TNF Inhibitors vs Standard Therapy in Rheumatoid Arthritis
The TACIT (tumour necrosis factor inhibitors against combination intensive therapy) trial was an open label pragmatic randomised two arm non-inferiority trial carried out over 12 months in multiple centres.
Patients were recruited from 24 rheumatology clinics in England. We included men and women aged over 18 with disease durations over 12 months who met the 1987 criteria for classification of rheumatoid arthritis and NICE criteria for starting biologics in England. The NICE criteria comprise disease activity score for 28 joints >5.1 twice over one month apart after treatment with methotrexate and one other disease modifying drug. We excluded patients who unable or unwilling to give informed consent, had not had successful results with or had contraindications to all combinations of disease modifying drugs (including possible pregnancy), had contraindications to tumour necrosis factor inhibitors, had serious inter-current illness, or were taking high dose corticosteroids (>10 mg prednisolone).
The trial compared two treatment strategies in patients who completed local screening procedures for treatment with tumour necrosis factor inhibitors, including tuberculosis screening. Safety monitoring followed national guidance. Treatment was guided by monthly changes in disease activity scores for 28 joints. The Box outlines the tumour necrosis factor inhibitor strategy, which replicated NICE guidance when the trial started, and the combined disease modifying drug strategy. Treatments used as required at standard doses included non-opiate analgesics, non-steroidal anti-inflammatory drugs, folic acid (5 mg/week) with methotrexate, bone protection (such as alendronate and calcium/vitamin D) with glucocorticoids, and intra-articular steroids.
Primary Outcome. The primary outcome was the score on the health assessment questionnaire at 12 months. Scores were measured initially at baseline and at six and 12 months. Trials of disease modifying drugs and biologics have shown that this patient assessed outcome is sensitive to change. Its performance is equal to disease activity measures like joint counts. Changes in health assessment questionnaire scores influenced NICE's decisions on biologics. It involves questions across eight domains: dressing and grooming, getting up, eating, walking, hygiene, reach, grip, and chores or activities. There are four possible answers: 0=without any difficulty; 1=with some difficulty; 2=with much difficulty; and 3=unable to do. The sum of the highest score per domain is divided by eight. The total score ranges from 0 to 3 (0=best; 3=worst) in 0.125 increments.
Secondary Outcomes. Every six months we assessed quality of life, erosive damage, and economics. Assessments comprised the EuroQol 5-dimension scale (EQ5D-3L), medical outcomes study short form 36 (SF-36), and radiographs of the hands (including wrists) and feet. An experienced observer (DLS) who was blinded to treatment read the digitised radiographs using modified Larsen scores after the trial ended. Use of resources was recorded with a modified client service receipt inventory administered as self completed questionnaires retrospectively for the previous three month period. Costs (for 2010–11 in £) were assessed from a health and social care perspective and were multiplied by two to represent the six month period before each assessment.
We assessed disease activity monthly by recording tender and swollen joint counts (28 joints), erythrocyte sedimentation rates, patients' global assessments of disease activity (100 mm visual analogue scale), and the disease activity scores for 28 joints. Withdrawals from treatment or the trial were also recorded monthly together with details of drugs and adverse events.
An anonymised electronic data capture system collected clinical data except radiographs (www. medscinet.net).
Sample Size Calculation. Published data on 12 month changes in scores on the health assessment questionnaire in rheumatoid arthritis trials of tumour necrosis factor inhibitors (mean baseline score 1.7; reduction after treatment of 25%; standard deviation of change 0.4) and combination therapy with disease modifying drugs (mean baseline score 1.6; reduction after treatment of 31%; standard deviation of change 0.6) gave an average standard deviation for changes in health assessment questionnaire scores of 0.5.
The minimal clinically important change in scores on the health assessment questionnaire is 0.22. We based our sample size on testing the null hypothesis of a difference of >0.22 between the two treatments. With a (one sided) testing level of 5%, we needed a sample size of 176 to achieve 90% power. We recruited 214 patients to allow for patients not receiving treatment or dropping out when treatment had started.
