All herbal components of DJT were prepared in a large computer-controlled boiler where volatile oils of therapeutic values were collected and sprayed into a two-story high vacuum drying chamber. The active ingredients of the herbs were transformed into granulated compounds five times more concentrated than the raw herbs. These granulated compounds were then vacuum dried at low temperatures before being siphoned into a separate sterile-room where they were bottled, labeled and sealed. For quality assurance of the active ingredients in DJT, HPLC (high performance liquid chromatography) fingerprinting was employed to identify substances in the final product 12. An extraction process required by the CCMP regulations was performed, involving the use of water to eliminate AA-I from Asarum heteropoides roots, a nephrotoxic substance contained mostly in the leaves and/or aerial parts of plants 13. Liquid chromatography – tandem mass spectrometry (LC/MS/MS) 14, a more sensitive method, had been carried out by the Bureau of Food and Drug Analysis of the Department of Health prior current study to ensure that AA-I was at an undetectable level. Each batch of DJT was also tested for E. coli, Salmonella (bacteria count) and heavy metals.

This prospective follow-up study was coordinated from the National Taiwan University, while participants were enrolled through two research clinics in northern Taiwan: Chinese Medicine Branch and Yangming Branch of the Taipei City Hospital. Approval of the study was obtained from the Joint Institutional Review Board for Traditional Chinese Medicine of Taiwan (JIRBTCM94-0426-01). All participants provided signed and informed consents before taking part in the study.

All nurses involved in the study attended a standardized training session to ensure consistency and to meet GCP (Good Clinical Practices) requirements in the study protocol. Participant eligibility was assessed during the first two visits to the clinic. Following the initial screening visit, participants entered a run-in phase to determine baseline data on their symptoms and their quality of life; each participant also underwent a health assessment including complete blood counts and biochemical function tests. These data were collected to ensure the eligibility criteria for each participant, and to screen out respondents with potential poor compliance. Following the run-in phase, each participant was provided with sufficient DJT to begin treatment at a dose of 2.5 g twice daily.

Subsequent study visits were scheduled for the first, second and fourth weeks, with any symptoms of the knee and adverse events being assessed at each visit. At the end of the trial, the participants received a further physical examination, including blood tests. Participants were contacted by telephone one to two days prior to each visit in order to encourage their continued compliance. Throughout the study period, leftover package counts were undertaken so as to monitor each participant's compliance. The consumption of non-steroidal anti-inflammatory drugs (NSAIDs) rescue medication was also recorded at each visit.

The primary outcome parameters evaluated prior to and during the four-week intervention were pain, stiffness and physical functioning subscales, along with the global pain assessment using a 100-mm 'visual analog scale' (VAS) from the Western Ontario and McMaster Universities Osteoarthritis (WOMAC) index 15. We purchased the WOMAC Chinese language form directly from the author 16. Changes in the quality of life were also assessed using the World Health Organization Quality of Life questionnaire – Taiwan brief version (WHOQOL-BREF) 17. Each subject's physical constitution was also categorized using a questionnaire for pattern identification, or 'bian zheng' in Chinese medical terminology, based on the criteria recommended in Zhongyao Xinyao Linchuang Yanjiu Zhidao Yuanze (The Guidelines for Clinical Research on New Chinese Medicines) 18. The questionnaire consisted of categories as follows: frequency of pain, aversion to cold and flaccidity of the lower back and knees (each of which were assessed in terms of 'occasional or less', 'often', or 'so frequent as to interfere with work'), and duration of stiffness ('less than one hour' or 'more than one hour').

The WOMAC index is a multi-dimensional, disease specific, self-administered, health status measure. It probes clinically important, patient-relevant symptoms in the areas of pain, stiffness and physical function in patients with knee osteoarthritis. The instrument comprises 24 questions (five on pain, two on stiffness and 17 on physical functioning) and can be completed in less than five minutes. Weightings were introduced to add together the five items on pain as a total sum score which predominantly quantifies the severity of knee pain in different daily activities, with the same process being carried out on stiffness and physical functioning 19. The VAS version of the WOMAC is valid, reliable and sufficiently sensitive for the detection of clinically-important changes in health status following a variety of interventions 2021. The Taiwan version of the WHOQOL-BREF comprises four domains (i.e. physical, psychological, social and environmental) containing 24 facets, and two national items on overall quality of life (QOL) and general health 17, with higher scores indicating superior QOL.

