Long-term follow-up analysis after rituximab therapy in children with refractory symptomatic ITP: identification of factors predictive of a sustained response


Ugo Ramenghi, Haematology Unit, Paediatric Department, University of Torino, Piazza Polonia n. 94 I-10126 Torino, Italy. E-mail: ugo.ramenghi@unito.it


We report the long-term follow-up (median 39·5 months) of 49 paediatric patients (33 females and 16 males) with refractory symptomatic immune thrombocytopenic purpura (ITP) treated with rituximab. The overall response rate was 69% (34/49 patients). Twenty-one responders had a platelet count >50 × 109/l at a median 20·2 months from treatment. Kaplan–Meier analysis showed a probability of relapse-free survival (RFS) of 60% at 36 months from the first rituximab infusion. The number of infusions and a previous splenectomy did not influence overall response rate. Patients who achieved complete response were significantly older at diagnosis and first rituximab infusion than partial responders (P = 0·027). Older children displayed a significantly greater probability of sustained response (RFS) at 36 months than younger children (88·9% vs. 56·7%, P = 0·037). Earlier responses (within 20 d from treatment) were significantly associated with both complete (P = 0·004) and sustained response (P = 0·002). Only mild and transient side-effects were observed in 9/49 children; no major infections nor delayed toxicities were recorded during the follow-up.

Immune thrombocytopenic purpura (ITP) is an autoimmune disorder characterised by low platelet count and mucocutaneous bleeding; during childhood, the disease is usually benign and resolves spontaneously 6–18 months after diagnosis in most cases (Blanchette & Bolton-Maggs, 2008). Few patients fail to achieve an increase in platelet count in response to first-line treatments with intravenous immunoglobulin (IVIG) or steroids. A few children with refractory ITP present with repeated haemorrhages and must be aggressively treated with second-line therapies, such as vincristine, cyclophosphamide, azathioprine and cyclosporin A (Kalpatthi & Bussel, 2008).

Rituximab is a chimeric humanised IgG1/κ monoclonal antibody (Reff et al, 1994) first developed for the treatment of adult B-cell non-Hodgkin lymphoma (Cvetkovic & Perry, 2006). It targets the CD20 antigen on the surface of normal and malignant premature and mature B lymphocytes, and induces B cells destruction by means of both complement-mediated lysis and antibody-dependent cellular cytotoxicity. Induction of apoptosis has also been demonstrated (Flieger et al, 2000; Shan et al, 2000; Alas et al, 2001).

On the assumption that selective B-cell depletion stops the production of autoantibodies, in the past 10 years the use of rituximab has been extended to the treatment of autoimmune diseases (Garcia Hernandez et al, 2007; Garvey, 2008). Its efficacy and tolerability in the treatment of ITP have been shown in adults (Arnold et al, 2007) and children (Taube et al, 2005; Wang et al, 2005; Bennett et al, 2006; Parodi et al, 2006; Penalver et al, 2006; Rao et al, 2008), though its exact mechanism is still unknown. Few data are available regarding long-term follow-up analysis and prognostic factors in children.

This paper reports the long-term follow-up of all paediatric patients with refractory ITP treated with rituximab in the Haematology Units affiliated to the AIEOP (Italian Association of Paediatric Haematology and Oncology) to date.

Material and methods

Patients and treatment

An ad hoc form was transmitted to the Haematology Units affiliated to the AIEOP in order to collect data about all patients with ITP aged 12 months to 18 years treated with rituximab since January 2002. Each institution was free to treat its patients in its own way in accordance with their clinical conditions. Informed written consent was obtained from the parents of all patients. Approval by the ethics committee to participate in the survey was obtained in centres where required. The 49 (33 female and 16 male) patients aged 12 months to 18 years treated to date in 11 AIEOP Centres were enrolled. Nineteen have been previously described (Parodi et al, 2006).

The median age at diagnosis was 7·42 years (range 0·7–17·6 years): eight girls were pubertal at diagnosis (more than 12 years old). The median platelet count at diagnosis was 7 × 109/l (range 3–23 × 109/l). No clinical data suggesting non-immune thrombocytopenia were present.

