One-year open-label trial of thalidomide in ankylosing spondylitis




To test in a large open study whether thalidomide is potentially useful in treating ankylosing spondylitis, and to see if thalidomide induces any change in expression of genes in peripheral blood mononuclear cells (PBMC).


Thirty male patients with treatment-refractory ankylosing spondylitis were recruited into a 12-month open study using thalidomide at a dosage of 200 mg/day. Seven indices were measured as primary outcome measures, and 6 other indices as secondary outcome measures. Transcripts in the PBMC of some of these patients were first screened with microarray, and then measured with reverse transcriptase–polymerase chain reaction.


Twenty-six patients completed the study. Of these, 80% showed a >20% improvement in 4 of 7 primary indices. Sharp declines in several parameters were noticed at 3–6 months. Nine patients became pain-free. There was also a statistically significant decrease in tumor necrosis factor α transcripts in the PBMC.


Thalidomide is a reasonably promising drug in treatment-resistant ankylosing spondylitis.


In ankylosing spondylitis (AS), the severity and progression of disease vary widely from patient to patient. Patients with severe disease develop disabling spinal fusion and joint destruction. Their pain is refractory to nonsteroidal antiinflammatory drugs (NSAIDs). Patients with mild disease have no spinal fusion or joint destruction. Their pain is episodic and can be effectively controlled by standard doses of NSAIDs alone. In the past, there were no convincingly effective therapies for patients with severe disease, especially for those with predominant axial involvement. Recently, preliminary short-term reports with anti-tumor necrosis factor α (TNFα) infusion in open studies promise dramatic and rapid relief in such patients (1, 2). These findings suggest that drugs targeting TNFα might be useful for AS. Another drug that has been reported to target TNFα is thalidomide. It selectively suppresses the production of TNFα by normal monocytes in vitro and lowers the levels of plasma TNFα in vivo in patients with erythema nodosum leprosum and pulmonary tuberculosis (3–5).

Thalidomide also has been reported effective in the collagen-induced arthritis animal model (6). The major deterrent to the clinical use of thalidomide is the teratogenic side effect of phocomelia. In spite of this, thalidomide has been tested in several small trials for rheumatoid arthritis (RA). In patients with RA, the drug is reasonably well tolerated; however, the clinical efficacy in RA is not encouraging (7, 8). In contrast to RA, the use of thalidomide in one report of 2 patients with treatment-refractory AS was described as being very successful (9). There was no description in the report on the effect of thalidomide on TNFα in those 2 patients. In the present study, 30 male treatment-refractory AS patients were recruited for an open trial with thalidomide. For the first 5 patients, peripheral blood mononuclear cells (PBMC) were collected before and after 3 months of therapy. Their gene expression profiles were tested by a 588 gene microarray. The purpose of the microarray assays was to serve as a convenient screening procedure for genes that might be suppressed by treatment with thalidomide. An arbitrary threshold was set to identify genes that might be suppressed. The results were then validated more stringently by semiquantitative reverse transcriptase–polymerase chain reaction (RT-PCR) using a larger panel of thalidomide-treated AS patients. Our clinical and experimental results support that thalidomide is potentially effective in the treatment of AS, and that the expression of TNFα is reduced with the therapy.


Patient selection

Thirty male patients were enrolled from the ankylosing spondylitis clinic of PLA General Hospital, a major tertiary referral center in Beijing, China. All 30 satisfied the modified New York classification criteria for AS (10), and had failed such conventional therapies as NSAIDs, sulfasalazine, methotrexate, and cortico-steroids. In addition, patients were required to meet 6 other entry criteria on 2 occasions in the month prior to study entry: Bath AS Disease Activity Index (BASDAI) values >4, total body or spinal pain values >2 when evaluated on a 4-point Likert scale, early morning stiffness >30 minutes, and erythrocyte sedimentation rate (ESR) as well as C-reactive protein (CRP) measurements higher than normal. Written informed consent and an agreement to use contraceptive measures during the period of study were obtained from each patient. The study was approved by the Human Protection Committee of the institute. Safety precaution protocol followed the System for Thalidomide Education and Prescription Safety (Celgene, Warren, NJ).

A detailed assessment was performed at baseline to document medical history. In particular, the following were recorded: disease duration, distribution of arthritis (axial, peripheral, or both), and extraarticular manifestations of disease.

Outcome measures

Of 30 patients enrolled, 4 withdrew, each after the 3-month assessment, all due to patient dissatisfaction with the response. Twenty-six patients received thalidomide for 12 months. After 12 months, thalidomide was stopped. These 26 patients were seen at 0, 1, 3, 6, 9, 12, and 15 months.

