This study (the trial registration number: UMIN000003128) was approved by the ethical committee of Katayama Orthopedic Rheumatology Clinic, and a written informed consent was obtained from all patients before performing any study procedures according to the Declaration of Helsinki. Thirty-eight voluntary patients with RA attending the Katayama Orthopedic Rheumatology Clinic and fulfilling the ACR diagnostic criteria of RA 31 were enrolled in this quasi-randomized study based on problems with authentic medications due to drug resistance, complications and/or risk factors. All patients continuously received the current treatments through this test period. The first twenty patients meeting the selection criteria received a whey protein product containing active antibodies in addition to their current medication (test group). The next 18 background-matched patients visiting the clinic were used as controls without administration of this supplement (control group). Eighteen out of 20 patients in the test group completed the study: 1 man and 17 women, average age 59.7 (range 31-80) years and average disease duration of 114.5 (range 3-360) months. Reasons for inclusion were resistance to concomitant drug therapy (5 patients), drug allergies (4), or complications and/or risk factors (9), such as interstitial pneumonitis (1), bilateral severe knee flexion contracture (1), cancer chemotherapy (2), breastfeeding (1), leucocytopenia (1), chronic heart failure (1), severe osteoporosis (1) and previous pneumonia (1). Mean prednisolone dose in the test group was 2.64 mg/day and mean 28 joint count disease activity score using ESR (DAS28-ESR) was 5.64 at the entry.

The eighteen background-matched control patients consisted of 1 man and 17 women, average age 62.8 (range 44-88) years and average disease duration of 88.1 (range 9-336) months. Mean prednisolone dose in this group was 2.72 mg/day, and mean DAS28-ESR at entry was 5.64. The basic pre-clinical data of patients in these 2 groups are shown in Table 1.

In the test group, one 10 g package of whey protein concentrate containing natural milk antibodies (Bonyuno Chikara^®) with water was added to the concurrent treatment daily after breakfast for 3 months, whereas control patients did not get this food supplement.

In 5 patients (ID 198, 3188, 3709, 3240 and 4119), who responded to this milk antibody treatment as described later, the treatment was repeated after cessation to assess the eventual effect of recall treatment. In addition, 1 patient (ID3240) volunteered for the 3rd time treatment.

A whey protein product, Bonyuno Chikara^®, was supplied by Asama Chemicals Inc. Tokyo, Japan. This product was prepared from normal cow milk instead of colostrums using special conditions to avoid heat denaturation of immunoglobulins, and has been shown to contains relatively high levels of active antibodies against at least 33 strains of pathogenic bacteria 30 . One 10 g packet contains 6 g of whey protein concentrate contains 240 mg natural immunoglobulins, equivalent to 500-600 ml of raw milk, and 3 g of fructooligosaccharide, 0.8 g of milk calcium, and 0.2 g of cellulose.

All patients in the test group were evaluated every month during the 3-month treatment period, whereas patients in the control group were evaluated before and after 3 months using the ACR response criteria with 7 core set variables: acute phase reactants (C-reactive protein: CRP or erythrocyte sedimentation rate: ESR), 0-66 swollen joint count (SJC), 0-68 tender joint count (TJC), modified Health Assessment Questionnaire (mHAQ), patient's and physician's global assessment of disease activity by visual analogue scale (VAS), and patient's pain assessment by VAS 32 , and the European League Against Rheumatism (EULAR) response criteria using DAS28-ESR 33 .

However, the ACR response criteria were not considered to be appropriate for evaluating the indirect disease modifying effects of milk antibodies, which do not possess direct anti-inflammatory or analgesic actions. Therefore, the effect of milk antibody treatment was evaluated by using an ad hoc "Evaluation Point", which was set up for the descriptive purposes of this study. For this analysis, the 7 core set variables used in the ACR response criteria were divided into 8 variables by separating the acute phase reactants into 2 variables, CRP and ESR, to evenly emphasize the importance of these independent marker values. The sum of the variables, which gained more than 20% improvement, was defined as an evaluation point. Evaluation points of 3 out of 8 and over were considered to be effective and defined as responders, whereas evaluation points of 2 and under were considered to be not effective and defined as non-responders. In addition, individual patients were further evaluated for features not related to arthritis but to general health, such as GI status, appetite, weight gain, activities of daily living (ADL), anemia, fatigue, sleep and daily physical feeling, and required dosage of steroid or NSAID.