All patients had been refractory (i.e. have failed to respond or were dependent on repeated administrations to avoid bleeding) to all treatments prior to rituximab. All had previously received one or more courses of IVIG at standard doses (0·8–1 g/kg/d for one or 2 d). Prednisone (PDN) or methylprednisolone (m-PDN) had also been given in 47/49 cases: four had received oral treatment only (PDN 1–2 mg/kg/d), 12 i.v. treatment only (m-PDN 10–20 mg/kg/d) and 31 both oral and i.v. treatment. Twenty-five patients had received high doses of dexamethasone (DXM 24 mg/m2/d for 4 d). Four patients had also received anti-D Immunoglobulins (50 μg/kg/d for 1 d) and three had been treated with cyclosporin A (3–5 mg/kg/d). Five patients had been splenectomized.

All presented sustained clinically significant spontaneous mucocutaneous bleeding requiring multiple hospital admissions and their quality of life was negatively influenced by the side-effects of treatments, multiple physician visits and restriction in life-style with avoidance of activities likely to result in injury. Patient characteristics are set out in Table I.

Table I.   Characteristics of the patients and response.
Patient nSexAge at ITP diagnosis (years)PLTs at ITP diagnosis (×109/l)Previous treatmentsSplenectomy, months from diagnosisRituximab, months from diagnosisAge at rituximab (years)Rituximab Infusions (n)Time to response (d)Response, typeRelapseDuration of response (months)
  1. IVIG, intravenous immunoglobulin; PDN, prednisone; m-PDN, methyl-prednisolone; DXM, dexamethasone; anti-D, anti-D immunoglobulins; CSA, cyclosporin A; CR, complete response (platelet count >150 × 109/l); PR, partial response (platelet count > 50 × 109/l); NR, no response (platelet count <50 × 109/l).

 1F153IVIG, m-PDN, DXM, Ig anti-D61216219CRNo51·8
 3F109IVIG, PDN, m-PDN, DXM101111427CRYes4
 6M6xIVIG, PDN, m-PDN711247CRNo12·9
11F26IVIG, PDN113418CRYes7
14F1426IVIG, m-PDN201642CRNo18·3
15F1112IVIG, m-PDN18191242CRNo74·8
16F1214IVIG, m-PDN201443CRNo24·4
17F812IVIG, m-PDN291152CRYes4
18M79IVIG, PDN, m-PDN, DXM30681342CRNo71·1
19F520IVIG, m-PDN35853CRYes3
20M1019IVIG, m-PDN201163CRYes5
21F411IVIG, m-PDN22642CRNo40·1
22F1310IVIG, PDN, m-PDN, DXM1214415CRNo27·2
23M37IVIG, PDN, m-PDN, Ig anti-D1444120CRYes8
25F35IVIG, PDN, m-PDN36627CRNo12·2
26F76IVIG, m-PDN1827CRNo12·6
32F97IVIG, PDN, m-PDN, DXM2311420CRNo20·2
30M14IVIG, PDN, m-PDN, DXM41415CRNo16·2
35F124IVIG, PDN, m-PDN, DXM2414421CRYes6
36F139IVIG, PDN, m-PDN, DXM91447CRNo16·2
39F99IVIG, PDN, m-PDN, DXM1110430CRNo41·5
42F1817IVIG, PDN118410CRNo10·1
45M138IVIG, PDN, m-PDN91447CRNo4·6
46M133IVIG, PDN, m-PDN11334CRNo3·7
47F108IVIG, PDN, m-PDN, DXM361347CRNo39·5
48F103IVIG, PDN, Ig anti-D161143CRNo8·1
24M64IVIG, PDN, m-PDN32827PRYes5
28M3xIVIG, m-PDN558314PRYes0·9
31M16IVIG, PDN, m-PDN, DXM364430PRYes1
34M43IVIG, PDN, m-PDN, DXM125490PRYes3
37F66IVIG, PDN, m-PDN, DXM217440PRYes1
40M87IVIG, PDN, m-PDN, DXM109415PRNo41·5
41F92IVIG, PDN, m-PDN, DXM2611490PRNo40·6
43F86IVIG, PDN, m-PDN, DXM119430PRYes1
 2F223IVIG, m-PDN, DXM, CSA2142NR
 4F56IVIG, PDN, m-PDN, DXM343074NR
 5F43IVIG, PDN, m-PDN, DXM, CSA4173104NR
 7F10xIVIG, PDN, m-PDN89184NR
 8F92IVIG, PDN, DXM3026113NR
 9F79IVIG, PDN, m-PDN, DXM252293NR
10M62IVIG, PDN, m-PDN574NR
12F14xIVIG, Ig anti-D16164NR
27M1xIVIG, PDN, m-PDN175154NR
29F53IVIG, PDN, m-PDN, DXM2774NR
33M23IVIG, PDN, m-PDN, DXM7084NR
38F72IVIG, PDN, m-PDN, DXM36104NR
44M1320IVIG, PDN, m-PDN5145NR
49F223IVIG, m-PDN, DXM, CSA334362NR

The median time between ITP diagnosis and rituximab treatment was 21 months (range 1–175). Eleven children were treated at a very early stage (<12 months from diagnosis) when refractory to first-line treatments and seriously symptomatic.