At baseline and at each followup visit, in addition to a detailed general clinical examination, the following 7 primary indices were measured: BASDAI, Bath AS Functional Index (BASFI) (11, 12), duration of morning stiffness, total body pain and spinal pain on a 4-point Likert scale, and patient and physician global assessment on a 4-point scale. These 7 parameters constituted components of the primary outcome measures. The primary definition of response to treatment was defined as >20% improvement in 4 of 7 indices after treatment.

The following physical measurements and blood tests were obtained as secondary outcome measures: chest expansion, finger to floor distance, occiput to wall distance, Schöber test, Westergren ESR, and CRP. All clinical evaluations were carried out by one investigator. Urine analyses and blood tests of standard clinical chemistry and hematology panels were also carried out prior to study entry and then at each visit.

Medications during the study

The starting dosage of thalidomide was 50 mg/day to be taken at bedtime (manufactured by Changzhou Pharmaceuticals, China, in 25 mg tablets). The dosage was doubled every 10 days until it reached 200 mg/day. In 1 patient, the dosage was increased to 300 mg/day 3 months after initiation of trial because his response was regarded as being unsatisfactory. Dosages were not decreased after upward adjustments. Any disease-modifying antirheumatic drug (DMARD) the patients were prescribed prior to thalidomide therapy was discontinued. If the patient was taking NSAIDs, they were continued, but at dosages not exceeding the pretrial ones. Four patients were on intermittent corticosteroid, but only at dosages less than 7.5 mg prednisone per day, and for only approximately 1 month prior to the study. These were discontinued 1 week before the baseline assessment point and not reinstituted during the trial. Following the advice of the ethics committee, these patients were allowed to use indomethacin suppository, not exceeding a dosage of 100 mg/day.

Microarray screening and RT-PCR

PBMC were separated from venous samples by histopaque centrifugation and preserved frozen in RNA later (Ambion, Austin, TX). Extraction of RNA and microarray using Atlas Human Array followed protocols provided by the manufacturer (Clontech, Palo Alto, CA). In the preliminary experiment, we assayed 1 sample at 2 different times. When the values derived from the 2 arrays were compared, a correlation coefficient of 0.85 was achieved. Competitive RT-PCR assay of TNFα followed the protocol provided by the manufacturer (Maxim Biolab, San Francisco, CA).

Statistical analysis

The degrees of statistical significance between values were analyzed by the paired Wilcoxon signed rank test. Probabilities of associations were assessed by calculating Spearman's rank correlation coefficient. Comparisons of number of individuals reaching a certain degree of improvement were calculated by chi-square test. P values were regarded as significant if they were <0.05. The effect of Bonferroni correction is discussed below.


Patient demographics

All 30 patients were HLA–B27 positive. The 26 patients who completed the 1-year study had a mean age of 34.67 ± 8.74 years (range 26–64) and a mean disease duration of 5.70 ± 7.08 years (range 2–36). Twenty-one patients had disease predominantly in the spine, and some had disease in the hip(s) as well. This pattern is designated here as being “axial.” In 5 patients, the pattern of disease was both axial and peripheral. Three patients had a history of uveitis. None of the patients had psoriasis, Crohn's disease, or ulcerative colitis. All patients had used NSAIDs in the past. Twenty patients continued their stable doses of NSAIDs throughout the thalidomide study. Eight patients had used corticosteroids at least some time in the past. Four were taking the equivalent of 10–20 mg/day of prednisone just prior to the trial. All patients had previously used sulfasalazine (n = 26), methotrexate (n = 14), or both (n = 24); 21 patients used the 2 drugs sequentially and 3 patients used them simultaneously (Table 1).

Table 1. Demographics of patients completing the study* (n = 26)
  • *

    All patients were male and HLA–B27 positive.

  • Steroids were replaced by 50–100 mg indomethacin suppository 1 week before study entry.

  • NSAIDs = nonsteroidal antiinflammatory drugs.