Serum TNF and IL-6 levels were determined using ELISA kits for human TNF-α/TNSF1A and human IL-6 (R & D Systems, Minneapolis, MN, USA). Serum antibody levels against human, bovine and chicken type II collagen, and LPS from E. coli O26, O55 and O111 (Sigma-Aldrich, St. Louis, MO, USA) were assayed by ELISA as previously reported 4 . Briefly, ELISA plates were coated with antigen (5 μg/ml) dissolved in phosphate buffer (μ = 0.4), pH 7.6. Antigen non-coated wells were used as a blank to determine background values of individual samples. A full strength buffered normal goat serum, pH 8.0, was used for blocking and sample dilution 4 . The secondary antibodies, biotin-conjugated goat anti-human IgG and IgA antibodies (Sigma, USA), were diluted in 2% normal goat serum, whereas streptavidin-peroxidase was diluted in 2% milk casein hydrolysate dissolved in 0.1 M Tris-HCl buffer, pH 7.5. All serum samples were diluted 1:100 and incubated with antigens at room temperature for 2 hours. Colour was developed using tetramethylbenzidine (TMB), and optical density (OD) values of antigen non-coated wells (blank values) of individual samples were subtracted from the OD values in antigen-coated wells.

For the anti-LPS antibody assay, affinity-purified bovine IgG anti-E. coli O111 LPS antibody was used as a reference to compare OD values in the same plate. For detection of bovine anti-LPS antibody, biotinylated goat anti-bovine IgG antibody (Jackson ImmunoResearch, West Grove, PA, USA) was used. Antibody concentrations in human serum samples are expressed as μg/ml.

HLA types were analyzed using peripheral blood cells collected from individual patients by rssop method at HLA Laboratory, Kyoto, Japan.

Data is expressed as the mean ± standard deviation except if otherwise indicated. For statistical evaluation, the paired Student's t-test was used to determine the significance of differences before and after treatment. The Mann-Whitney U-test was used for comparisons between the control and test groups, and the responder and non-responder groups. Chi-square test with Fisher's exact probability test was used for analysis of incidence and prevalence data. Results are shown using P values with 5% level of significance if not otherwise stated. The relationship of HLA haplotypes and therapeutic effect of milk antibodies was analyzed by a prospective cohort analysis using an equation, Relative risk = [a/(a+b)]/[c(c+d)], whereas a: number of patients with DR15+ (non-responder), b: DR15+ (responder), c: DR15- (non-responder), and d: DR15- (responder).

Baseline clinical demographics of the test and control groups were similar (Table 1). Based on the number of milk packages returned at the monthly controls, compliance with taking the whey protein was high, and only 2 of 20 patients withdrew due to disliking of the milk flavour. One patient (ID3188) reported aggravated constipation during the second treatment period, but otherwise no adverse effects were reported.

Apparent improvement of disease marker values such as CRP and SJC was observed as early as 1 month in several patients in the test group and lasted until the end of the study, where such changes were not seen in the control group. After 3-months of treatment, all patients were evaluated by 3 methods, the evaluation point, ACR, and EULAR response criteria (Table 2), in addition to general health assessment by both the patient and physician (see remarks in Table 2).

By the evaluation point analysis, milk antibody treatment was considered to be effective (evaluation point: 3 and more) in 8 of 18 (44.4%) patients but not effective (evaluation point: fewer than 2) in 10 of 18 (55.6%) patients in the test group. On the other hand, no improvement was observed in18 control patients during this 3-months test period, except 1 patient scored 3 (data not shown). Among 10 patients with on evaluation point less than 2 in the test group, 2 patients (ID169 and 511) with mutilating disease and diagnosed as most functionally severe subtype of RA by Ochi's classification 34 were classified as partial responders, because clinical and general health parameters were apparently improved as shown in " Remarks" in Table 2.

