unforeseen challenges in the conduct of the trial as well as allowing necessary changes to be made to the design that will increase the quality of the subsequent research. & 2004 Elsevier Ltd. All rights reserved.
*Corresponding author. Tel.: þ 44-7970-7572-42; fax: þ 44-1925-6623-84. E-mail address: [email protected] (S. Carfoot). 0266-6138/$ - see front matter & 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.midw.2003.11.002
ARTICLE IN PRESS The value of a pilot study in breast-feeding research
Introduction The UNICEF UK Baby Friendly Hospital Initiative (WHO, 1989) includes a programme to encourage best practice in breast feeding in the form of the Ten Steps to Successful Breast Feeding. However, one of these steps, encouraging prolonged and uninterrupted skin-to-skin contact for all healthy newborn babies, has not been shown in existing reviews to have an effect on initiation or duration of breast feeding (Bernard-Bonnin et al., 1989; Perez-Escamilla et al., 1994; Fairbank et al., 2000) and is recommended on the basis of limited evidence (WHO, 1998). The authors previously carried out a systematic review (Carfoot et al., 2002), however no conclusions could be made regarding the effectiveness of skin-to-skin care due to the lack of sufficient high quality randomised controlled trials (RCTs). Furthermore, implementation of this practice may require significant changes in practice and many maternity units have found it difficult to successfully integrate skin-to-skin care into routine practice (Rapley, 2002). Given the limited evidence of effectiveness of skin-to-skin contact, the authors decided to conduct an RCT to assess the effect of skin-to-skin contact versus routine care on the first breastfeeding experience and duration of breast feeding. In planning this trial there were a number of uncertainties: the feasibility of the study, the sample size required, the consent rate and the number of research assistants required to provide 24-h cover. The authors decided to undertake a pilot study to test as many elements of the research proposal as possible. The literature on the methods for pilot studies in general (Lancaster et al., 2004) and in midwifery research in particular (Hundley and van Teijlingen, 2002) is limited. This pilot study is presented as an illustration of its ability to inform the design of the main study.
Methods North Cheshire Research Ethics Committee approved the study. Access to undertake the study was provided by the obstetricians and senior managers within the Trust.
Definitions of care Routine care (control group) At birth babies were quickly dried and wrapped in a towel before being handed to their mother or
189
father. Mother/baby contact was interrupted for weighing, dressing and measuring the baby, or for suturing the mother’s perineum after delivery. The midwife offered assistance with the first feed when both mother and baby were ready. Skin-to-skin care (intervention group) Mothers were given their babies to hold with skinto-skin contact as soon as possible following birth. Where babies were delivered by caesarean section, prolonged skin-to-skin contact began in the recovery room. Babies were placed prone, between the mother’s breasts and covered with a warm blanket or towel. Both mother and baby were covered with a blanket. Mother/baby pairs were left undisturbed until the baby showed signs of readiness to feed and the mother was ready to feed her baby. The midwife offered assistance with the first feed. The specific objectives of the pilot study were: to test the procedure for identifying women eligible to participate; to confirm enrolment rates; to test the randomisation procedure; to inform sample size calculations; to test data collection; and to confirm the number of research assistants needed to provide 24 h on-call cover. Testing the procedure for identifying women eligible to participate Women were identified through the existing computer system at 36 weeks gestation. At the time that this study was undertaken this approach complied with the requirements of the Data Protection Act. All women who fulfilled the inclusion criteria were contacted by telephone by the clinical co-ordinator to provide information and asked if they were willing to be part of the trial. Women were eligible to participate if they were healthy, more than 36 weeks pregnant, there was no known abnormality in the fetus, intended to breast feed and were ‘booked’ for maternity care at Warrington Hospital. They were unable to participate if they requested either skin-to-skin or no skin-to-skin contact after birth or were carrying a multiple pregnancy. Eligible women were provided with verbal information about the trial and asked for their consent to participate. They were also informed that the clinical co-ordinator would attend the delivery and observe the first breast feed. Women who agreed to participate in the trial were sent a written information leaflet about the trial, two consent forms to sign (one of which was returned in a stamped addressed envelope) and a sticker to put on the front of their hand-held case notes. The clinical co-ordinator put a sticker on their hospital case notes.