Age in years, mean (range)34.67 (26–64)
Duration of arthritis in years
 Total, median (range)7.08 (2–36)
 Predominately axial, median (range)5.98 (2–14)
 Axial and peripheral, median (range)15.64 (3–36)
Drug treatment, number of patients past  use (at entry)
 NSAIDs26 (22)
 Corticosteroids8 (4)
 Methotrexate14 (0)
 Sulfasalazine26 (0)

Response to treatment

For this analysis, we defined positive response as having >20% improvement in 4 of 7 primary indices. For the 26 patients who completed the 12 months, 5 patients were nonresponders. If the 4 early dropouts were also regarded as nonresponders, the total would be 30%. Disregarding the early dropouts, 21 (80.8%) showed a positive response (Figure 1). Figure 1 also shows the percentages of patients who demonstrated various levels of responses as measured by the 7 indices. For improvement of 20% in 4 of 7 indices, the numbers were significantly different at 3 months compared with 1 month, and at 6 months compared with 3 months (P < 0.001 for both comparisons). Six months appears to be a crucial time point, because there was statistically significant improvement compared with 3 months for the following categories: 20% improvement in 7 of 7 indices, 50% improvement in 4 of 7 indices, and 75% improvement in 4 of 7 indices (P < 0.05, P < 0.03, and P < 0.04, respectively). However, the percentage of patients showing responses exceeding 75% was small. Also noted is that the response reached a maximum at 12 months.

Figure 1.

Percentage of patients meeting various response criteria at various months after initiation of thalidomide. The response criteria are inserted into the upper right corner of each panel. The numbers of patients assessed at times 0, 1, 3, 6, 12, and 15 months were 30, 30, 30, 26, 26, and 26, respectively.

Values of the 7 individual components of primary outcome measures in the 26 patients are shown in Table 2. Of the 7 indices, each shows a statistically significant decrease after 3 months. Values of total body and spinal pain are good illustrations of the degree and rate of change. There was a sharp decline in total body pain at 6 months and spinal pain at 3 months, with further improvement afterward. At 12 months, the number of patients reporting absence of total body pain or spinal pain were 10 and 9, respectively. Nine patients reported absence of any pain (Figure 2). A sharp decline in BASDAI and BASFI was observed at 3 months and 6 months.

Table 2. Primary outcome measures in the 26 patients completing the study*
MeasureBaseline0 vs. 1 month0 vs. 3 months0 vs. 6 months0 vs. 12 months0 vs. 15 months3 vs. 1 month6 vs. 3 months12 vs. 6 months15 vs. 12 months
  • *

    P derived from the nonparametric Wilcoxon paired signed rank test. They are not adversely affected when the 4 dropout patients are taken into account. Thalidomide was stopped after 12 months.

  • BASDAI = Bath ankylosing spondylitis disease activity index; BASFI = Bath ankylosing spondylitis functional index.

 Mean difference −6.50−16.97−30.34−33.49−25.41−10.47−13.37−3.158.08
 Percent change 10.2326.7147.7652.7240.0018.3628.729.49−26.90
 P <0.001<0.001<0.001<0.001<0.001<0.001<0.0010.100.002
 Mean difference −6.47−16.8−31.54−34.88−29.47−10.33−14.74−3.345.41
 Percent change 9.9525.8548.5253.6645.3417.6530.589.98−17.96
 P <0.001<0.001<0.001<0.001<0.001<0.001<0.0010.0100.001
Total body pain2.47
 Mean difference −0.23−0.60−1.08−1.49−1.51−0.37−0.48−0.41−0.02
 Percent change 9.3124.2943.7260.3261.1316.5225.6729.502.04
 P <0.001<0.001<0.001<0.001<0.0010.0010.0040.0051.0
Total spinal pain2.77
 Mean difference −0.27−0.7−1.25−1.75−1.69−0.43−0.55−0.500.06
 Percent change 9.7525.2745.1363.1861.0117.2026.5732.89−5.88
 P <0.001<0.001<0.001<0.001<0.001<0.0010.0010.0010.63
Patient global assessment2.75
 Mean difference −0.17−0.68−1.27−1.48−1.54−0.51−0.59−0.21−0.06
 Percent change 6.1824.7346.1853.8256.0019.7728.5014.194.72
 P 0.004<0.001<0.001<0.001<0.001<0.001<0.0010.050.65
Physician global assessment2.77
 Mean difference −0.25−0.73−1.23−1.56−1.56−0.48−0.50−0.330.00
 Percent change 9.0326.3544.4056.3256.3219.0524.5121.430.00
 P <0.001<0.001<0.001<0.001<0.001<0.0010.0030.0040.93
Duration of morning stiffness70 min
 Mean difference −8.00−26.00−42.60−47.50−50.19−18.00−16.60−4.90−2.69
 Percent change 11.4337.1460.8667.8671.7029.0337.7317.8811.96
 P 0.10<0.001<0.001<0.001<0.0010.0010.0120.220.31
Figure 2.

Response of total body and spinal pain to thalidomide treatment. Data were derived from the 26 patients who completed the study. Pain was measured on a scale of 0 to 4. Bars in front represent total body pain. Bars at the back represent spinal pain. Thalidomide was stopped after 12 months of treatment.