Potentially eligible patients were screened and reasons for non-entry recorded. Consenting patients were randomised in blocks of four with allocation stratified by region. MedSciNet generated the allocation sequence; trial staff had no prior knowledge of the allocation sequence. There were variable delays between randomisation and baseline assessments because patients received combination disease modifying drugs and tumour necrosis factor inhibitors through routine NHS systems.
The trial was unblinded because it used treatment specific algorithms with the adjustment of multiple drug doses. Patients assessed their own disability and quality of life. Radiographs were read blinded.
Baseline characteristics were summarised by randomisation group as means and standard deviations (continuous normally distributed variables), medians and interquartile ranges (non-normally distributed variables), and frequencies and percentages (categorical variables).
Randomised patients who received treatment were assessed on an intention to treat basis. All participants had complete observations at baseline. Missing data at follow-up was imputed regardless of the reason(s) they was missing. For participants with missing outcomes, we used the baseline outcomes and other explanatory covariates (treatment group, sex, age, ethnicity, region, and disease duration) to impute the missing data, assuming unobserved measurements were missing at random (see Appendix Table A ).
Linear regression evaluated 12 month outcomes. Univariate analyses were adjusted for region (design variable). Multivariable analyses were adjusted for sex, ethnicity, age, region, and duration of disease and baseline scores, which all influence disability. Generalised estimating equations with working correlations (auto-regressive with lag one) evaluated disease activity score for 28 joints and its components measured monthly. Treatment withdrawals and toxicities were compared with Fisher's exact tests. These analyses compared treatment strategies ignoring subsequent treatment switches. In all analyses patients in the tumour necrosis factor inhibitor strategy were the reference group. Only the primary outcome was tested for non-inferiority. Other outcomes were compared for evidence of superiority of one or other treatment strategy.
Exploratory analyses examined good responders according to the European League Against Rheumatism (EULAR) with odds ratios and the development of remission based on disease activity scores for 28 joints with log rank tests on all observed data. No imputations were performed for these data. These analyses also examined differences between patients in the disease modifying drug strategy who stayed taking these drugs or switched to tumour necrosis factor inhibitors. Finally complete case analyses evaluated patients who followed the protocol and received 12 months' treatment.
We also undertook a full economic evaluation, which will be reported separately.
Methods
Design
The TACIT (tumour necrosis factor inhibitors against combination intensive therapy) trial was an open label pragmatic randomised two arm non-inferiority trial carried out over 12 months in multiple centres.
Participants
Patients were recruited from 24 rheumatology clinics in England. We included men and women aged over 18 with disease durations over 12 months who met the 1987 criteria for classification of rheumatoid arthritis and NICE criteria for starting biologics in England. The NICE criteria comprise disease activity score for 28 joints >5.1 twice over one month apart after treatment with methotrexate and one other disease modifying drug. We excluded patients who unable or unwilling to give informed consent, had not had successful results with or had contraindications to all combinations of disease modifying drugs (including possible pregnancy), had contraindications to tumour necrosis factor inhibitors, had serious inter-current illness, or were taking high dose corticosteroids (>10 mg prednisolone).
Interventions
The trial compared two treatment strategies in patients who completed local screening procedures for treatment with tumour necrosis factor inhibitors, including tuberculosis screening. Safety monitoring followed national guidance. Treatment was guided by monthly changes in disease activity scores for 28 joints. The Box outlines the tumour necrosis factor inhibitor strategy, which replicated NICE guidance when the trial started, and the combined disease modifying drug strategy. Treatments used as required at standard doses included non-opiate analgesics, non-steroidal anti-inflammatory drugs, folic acid (5 mg/week) with methotrexate, bone protection (such as alendronate and calcium/vitamin D) with glucocorticoids, and intra-articular steroids.
Primary Outcome. The primary outcome was the score on the health assessment questionnaire at 12 months. Scores were measured initially at baseline and at six and 12 months. Trials of disease modifying drugs and biologics have shown that this patient assessed outcome is sensitive to change. Its performance is equal to disease activity measures like joint counts. Changes in health assessment questionnaire scores influenced NICE's decisions on biologics. It involves questions across eight domains: dressing and grooming, getting up, eating, walking, hygiene, reach, grip, and chores or activities. There are four possible answers: 0=without any difficulty; 1=with some difficulty; 2=with much difficulty; and 3=unable to do. The sum of the highest score per domain is divided by eight. The total score ranges from 0 to 3 (0=best; 3=worst) in 0.125 increments.