Among the 8 patients classified as responders by the evaluation point, 2 patients improved by 20% (ACR20) and 1 patient by 50% (ACR50) according to the ACR response criteria, whereas DAS28-ESR improvement was good in 1 patient, moderate in 4 patients, and poor in 3 patients by EULAR response criteria as shown in Table 2. None of the 2 partial responders, 8 non-responders and 18 control patients showed improvement by either ACR or EULAR evaluation.

General health assessment was performed by both the patient and physician for fatigue, anemia, mobility, ADL, appetite, and GI status such as constipation and diarrhea (see Remarks in Table 2). Notably, constipation or diarrhea was observed in 12 and 3 patients out of 18 patients, respectively, in the control group, and 12 and 2 patients out of 18 patients in the test group (Table 1). In the test group, these symptoms were alleviated in 11 of 14 patients: 7 of 7 responders, 1 of 1 (ID 511) partial responder, and even in 3 of 6 patients (ID 721, 2264 & 9291) in the non-responder group, but in none of the 15 patients in the control group (P < 0.01).

The temporal changes of the 8 core variables during the treatment with milk antibody are shown in Figure 1 to compare the trends between responders and non-responders in the test group and non-treated patients in the control group. In the control group, ESR (p < 0.05) and TJC (p < 0.01) increased due to the persistence of the disease during the 3-month period. In contrast, the mean values of all 8 variables declined or tended to decline in responders of the test group, whereas all these values tended not to change or increase in non-responders. The most remarkable and steady improvement was observed in CRP and SJC, although significant improvement was also observed in the other 4 variables, ESR, patient's pain assessment, and patient's and physician's global health assessments by VAS at the end of the 3-month treatment. Associated with these trends, the mean of DAS28 values in the control group and the non-responder group increased or tended to increase from 5.64 ± 0.86 to 6.29 ± 0.94 (p < 0.05) and 5.45 ± 1.44 to 6.18 ± 1.15 (NS), whereas DAS28 values in the responder group decreased from 5.63 ± 1.26 to 4.66 ± 1.43 (p < 0.01) (Table 2).

Five patients from the responder group (ID198, 3188, 3709, 3240 and 4119) volunteered an additional 3-months of treatment after a 4-month washout period. CRP and ESR levels had decreased and remained at low levels during the first treatment period, but gradually increased and reached the pre-treatment levels approximately 2 months after discontinuation (Figure 2a and 2b). During the second treatment period, CRP and ESR levels decreased again, but increased after termination of the treatment. Similarly, TJC and SJC values tended to improve during the second treatment (the mean of TJC value decreased from 12.4 to 9.6, and the mean of SJC value from 13.2 to 8.8). DAS28-ESR also tended to decrease during the second treatment (Figure 2c), except in 1 patient (ID 3188) whose constipation worsened during the second treatment period.

In order to search for the fundamental difference between the responders and the non-responders, the pre-clinical biological marker values before starting this treatment were compared (Figure 3). There were no significant differences in serum TNF-α and IL-6 levels, which represent the inflammatory reactions. However, anti-human type II collagen antibody levels in the responder group were higher than those in the non-responder group (P < 0.05). Similarly, serum IgG and IgA anti-LPS antibody levels tended to be higher or were higher in the responder group compared to the non-responder group (IgG: P = 0.052, IgA: P < 0.05) as shown in Figure 3.

Notably, only 1 of 7 responders who had been tested for their HLA type was positive for the DR15 (DRB1-1501 & 1502) haplotype, whereas 7 of 8 non-responders carried DR15 (P < 0.05). The risk factor value of DR15, calculated by a prospective cohort analysis, was 6.1 for non-responsiveness (p < 0.01 by Fisher's exact test), whereas the risk factor value of DR4 (DRB1-0404 & 0405) was 0.88 (Table 2).