As for the secondary outcome measures in the 26 patients, there was a significant improvement in Schöber test (P < 0.05) and chest expansion (P < 0.001). The only parameter in which changes were not statistically significant throughout the study was finger to floor distance (P > 0.05) (Table 3). As for inflammatory markers, the mean baseline ESR was 53.07 mm/hour (normal = < 15 mm/hour). The corresponding value for CRP was 3.08 mg/dl (normal was < 0.8 mg/dl). At 12 months, ESR and CRP reached normal levels in 4 and 14 patients, respectively. There was no correlation, however, between the values of ESR and CRP (Table 3).

Table 3. Secondary outcome measures in the 26 patients completing the study*
MeasureBaseline0 vs. 1 month0 vs. 3 months0 vs. 6 months0 vs. 12 months0 vs. 15 months3 vs. 1 month6 vs. 3 months12 vs. 6 months15 vs. 12 months
  • *

    P derived from the nonparametric Wilcoxon paired signed rank test. They are not adversely affected when the 4 dropout patients are taken into account. Thalidomide was stopped after 12 months.

  • ESR = erythrocyte sedimentation rate; CRP = C-reactive protein.

ESR53.07 mm/hour
 Mean difference −3.77−16.4−24.49−28.15−30.75−12.63−8.09−3.66−2.60
 Percent change 7.1030.9046.1553.0457.9425.6222.0612.8110.43
 P 0.210.001<0.001<0.001<0.0010.0020.0040.20.24
CRP3.08 mg/dl
 Mean difference −0.75−1.02−1.74−2.22−2.11−0.27−0.72−0.480.11
 Percent change 24.3533.1256.4972.0868.5111.5934.9535.82−12.79
 P 0.030.02<0.001<0.001<0.0010.08<0.0010.030.26
Schöber2.57 cm
 Mean difference 0.230.720.931.161.110.490.210.23−0.05
 Percent change 8.9528.0236.1945.1443.1920.9411.3514.02−3.55
 P 0.2210.0150.0040.0010.0010.110.550.510.88
Chest expension1.76 cm
 Mean difference 0.160.400.570.660.700.
 Percent change 9.0922.7332.3937.5039.7715.0012.507.563.64
 P 0.090.001<0.001<0.001<0.0010.
Occiput-wall2.33 cm
 Mean difference −0.23−0.99−1.26−1.45−1.45−0.76−0.27−0.190.00
 Percent change 9.8742.4954.0862.2362.2336.1920.1517.760.00
 P 0.360.
Finger-floor7.37 cm
 Mean difference −0.37−0.97−0.83−2.21−2.13−0.600.14−1.380.08
 Percent change 5.0213.1611.2629.9928.908.57−2.1921.10−1.55
 P 0.440.340.370.

Thalidomide was stopped at 12 months. When patients were assessed at 15 months, using >20% change in 4 of 7 indices as definition of primary response, only 1 patient showed a significant deterioration. When the 7 components of the primary outcome measures were analyzed individually, there was statistically significant deterioration in BASDAI and BASFI. However, relapses in total body pain and spinal pain were still negligible. There was no statistically significant deterioration in the secondary outcome measures (Tables 2 and 3).

Adverse events

No patients withdrew from the study because of adverse effects. The major side effects were slight drowsiness in 8 and dry mouth in 6, both of which resolved after 4 weeks of continued therapy. Three patients reported an increase in dandruff. Three had a mild increase of liver enzymes of less than 2-fold above maximum normal values. Values had returned to normal when tested 1 week later. Two patients had microscopic hematuria on 1–2 occasions. All patients were questioned about symptoms of peripheral neuropathy at every visit. One patient had 2–3 seconds per day of tingling sensation at digit tips at the 12th month that lasted only 3 days. Because thalidomide is known to cause peripheral neuropathy, it was immediately stopped (13). Subsequent clinical neurologic examination revealed no peripheral neuropathy. No neuropathy symptoms were observed in any other patient.

Screening by microarray

PBMC were collected from 5 patients before and after 3 months of treatment with thalidomide. Differential gene expression was first screened by a 588 gene microarray analysis. The result of each gene in each patient was recorded as intensity of hybridization signals. For all 5 patients, the intensity value of each gene before and after treatment was compared. Seven thalidomide-suppressed genes were identified, of which the differences in intensities were more than the background value, and in which the P values were <0.05 when tested with paired t-test: TNFα, interleukin-1β, M1P-1α, c-jun, M1P-2α, OX40 ligand, and the T lymphocyte maturation-associated protein MAL. When applying the Bonferroni correction to the P values to take into account the number of genes being assayed simultaneously, the P values became >0.05. In addition, because of the high experimental variability in microarray, and the arbitrary threshold set here, these microarray results can only be regarded as preliminary. Because the TNFα identified in microarray has been reported as being a target for thalidomide, we tested the samples for TNFα using a competitive RT-PCR assay. In addition, we extended the number of patients to 10 before and after 3 months, and 5 patients after 6 months. Differences of values between before treatment and 3 months as well as 6 months were statistically significant (P < 0.04 and 0.03, respectively) (Figure 3).