Secondary Outcomes. Every six months we assessed quality of life, erosive damage, and economics. Assessments comprised the EuroQol 5-dimension scale (EQ5D-3L), medical outcomes study short form 36 (SF-36), and radiographs of the hands (including wrists) and feet. An experienced observer (DLS) who was blinded to treatment read the digitised radiographs using modified Larsen scores after the trial ended. Use of resources was recorded with a modified client service receipt inventory administered as self completed questionnaires retrospectively for the previous three month period. Costs (for 2010–11 in £) were assessed from a health and social care perspective and were multiplied by two to represent the six month period before each assessment.
We assessed disease activity monthly by recording tender and swollen joint counts (28 joints), erythrocyte sedimentation rates, patients' global assessments of disease activity (100 mm visual analogue scale), and the disease activity scores for 28 joints. Withdrawals from treatment or the trial were also recorded monthly together with details of drugs and adverse events.
An anonymised electronic data capture system collected clinical data except radiographs (www. medscinet.net).
Sample Size Calculation. Published data on 12 month changes in scores on the health assessment questionnaire in rheumatoid arthritis trials of tumour necrosis factor inhibitors (mean baseline score 1.7; reduction after treatment of 25%; standard deviation of change 0.4) and combination therapy with disease modifying drugs (mean baseline score 1.6; reduction after treatment of 31%; standard deviation of change 0.6) gave an average standard deviation for changes in health assessment questionnaire scores of 0.5.
The minimal clinically important change in scores on the health assessment questionnaire is 0.22. We based our sample size on testing the null hypothesis of a difference of >0.22 between the two treatments. With a (one sided) testing level of 5%, we needed a sample size of 176 to achieve 90% power. We recruited 214 patients to allow for patients not receiving treatment or dropping out when treatment had started.
Randomisation
Potentially eligible patients were screened and reasons for non-entry recorded. Consenting patients were randomised in blocks of four with allocation stratified by region. MedSciNet generated the allocation sequence; trial staff had no prior knowledge of the allocation sequence. There were variable delays between randomisation and baseline assessments because patients received combination disease modifying drugs and tumour necrosis factor inhibitors through routine NHS systems.
Blinding
The trial was unblinded because it used treatment specific algorithms with the adjustment of multiple drug doses. Patients assessed their own disability and quality of life. Radiographs were read blinded.
Statistical Methods
Baseline characteristics were summarised by randomisation group as means and standard deviations (continuous normally distributed variables), medians and interquartile ranges (non-normally distributed variables), and frequencies and percentages (categorical variables).
Randomised patients who received treatment were assessed on an intention to treat basis. All participants had complete observations at baseline. Missing data at follow-up was imputed regardless of the reason(s) they was missing. For participants with missing outcomes, we used the baseline outcomes and other explanatory covariates (treatment group, sex, age, ethnicity, region, and disease duration) to impute the missing data, assuming unobserved measurements were missing at random (see Appendix Table A ).
Linear regression evaluated 12 month outcomes. Univariate analyses were adjusted for region (design variable). Multivariable analyses were adjusted for sex, ethnicity, age, region, and duration of disease and baseline scores, which all influence disability. Generalised estimating equations with working correlations (auto-regressive with lag one) evaluated disease activity score for 28 joints and its components measured monthly. Treatment withdrawals and toxicities were compared with Fisher's exact tests. These analyses compared treatment strategies ignoring subsequent treatment switches. In all analyses patients in the tumour necrosis factor inhibitor strategy were the reference group. Only the primary outcome was tested for non-inferiority. Other outcomes were compared for evidence of superiority of one or other treatment strategy.
Exploratory analyses examined good responders according to the European League Against Rheumatism (EULAR) with odds ratios and the development of remission based on disease activity scores for 28 joints with log rank tests on all observed data. No imputations were performed for these data. These analyses also examined differences between patients in the disease modifying drug strategy who stayed taking these drugs or switched to tumour necrosis factor inhibitors. Finally complete case analyses evaluated patients who followed the protocol and received 12 months' treatment.
We also undertook a full economic evaluation, which will be reported separately.
SHARE