Median age at treatment was 10·7 years (range 1·2–17·7). Fourteen patients were pubertal at rituximab administration (11 females >12 years, three males >14 years).

Rituximab (Roche, Milan, Italy) was administered i.v. weekly at 375 mg/m² in all patients. Most patients (35/49, 72%) received four weekly infusions, according to the treatment schedule used for oncology patients. Six (12%) received two infusions, four (8%) three infusions, three (6%) five infusions and one (2%) six infusions. All patients received premedication with acetaminophen, chlorphenamine maleate and/or hydrocortisone. No other treatment was administered in addition to rituximab during the follow-up of the responders.


Overall response (OR) was defined as an increase in platelet count above 50 × 109/l for at least 7 d. A complete response (CR) was recorded if the platelet count rose to normal values (i.e. above 150 × 109/l), a partial response (PR) if it rose to between 50 × 109/l and 150 × 109/l. Relapse was defined as a decrease of the platelet count to below 50 × 109/l. Duration of response was considered from the day of the initial infusion to the first time of relapse or to time of analysis.

Statistical analysis

Data were analysed as at 30 June 2008. Results were expressed as medians and ranges, or absolute numbers and percentages, as appropriate. The association between categorical variables were assessed with Fisher’s exact test, while differences between groups for continuous variables were checked with the Mann–Whitney test.

Kaplan–Meier analyses were performed in order to estimate the relapse-free survival (RFS); the equality of survival times between groups identified by categorical variables was assessed with Log Rank Test.

A receiver-operating characteristic (ROC) curve was used to find the best cut-off for the length between treatment and response as a predictive value of a sustained response. All statistical tests are two-sided with P < 0·05 as the significance cut-off. Data were analysed with the Statistical Package for the Social Sciences (spss) v.15 for Windows (SPSS Inc., Chicago, IL, USA).



The median follow-up from the first infusion was 39·5 months (range 3·7–88·2).


An OR was detected in 34/49 (69%) of patients: 8/49 (16%) achieved a PR, 26/49 (53%) achieved a CR. The other 15 patients (31%) did not present a platelet count above 50 × 109/l and were classed as non-responders (Table I). OR was detected in 3/5 splenectomized patients, and 31/44 non-splenectomized patients.


Thirteen of 34 (38%) responders relapsed at a median 4 months (range 1–8) from the first infusion. The other 21 (62%) continued to maintain a platelet count above 50 × 109/l at a median 20·2 months (range 3·7–74·8) from treatment (Table I).

Kaplan–Meier analysis showed a probability of RFS of 60% at 36 months from the first infusion. All relapses occurred within 8 months of treatment (Fig 1).

Figure 1.

 Overall response. Relapse-free survival (RFS) after rituximab therapy. X-axis shows the number of months from the first rituximab infusion; Y-axis shows the proportion of RFS.

Patterns of response

Responders reached the cut-off platelet count >50 × 109/l at a median 8·5 d (range 2–120) from first rituximab infusion; 24/34 (70%) displayed the response within 20 d, six (18%) between 20 and 40 d, one (3%) between 40 and 60, three (9%) between 80 and 120 d.

Factors predictive of an OR

Comparison between patients who achieved a platelet count >50 × 109/l and patients who failed to respond are summarised in Table II. The OR rate was not influenced by sex, age and platelet count at diagnosis and a previous splenectomy or the number of weekly administered infusions (P > 0·05). Median duration of the disease was significantly shorter in responders than in non-responders (19·5 vs. 30 months; P = 0·01).

Table II.   Factors predictive of overall response.
FactorNon respondersRespondersP-value
  1. Data are shown as median (range), unless otherwise indicated.

  2. *Fisher’s exact test.

  3. †Mann–Whitney’s test.