Figure 3.

Effect of thalidomide on the molar quantity of tumor necrosis factor (TNF) α transcripts in peripheral blood mononuclear cells. The numbers of patients at 0, 3, and 6 months were 10, 10, and 5, respectively. Error bars represent standard error of means. atto = 10−18.


A large number of DMARDs have been tried in the treatment of severe AS; however, most do not show much promise. Thalidomide has been reported to be potentially useful in suppressing disease, even in patients already refractory to other DMARDs. However, so far only the results of 2 AS patients are reported in the literature. A larger number of documented cases in open study are essential before consideration for a double-blind controlled trial. This report extends the number by 30. An additional challenge for this study is that there are as yet no internationally agreed on endpoints to measure response to DMARDs in AS patients. Commonly used parameters are global assessment by patients and physicians, BASDAI, BASFI, ESR, and CRP. Using some of these and additional parameters, 2 recent studies have shown very significant clinical improvement in AS patients who received infusions of anti-TNFα antibody (1, 2). In the current study, we measured a total of 13 parameters that have been used previously in those studies. Following the example of those reports, we also set up a panel of indices as primary outcome measures. In our study, we defined a >20% improvement of 4 of 7 indices as a primary response. Using this definition, 80% of patients who completed our 1-year thalidomide study demonstrated a positive response. In fact, about 50% of patients showed a >50% improvement in 4 of 7 indices. However, complete remission with improvement >75% in 7 of 7 indices was not attained. This observation of primary outcome measure improvement was supported by improvement in 5 parameters of secondary outcome measures. Most notable were significant decreases in both ESR and CRP; ESR and CRP were high in all patients prior to therapy. They became completely normal in 4 and 14 patients, respectively, after 6–12 months of thalidomide therapy. These 2 parameters are the only laboratory parameters shared in the 2 anti-TNFα antibody infusion studies that have been reported (1, 2). They are accepted as useful laboratory parameters to monitor disease activity in AS, both on account of their feasibility and because there is some evidence that they correlate with disease activity (14).

The rate of response of AS patients to thalidomide was certainly not as dramatic as those taking anti-TNFα antibody, where there can be subjective improvement as soon as 1 day after the first infusion (1, 2). As a matter of fact, 4 of our recruited patients withdrew from the study at the 3rd month because the patients were not satisfied with the progress. The patients who completed the 1-year of thalidomide showed maximum response at 6–12 months. Sharp declines in several parameters were observed after 3–6 months of therapy. The response to thalidomide can be highly satisfactory to the patients. Nine patients reported absence of any arthritis pain. Similar to infusion therapy with anti-TNFα antibody, maintenance of response requires continual use of the drug. In our patients, there was a significant clinical relapse 3 months after stopping thalidomide.

In such open studies, it would be preferable to include additional objective parameters to enhance the validity of the clinical observations. Studies of anti-TNFα antibody infusion, for example, include assessment by magnetic resonance imaging or histologic changes. In this study, we first screened the PBMC of 5 patients by microarray for changes in gene expression profiles. Because the microarray technique is susceptible to considerable experimental variation and difficulties in interpretation, the technique was being used here simply as a screening procedure. Using an arbitrary criterion, 7 out of 588 genes were discovered to be potentially affected by thalidomide treatment. Because the most interesting would be a decrease in expression of TNFα, we tested this with a semiquantitative RT-PCR and found statistically significant suppression. This observation does not imply that TNFα is the target of thalidomide therapy, it simply provides a parameter to strengthen the clinical observation that thalidomide leads to changes in the patients.

In the past, the possibility of teratogenic side effects has been the major deterrent to using thalidomide. Because of that, in the US, the manufacturer (Celgene, Warren, NJ) has provided a controlled distribution system that requires compliance before a prescription is effective. The risk of treating with thalidomide should be reasonably controlled, even in other parts of the world, if these guidelines are stringently followed. Here, we have documented that thalidomide is a potentially useful drug in patients with refractory AS and that almost all our patients tolerated thalidomide reasonably well. This drug could have a special place in the treatment of those AS patients who have a high chance of disabling progression, especially when cost of treatment is a limiting factor.