ARTICLE IN PRESS 190
Confirmation of enrolment rates The hospital had approximately 3200 deliveries per year and approximately 52% of women intended to breast feed. It was anticipated that 50% of mothers would fulfil the inclusion criteria and 50% would give consent to participate in the study. Testing the randomisation procedure When the study participant was close to delivery, i.e. second stage of labour had been confirmed, a call was made to the clinical co-ordinator to attend the delivery in order to observe the first breast feed. A sealed envelope was opened to reveal the treatment group for the trial participant, and the midwife responsible for care was also informed. The sealed envelopes were initially held in a box in labour ward. Early in the pilot study, two envelopes were opened in the labour ward but the clinical coordinator was unable to attend the deliveries and observe the first breast feeds. To better control randomisation the clinical co-ordinator subsequently took responsibility for the sealed envelopes. When she had confirmed her ability to attend, she opened the sealed envelope to reveal the treatment group for the trial participant, and informed the midwife responsible for care. Informing sample size requirements The baby’s first breast feed was assessed using the Breast-feeding Assessment Tool (BAT). The purpose of the tool was to systematically assess and score feeding at the breast. A validated tool to assess the effectiveness of this first feed had not been identified. The tool used in the pilot study (available from the first author on request was adapted by the clinical co-ordinator from the Infant Breast feeding Assessment Tool (IBFAT). This IBFAT was developed and tested by Matthews (1988) and is a short assessment of baby breast-feeding competence measuring readiness to feed, rooting and suckling. The IBFAT also measures the mother’s perception of and satisfaction with the feeding. An assessment is completed by the mother after each breast feed and at least two breast feeds are observed by a researcher, to score the IBFAT for inter-rater reliability and to validate the mothers’ observations. However, Matthews (1988) found that inter-rater reliability between mothers and researchers’ assessments was 91%. The clinical co-ordinator adapted the IBFAT to include a ‘latching-on’ component since a correct latch is crucial to effective breast feeding. The Breast-feeding Assessment Tool (BAT) therefore included specific criteria that form the key components of breast feeding: readiness to feed, rooting
S. Carfoot et al.
and hand movements, latching on to the breast and suckling. The authors considered the amended tool to have face validity. The range of scores for each of the four components is 0–3. A total score can thus range from 0–12. The authors agreed that the baby should be able to lose one point in each category and still be considered to have had a successful feed. A loss of one point in category 1 and category 2 would suggest that the baby is quiet and alert and needs encouragement to ‘root’ for the breast. However, if the baby latches on to the breast correctly after a few attempts and begins to suckle fairly well (thus losing a point from categories 3 and 4), it is possible for the baby to breast-feed successfully. A ‘successful breast feed’ was defined, a priori by the authors as, a BAT tool score in the range 8–12 points. The sample size for the main study was based on the assumption that between 40% and 50% of the routine care group would have a successful first feed. This estimate was based on the experience of the clinical co-ordinator, who as part of her role collected breast-feeding statistics for the hospital. This needed to be confirmed through formal assessment using the tool. Pilot of data collection The clinical co-ordinator collected demographic, labour and birth history data (past and current pregnancies) from the records. Post delivery data collection was carried out prior to hospital discharge, through completion of a questionnaire with the clinical co-ordinator. Post discharge data collection was carried out through telephone contact with the mothers at four months and included information on duration of breast feeding. The decision to follow-up by telephone was based on an assumption that personal contact would ensure a higher response rate. Data were entered in a database specifically designed in conjunction with a database programmer. Confirmation of the number of research assistants necessary to provide 24-h on-call cover The day-to-day activities of the pilot study were managed by the clinical co-ordinator and she provided 24 h on-call cover for all trial participants.
Pilot study sample size The timescale for the proposed RCT allowed 2 months for the pilot study. It was anticipated that approximately 30 mother/baby pairs could be recruited in this time.
ARTICLE IN PRESS The value of a pilot study in breast-feeding research
Data analysis As this was a pilot study, no formal hypothesis testing was undertaken. However, the characteristics of the participants have been described using means and ranges for quantitative variables and proportions for categorical variables. The reasons for non-participation in the study are also listed.
