This was a three-armed randomised, controlled trial. Data were collected at three time points: T1 (6±1 week post breast surgery), T2 and T3 (4 and 10 weeks post completion of CAM, respectively) (Fig. 1).

Ethical approval was obtained from Hull and East Yorkshire Local Research Ethics Committee (reference 01/01/010), and the study was registered with the International Standard Randomized Controlled Trial Registry (ISRCTN 87652313).

Women over 18 years of age with early breast cancer [T1, T2 (<3 cm), N0, N1a, M0], awaiting adjuvant therapy, were recruited consecutively after surgery. A diagnostically homogeneous group of patients was chosen to minimise the effects of disease and stage-related variables (Table I) (45). Patients with a previous cancer diagnosis or more advanced disease were not eligible for recruitment, as were those participating in other clinical trials and those suffering from clinically significant cognitive impairment or dementia.

Patients who gave written informed consent (n=183) were randomised to one of three interventions in the Oncology Health Centres at Castle Hill or Princess Royal Hospitals in Hull: self-initiated support (SIS) plus foot reflexology (n=60), SIS plus scalp massage (identical amount of comparator physical and social contact intervention from the same therapists who administered reflexology; n=61), or SIS alone (treatment as usual; n=62) (44).

A permuted block randomisation sequence for each stratum (menopausal status, chemotherapy and radiotherapy) was generated using Graph-Pad (http://www.graphpad.com); block size was 8 and was concealed. The sequences were stored in sealed, opaque, numbered envelopes. Randomisation was carried out remotely at the Clinical Trials Section of the Institute of Rehabilitation, University of Hull. Biological assessments were carried out in a completely blinded manner.

Patients randomised to reflexology or massage received 8 sessions at weekly intervals for 8 weeks commencing 7 weeks after surgery. Eight sessions at weekly intervals was chosen on the recommendation of an external consultant who was formerly the Secretary of the Scottish Institute of reflexology.

Reflexology was administered by two part-time staff who had been trained to the standards required for membership of the Scottish Institute of Reflexology. Their performance and adherence to the reflexology protocol was monitored at regular intervals during the study by an external consultant experienced in administering reflexology to patients with cancer.

Scalp massage was used as a control for attention, physical contact and non-specific therapist effects. Patients randomised to massage received gentle scalp massage from the same therapists according to a quality assured protocol. Since reflexologists believe that the ears and neck have ‘terminals’, care was taken to avoid these areas. Scalp massage was chosen, rather than foot or hand massage, because any manipulation or pressure to the feet and hands, according to reflexology theory, will stimulate pressure points and, therefore, will be a weak form of reflexology rather than an appropriate ‘placebo’.

Women randomised to SIS were invited to attend, or telephone, one of the Oncology Health Centres whenever they wished. They received ‘treatment as usual’ in the Centres when they attended, as did those randomised to reflexology or massage. The Oncology Health Centres are staffed by clinical health psychologists and nurses, and provide psychosocial support services for more than 1,500 new patients per year and almost as many new relatives. Emphasis is placed on the prevention of psychological and psychiatric morbidity, and evidence-based psychopharmacological and psychotherapeutic interventions are offered to patients who develop clinically significant problems. Patients can access the service without referral, and appointments are not necessary.

In order to control for practitioner variables, each practitioner saw a similar number of patients in each of the two physical contact arms of the study. As far as possible, each patient had the same therapist throughout, and both treatments were given in the same rooms.

All patients underwent conventional treatment according to current best practice (surgery, radiotherapy, chemotherapy and hormone therapy) as clinically indicated.

Venous blood (50 ml) was collected into syringes containing 1,250 IU heparin from all patients at T1, T2 and T3 and transported to the Centre for Biomedical Research at the University of Hull. PBMCs were isolated using Ficoll-Hypaque (Sigma), density gradient centrifugation (46). The PBMCs were washed with phosphate-buffered saline (PBS; pH 7.4), enumerated using a haemocytometer and assessed for viability by trypan blue exclusion, before resuspension in foetal bovine serum (Invitrogen) containing 10% (v/v) dimethylsulphoxide (Sigma). Aliquots were then frozen at 1°C/min and stored in liquid nitrogen until use.

Serum separation. Venous blood (8 ml) was collected into serum separator tubes at approximately the same time of day (to control for diurnal variation) for each patient at T1, T2 and T3. These were incubated at 4°C for 30 min before centrifugation at 1,500 × g for 10 min. The top serum layer was aliquoted and stored at −80°C until ELISA analysis.