Routine hematology and biochemistry data were collected at the baseline and at week 4, including complete blood count and platelets, serum levels of creatinine, BUN, AST, ALT, albumin/globulin (A/G), uric acid and urine levels of N-acetyl-β-D-glucosaminidase and retinol binding protein. A research nurse also actively monitored any adverse events and recorded any unexpected signs, symptoms or feelings during the study period.

Our analysis focused mainly on changes in the three-domain construct of the WOMAC assessment (pain, stiffness and physical function) along with the WHOQOL measures. The treatment effects were described as changes of the mean differences among the participants between the scores measured during each visit at weeks 1, 2, and 4 and at baseline. Separate linear regression models were constructed for the response variables of the differentiated scores on the pain, stiffness and physical functioning subscales, the VAS-pain and the WHOQOL-BREF domains.

The explanatory variables in the final regression model were gender, age, baseline severity of osteoarthritis, the interaction between age and gender, the BMI (in kg/m²), aversion to cold, flaccidity of the lower back and knees and usage of NSAID rescue drugs. For us to examine the improvement across the 3 visits, two indicator variables for visits at weeks 2 and 4 were also included in the regression model. The coefficients of the two indicator variables represent the differences between the average value of measured scores for the visits at weeks 2 and 4 versus those at week 1. The error terms were assumed to be correlated among the three repeated measurements on each participant. Finally, each single WOMAC score item was also examined separately so as to identify any WOMAC items sensitive to treatment by DJT. The estimates of coefficients and standard errors of the estimates in the multiple regression models with correlated error terms were obtained by using generalized least squares function of the package nlme version 3.1–60 in a free statistical software R version 2.1.1 22.

Of the 87 sample patients, 18 were deemed ineligible. The principal reasons for ineligibility were uninterested in participation (n = 9), abnormal liver function (n = 5), abnormal renal function (n = 2) and no evidence of osteoarthritis (n = 2). Sixty-eight of the initial 69 participants completed the four-week study. The main reason for the withdrawal was lack of efficacy during the first week of treatment. The demographic and clinical characteristics of the study subjects are summarized in Table 2. Among them, 37% were male, 82% married and 59% having attained senior high school education or above. Seven of the 68 took less than 80% of the prescribed dosage over the study period for the reason of lack of response to DJT. Patients with poor compliance had significantly lower scores in severity of pain and physical functioning and had significantly higher scores in physical QOL than those with good compliance.

After two weeks of treatment, there were significant and persistent improvements in each of the WOMAC subscales and also in the 10 cm VAS score for knee pain (Table 3). The magnitudes of the improvements in the WOMAC subscales were 27.5% for pain, 34.2% for stiffness and 19% for physical functioning. There was also a 28.2% improvement in the VAS score.

After four weeks of treatment, 44 of the 68 patients reported no change in the symptom of flaccidity; nine reported improvements, whereas fifteen reported deterioration. Thirty-eight patients reported no change in the symptom of aversion to cold after four weeks of treatment; fourteen reported improvements, while16 reported deterioration.

There were no significant improvements in these two symptoms after the treatment.

The results of the multiple linear regression analyses indicate the effects of the different determinants on the outcome scores of the VAS for pain and different WOMAC subscales (Table 4). After rescue medication and other determinants had been controlled, the model constructions demonstrated significant improvements in the severity of pain, stiffness and physical functioning after week 2 (which also persisted through week 4), with substantial treatment effects on the stratum of the most severe pain (baseline VAS score >4 cm, n = 32). The frequency of either flaccidity or aversion to cold were not significantly affected in any of the above major outcome scores, and there were no consistent improvements with regard to these two symptoms after week 4.

There were no significant changes in all tested biomarkers including the highly sensitive indicators of nephrotoxicity 2324 after the 4-week DJT intervention. Only four adverse drug reactions were potentially related to DJT treatment, involving single events of skin discoloration, flashes, diarrhea and tachycardia.