Findings Eligible women were easily identified through the existing computer system and telephone contact was made by the clinical co-ordinator. The consent rate was higher than anticipated: 66 women were asked if they were willing to participate in the pilot study and 44 (67%) women consented to participate. Four of the 22 women who declined to participate wanted to have routine contact with their baby following delivery, two women had planned to have a home birth, three women had changed their feeding intention and wished to artificially feed, and the remaining 13 preferred to have skin-to-skin contact with their babies following birth. These 13 women had attended antenatal classes and early postnatal care, including skin-toskin contact, had been discussed with a midwife. None of the reasons implied that this trial should not proceed. Twelve women who agreed to participate in the pilot were not randomised. In six cases the clinical co-ordinator was not available to attend the delivery, in five because the midwife responsible for care of the trial participant failed to call the clinical co-ordinator and in one case because the time from notification of the admission of the participant was too short for the clinical coordinator to get to the hospital. A further two participants withdrew from the study before randomisation. In one case, the Table 1
191
mother had an emergency caesarean section and did not feel well enough to participate. In the other case, the mother felt that it was inappropriate to call the clinical co-ordinator during the night to attend the delivery. Two more participants were not included because researchers suspended the pilot study during the Christmas period. The remaining 28 women were randomised. A midwife randomised two women, at a time when the clinical co-ordinator was not available. As a result, it was agreed that the clinical co-ordinator should keep hold of the box containing the randomisation envelopes and that participants would only be randomised when it was possible for the first breast feed to be observed. In the control group, all 14 mother/baby pairs received routine care. In two cases the babies were offered a breast feed immediately following birth, but the rest were offered breast feeds when the mothers and midwives were ready. Four of the routine care babies were not offered a breast feed when they were ready (as observed by the clinical co-ordinator) and did not feed well on their first attempt. However, three of these babies appeared to be feeding well before discharge from hospital. In the intervention group, all 14 mother/baby pairs received skin-to-skin care and were offered help with breast feeding when the mothers were ready and the babies showed signs of readiness to feed. In total, 14 (50%) of the babies observed fed for longer than one hour from one or both breasts. The baseline characteristics of the two groups are shown in Table 1. This indicates why formal hypothesis testing in pilot studies should be treated with caution. Pilot studies tend to be conducted with small numbers of participants, which is more likely to lead to chance imbalances in characteristics. The success rate of 62% for routine care (Table 2) was slightly higher than the a priori estimate of (50%) In retrospect it was felt that perhaps the involvement of the clinical co-ordinator may have
Description of participants.
Characteristics
Primigravida Attended ante-natal classes Male baby Vaginal delivery Pethidine Epidural/spinal Previous breast feeding experience (multips): mean duration, days Age: mean, years
Skin-to-skin n ¼ 14
Routine n ¼ 14
N
%
N
%
9 10 7 10 4 5
64 71 50 77 29 36 217 31
6 12 7 9 9 6
43 86 50 75 64 43 109 30
ARTICLE IN PRESS 192
S. Carfoot et al.
resulted in this increase. The original estimated sample size for the trial was based on an assumption of a 25% increase in breast-feeding success of first feed, from 50% to 75% period. A change in these estimates to using a base success rate of 60% (maintaining a 25% improvement) would have resulted in a smaller sample size for the main trial. Therefore, it was decided to keep the original sample size. The four-month follow-up data were collected for 26/28 (93%) participants (Table 3). Complete follow-up data were only available for 26 (87%) of the 30 randomised women and therefore it was decided to allow for 20% loss to follow-up in the main trial. The planned sample size of 200 for the main trial was therefore confirmed by the pilot outcome data. The pilot study confirmed that it would not be logistically possible for the clinical co-ordinator to provide 24-h on-call cover for the main trial. Assessment of the workload prompted the research team to employ a part-time assistant to help the clinical co-ordinator with recruitment, data collection and data entry during daytime hours. In addition, it was estimated that six research assistants would be required to provide cover for 12 h during the night, in order to cover the entire week. Provision of 24-h on-call cover would ensure that the recruitment and observation rate remained high throughout the study. Given this new strategy, it was estimated that recruitment for the main trial would take six months.
Table 2 8–12).
Success of first breast feed (BAT score
Skin-to-skin Routine
Table 3
Success rate
Failure rate
n
%
n
%
13 8
100 62
0 5
0 38
Discussion It was important to consider the reasons for nonparticipation in the pilot study to ensure that there was no major objection to the study and to ensure that the consent rate for the main trial would be high. Our consent rate of 67% indicates that women were in fact quite eager to participate in the study. The primary reason to decline was a preference for skin-to-skin care. Although the success rate in the skin-to-skin group was 100% (Table 2), the main trial is justified because the pilot study was small and there were imbalances between the groups that could have biased the result. Experiences in the pilot study led to several changes to the design of the main trial. First, all breast feeds in the pilot study were observed by the clinical co-ordinator who was experienced in the practice of skin-to-skin contact and breast feeding. In the main trial, it was planned that there would be research assistants (observers), who would be trained in observing the success of the first breast feed and use of the BAT tool. The addition of research assistants would provide 24-h cover. Due to the subjective nature of the primary outcome assessment, it could be assumed that the involvement of different observers might decrease observer bias. However, the use of multiple observers decreases the chances of inter-rater reliability. A training course for observers in the use of the BAT tool and an assessment of inter-rater reliability was planned prior to the main trial. For the purpose of the pilot study, it was agreed that we would observe the entire breast feed. However, for the purpose of the main trial, it was decided that a period of 45 min would be sufficient time to assess the success of the first breast feed because BAT scores were completed during the early part of the feed. For the main trial it was decided that if babies were feeding for longer then this would be documented on the form but observation discontinued. It was also evident to the clinical coordinator during the pilot study that several babies from the ‘routine care’ group (three out of four
Feeding at 4 months post birth.