Immunophenotyping of peripheral blood mononuclear cells. An aliquot of PBMCs from each of the three time points was resuspended in complete medium [RPMI-1640 medium supplemented with 10% (v/v) foetal bovine serum, penicillin (100 U/ml)/streptomycin (100 μg/ml) and L-glutamine (2 mM), all purchased from Invitrogen]. Approximately 2×10⁵ PBMCs were labelled with 5 μl (0.1 mg/ml) of one of a panel of fluorescein isothiocyanate (FITC)-conjugated monoclonal antibodies (AbD Serotec, Oxford, UK) for 30 min in the dark at room temperature. These antibodies were specific for the surface markers: CD2 and CD3 (T cells), CD4 (T helper cells), CD8 (T cytotoxic cells), CD16 and CD56 (NK cells), CD14 (monocytes), CD19 (B cells) and CD25 (activated lymphocytes/regulatory T cells). Purified mouse IgG1-FITC was used as an irrelevant control. Following labelling, the cells were washed with PBS, pH 7.4, containing 0.1% (w/v) bovine serum albumin and 10 mM NaN₃ (PBS/BSA/azide; Sigma) and recovered by centrifugation before immediate acquisition of 10,000 cells/sample using a FACS Calibur™ machine (Becton Dickinson, Biosciences, Oxford, UK). Analysis was performed using CellQuest Pro V software (Becton Dickinson) with gates being set around the lymphocytes and the monocytes based on forward scatter/side scatter distribution. Histograms were drawn for each antibody using the lymphocyte gate, except for CD14 which used the monocyte gate. The plots using the irrelevant control were used to set a marker whereby ≤3% of cells were positive with this reagent. To calculate the percentage of specific binding, the irrelevant value was subtracted from the percentage of cells staining with the test antibody.

Th1/Th2 cellular determination. The method used was a modification of that by Jung et al (47). Briefly, PBMCs were incubated for 4 h at 37°C with 5% CO₂ in either activation medium [complete RPMI containing ionomycin (2 μg/ml), brefeldin A (20 μg/ml) and phorbol 12-myristate 13-acetate (20 ng/ml), all from Sigma] or control medium [complete RPMI containing solely brefeldin A (20 μg/ml)]. Approximately 2×10⁵ PBMCs, were then incubated for 30 min in the dark with 5 μl (0.1 mg/ml) of FITC labelled monoclonal antibody specific for the surface markers CD3 or CD8; purified mouse IgG1 provided the irrelevant control (AbD Serotec). After incubation, the cells were washed with PBS/BSA/azide, and fixed using Leucoperm A (AbD Serotec) for 15 min in the dark. Following further washes, the cells were permeabilised with Leucoperm B before being incubated with 5 μl r-phycoerythrin (RPE)-labelled anti-cytokine antibody [IL2, IFNγ (Becton Dickinson), IL4, IL10 (AbD Serotec), 0.1 mg/ml] for 30 min at room temperature in the dark. TNFα (Becton Dickinson) provided a positive activation control, and purified mouse IgG1 provided the negative control. Cells were acquired and analysed as described previously using a FACS Calibur™. The lymphocyte subset was gated on the basis of forward and side scatter characteristics (Fig. 2A). Stimulated lymphocytes produced Th1 (IL2/IFNγ) and Th2 (IL4/IL10) cytokines (Fig. 2B). CD3 was used as a total lymphocyte marker, and the CD4 fraction was determined by subtraction of the percentage of CD8⁺ cells. Unstimulated controls (Fig. 2C) were used to set the quadrants so that <1% of these cells were positive for both the surface and the cytokine antibodies (upper right quadrant).

NK and LAK cell cytotoxicity. The cytotoxic activity of both NK and LAK cells within the PBMC population was determined using a modification of the Live/Dead Cell-mediated cytotoxicity kit (Molecular Probes/Invitrogen) (48). Briefly, 2×10⁶ log-phase growing target cells [erythroleukaemic cell line (K562)] for NK and Burkitt lymphoma cell line (Daudi) for LAK] were prelabelled with a green fluorescent membrane dye (DiOC₁₈(3), 30 μM) for 1 h at 37°C. Once thawed PBMCs were incubated for 48 h in complete RPMI medium with or without recombinant IL2 (500 U/ml; AbD Serotec) (49) to stimulate LAK formation and for NK determination respectively; this also allows for the adherence and removal of monocytes. Following washing, viable effector cells were enumerated using trypan blue exclusion, and 1.5 or 3×10⁵ cells were added to 3×10⁴ target cells to give a 5:1 and 10:1 ratio, respectively, in a total volume of 140 μl. An equal volume of the membrane impermeable dye, propidium iodide (PI; 150 μM; Sigma), was added to each tube before centrifuging briefly at 1,000 × g for 1 min and incubation at 37°C for 3.5 h in a humidified atmosphere. Appropriate controls of target and effector cells alone were also prepared. Following incubation, the cytotoxic activity was analysed by flow cytometry; samples were acquired for 45 sec with no gating and data were obtained for both the green (FL-1, DiOC₁₈(3)) and red (FL-3, PI) fluorescence, as well as forward and side scatter characteristics. Data analysis was performed on dot plots of FL-1 vs. FL-3. Quadrants were set using the appropriate controls to exclude the effector cells from the analysis, and the lysis of target cells was determined from the percentage of cells present in the upper right (UR) quadrant (green and red positive, i.e., dead target cells) divided by the total number of green target cells (UR + LR).