Skin-to-skin Routine
Exclusive breast feeding
Mixed feeding
Artificial feeding
N
%
N
%
N
%
5 4
36 33
2 1
14 8
7 7
50 59
N.B. two participants from routine group could not be contacted and are not included in the four-month follow up.
ARTICLE IN PRESS The value of a pilot study in breast-feeding research
who had scored poorly at the first breast-feeding assessment) were feeding successfully within 24 h. It was therefore felt that it was appropriate to complete another breast-feeding assessment before mother and baby were discharged from hospital. However, it was decided that the success of the first feed would remain the primary outcome and the quality of the feeding prior to discharge would be included as a secondary outcome. The breast-feeding rates in the pilot study confirmed that the original sample size calculations were appropriate. Valuable lessons can be learnt from pilot studies. It has been reported from other studies that valuable lessons can be learnt from conducting pilot studies. For instance, in a pilot study conducted in preparation for the Scottish Births Survey, a national survey of maternity services in Scotland by van Teijlingen et al. (2001), a number of issues were identified. These included, the method of distributing the questionnaire, gaining access to participants and reliance on ‘gatekeepers’. The information gained during this pilot study was deemed to be invaluable because it demonstrated the feasibility of a large RCT. It identified several areas where the main trial protocol needed to be amended and enabled the researchers to test the Breast-feeding Assessment Tool and the data collection forms. The pilot study also provided an estimate of the recruitment rate. Most importantly the pilot study identified procedural issues that may have proved costly had they not been changed prior to the commencement of the main trial (i.e. the process of randomisation). Pilot studies should be recommended as routine practice within the protocol of conducting any RCT.
193
Acknowledgements Funding for this project has been given from the North West Executive Regional R&D Directorate.
References Bernard-Bonnin, A., Stachtenko, S., Girard, G., Rousseau, E., 1989. Hospital practices and breast feeding duration: metaanalysis of controlled trials. Birth 16, 64–66. Carfoot, S., Williamson, P., Dickson, R., 2002. Successful breastfeeding: a systematic review of randomised controlled trials evaluating the effect of mother/baby skin-to-skin care. Midwifery 19, 148–155. Fairbank, L., O’ Meara, S., Renfrew, M., Woolridge, M., Sowden, A., Lister-Sharp, D., 2000. A systematic review to evaluate the effectiveness of interventions to promote the initiation of breast feeding. Health Technology Assessment 4, 25. Hundley, V., van Teijlingen, E., 2002. The role of pilot studies in midwifery research. Midwives 5, 372–374. Lancaster, G., Hinds, S., Williamson, P.R., 2004. Use of pilot studies in medical research: what is a pilot study? Journal of Evaluation in Clinical Practice 10 (2). Matthews, M.K., 1988. Developing an instrument to assess baby breast feeding behaviour in the early postnatal period. Midwifery 4, 154–165. Perez-Escamilla, R., Pollitt, E., Lonnerdal, B., Dewey, K., 1994. Baby feeding policies in maternity wards and their effect on breast-feeding success: an analytical overview. American Journal of Public Health 84 (1), 89–96. Rapley, G., 2002. Keeping mothers and babies together- breastfeeding and bonding. Midwives 5, 332–334. Van Teijlingen, E.R., Rennie, A.M., Hundley, V., Graham, W., 2001. The importance of conducting and reporting pilot studies: the example of the Scottish births survey. Journal of Advanced Nursing 34 (3), 289–295. WHO, 1989. Protecting, promoting and supporting breastfeeding. The special role of maternity services. A joint WHO/UNICEF statement. Geneva. WHO, 1998. Evidence for the ten steps to successful breast feeding. Division of Child Health. World Health Organisation, Geneva.