Serum hormone measurements. The hormones prolactin, cortisol and growth hormone were all measured in duplicate, using Enzyme Linked Immunosorbant Assay (ELISA, DRG Instruments GmbH, Germany). All samples from the same patient were analysed on the same ELISA plate to minimise intra-patient variability. Prolactin levels were measured using a standard solid phase sandwich ELISA technique, with a lower detection limit of 2 ng/ml, according to the manufacturer’s protocol. Cortisol levels were analysed using a competitive ELISA with a lower detection limit of 2.5 ng/ml, according to the manufacturer’s protocol. Growth hormone levels were determined using a solid phase Enzyme Amplified Sensitivity Immunoassay, in which monoclonal antibodies against distinct epitopes of human growth hormone are used to create the sandwich; with a lower detection limit of 0.11 μIU/ml. Data for each sample were extracted from the standard curve.

Hormone receptor measurements. Following activation of PBMCs as described above, the cells which were to be used for the detection of the prolactin receptor (PRL-R) and the glucocorticoid receptor (Gluc-R) were permeabilised. PBMCs were then incubated for 30 min at 4°C with 5 μl (1 mg/ml) mouse anti-human antibodies: GH-R unconjugated, Gluc-RFITC conjugated (both AbD Serotec) and PRL-R Ab-1 (B6.2) unconjugated (Neomarkers, Fremont, CA). Purified mouse IgG1 was used as the negative control. Following washing with PBS/BSA/azide unconjugated antibodies were detected using a secondary rabbit anti-mouse F(ab’), IgG:FITC antibody (AbD Serotec) for 30 min at 4°C. PBMCs were washed again before flow cytometric analysis.

Data were analysed using SPSS v13 for MS Windows. α was set at 0.05 (two-tailed). The comparability of the three groups at baseline (T1; clinical, socio-demographic and psychosocial variables) was assessed using one-way analyses of variance (ANOVA) for continuous variables, and the Chi-square exact test for categorical variables. All data were included in the analyses of categorical variables, and missing data were not inputed, as cohort averages would not have been appropriate.

An intention to treat analysis was carried out (50), and continuous variables were analysed using univariate analyses of covariance (ANCOVA), with age, tumour stage and baseline (T1) values as covariates. To minimise the risk of a Type I error, paired comparisons were only considered when the f-value for the three-group comparison was significant, and Bonferroni corrections were applied for subsequent paired comparisons. Data were log transformed when the distributions differed significantly from the normal.

A consecutive series of 243 women was assessed for eligibility (Fig. 3). Of the 234 who were eligible, 183 (78.2%) agreed to be randomised. The most common reason for not wishing to participate was lack of interest, often because women expressed a desire to ‘get on with their lives’. Four (1.7%) eligible patients refused randomisation because they did not wish to partake in reflexology.

Sixty patients were randomised to reflexology, 61 to massage and 62 to SIS alone. The characteristics of the women by randomisation are shown in Table I. The three groups did not differ significantly for any of the demographic or clinical variables, including radiotherapy and/or chemotherapy.

The CONSORT diagram (Fig. 3) indicates the samples used for the different biological assays. The number of aliquots of PBMCs obtained from each patient varied, and there were insufficient aliquots from each patient to be used for all techniques. Samples were selected for each analysis on the basis that there were sufficient aliquots of PBMCs present for each of the three time points to enable a full dataset to be collected.

Flow cytometry was used to determine the changes which occurred in the distribution of the mononuclear cell populations (Table II). At T3, ANCOVA showed that the percentage of CD25⁺ lymphocytes in the patients receiving massage was significantly higher than for those in the SIS group (p=0.05). No significant between-group differences were found for the remainder of the phenotypic markers.

The percentages of NK, B cells, CD3⁺ and CD4⁺ T cells were very similar to those previously reported both pre- and post-psychosocial intervention by Carlson et al in a cohort of patients with either breast or prostate cancer (51).

Table III shows the results from the flow cytometry method used to determine the percentage of lymphocytes producing Th1 (IL2/IFNγ) and Th2 (IL4/IL10) cytokines. ANCOVA showed a significantly lower percentage of CD3⁺ cells expressing IL4 at T3 in the massage patients compared with the SIS patients (p=0.02, Fig. 4A). The same was true in the CD4⁺ subset of T cells expressing IL4 which mirrored the results of the CD3⁺ cells (Fig. 4B; p=0.02).

At T3, ANCOVA showed that a significantly higher percentage of CD8⁺ cells were expressing IFNγ in the massage group compared with the SIS group (p=0.02, Fig. 4C). No significant between-group differences were found for any of the other T cell subsets expressing Th1- or Th2-like cytokines.

The NK and LAK cell activity was determined at effector:target ratios of 5:1 and 10:1 using a flow cytometry-based method. These ratios were chosen since they provided the most reproducible results in initial studies and spared sufficient PBMCs for use in other experiments. There were no significant between-group differences in the cytotoxic activity of either NK or LAK cells at any time point (Table IV).

Analysis of the serum hormones or hormone receptors (cortisol, prolactin and growth hormone) also found no significant between-group differences (Tables V and VI). All the values obtained for the serum hormone concentrations were within the normal range described in the manufacturer’s protocols.