Sex, n (%)
 M4 (26·7)12 (35·3)0·743*
 F11 (73·3)22 (64·7)
 Age at diagnosis (years)4·9 (0·7–14·2)8·6 (0·7–17·6)0·061†
 Platelet count at diagnosis (×109/l)3 (2–23)7 (2–26)0·265†
Splenectomy, n (%)
 Yes2 (13·3)3 (8·8)0·635*
 No13 (86·7)31 (91·2)
Age at treatment (years)9·6 (4·0–17·6)11·0 (1·2–17·7)0·703†
Time from diagnosis (months)30·0 (5·0–175·0)19·5 (1·0–71·0)0·011†
No. infusions, n (%)
 <44 (26·7)6 (17·6)0·470*
 ≥411 (73·3)28 (82·4)

Factors predictive of CR

Partial response and CR patients are compared in Table III. Patients who achieved CR were significantly older at ITP diagnosis than partial responders: median age 9·7 vs. 5·6 years (P = 0·027).

Table III.   Factors predictive of complete response.
FactorPartial respondersComplete respondersP-value
  1. Data are shown as median (range), unless otherwise indicated.

  2. *Fisher’s exact test.

  3. †Mann–Whitney’s test.

Sex, n (%)
 M5 (62·5)7 (27)0·098*
 F3 (37·5)19 (73)
Age at diagnosis (years)5·6 (0·7–9·3)9·7 (0·8–17·6)0·027†
Platelet count at diagnosis (×1093/l) 6 (2–7)9 (3–26)0·015†
Splenectomy, n (%)
 yes0 (0)3 (11·5)1·000*
 no8 (100)23 (88·5)
Age at treatment (years)8·2 (3·7–17·4)11·5 (1·2–17·7)0·027†
Time from diagnosis (months)23·5 (10·0–55·0)17·5 (1·0–71·0)0·331†
No. infusions, n (%)
 <42 (25·0)4 (15·4)0·894*
 ≥46 (75·0)22 (84·6)

Median age at first infusion was 11·5 years (range 1·2–17·7) in complete responders and 8·1 years (range 3·7–17·4) in partial responders. The difference was statistically significant (P = 0·027). Sex, number of infusions and a previous splenectomy did not influence the probability of complete response (P > 0·05).

The duration of the disease at time of rituximab was not significantly different between complete and partial responders (17·5 vs. 23·5 months, P > 0·05). Patients who achieved CR reached the platelet count cut-off (50 × 109/l) at a median 7 d (range 2–120), patients who achieved PR reached this cut-off point at a median 30 d (range 7–90) (P = 0·004).

Factors predictive of relapse

Kaplan–Meier analysis showed that males older than 14 years and females older than 12 years at time of rituximab administration had a significantly higher probability of RFS: 88·9% vs. 56·7% RFS at 36 months (P = 0·037).

A higher probability of RFS was observed in patients who reached the platelet count cut-off level within 20 d from treatment than patients who responded later (P = 0·002). RFS at 36 months was 73·7% and 22·2% respectively (Fig 2).

Figure 2.

 Stratification by time to response. Relapse-free survival (RFS) after rituximab therapy in patients who displayed response within 20 d from the first rituximab infusion (continuous line) and in patients who responded later (dotted line). X-axis shows the number of months from the first rituximab infusion; Y-axis shows the proportion of RFS.

Patients with a CR presented a higher probability of RFS than those with a PR (P = 0·003). RFS at 36 months was 70·2% and 25·0% respectively (Fig 3).

Figure 3.

 Stratification by type of response. Relapse-free survival (RFS) after rituximab therapy in patients who achieved complete response, i.e. platelet count >150 × 109/l (continuous line), and in patients who achieved partial response, i.e. platelet count >50 × 109/l (dotted line). X-axis shows the number of months from the first rituximab infusion; Y-axis shows the proportion of RFS.

The other parameters, including number of infusions administered (Table IV), did not influence the probability of RFS.

Table IV.   Factors predictive of relapse-free survival (RFS).
FactorProbability of RFS at 36 months (%)P-value
  1. *Probability of RFS at 16 months.

Age at diagnosis (years)
Age at treatment (years)
Time from diagnosis (months)
No. infusions
Time to response (d)
Type of response


Only mild and transient side-effects, which promptly resolved with appropriate therapy, were recorded in 9/49 children: four urticarial rash, two mild headache, one headache and chills, two fever and chills. No major infections or delayed toxicities attributable to rituximab were experienced during the follow-up period.


The clinical data for 49 paediatric patients with refractory ITP who underwent rituximab treatment are reported. The efficacy and tolerability of this drug in both adult and childhood ITP have been already proven, but few data are available regarding long-term follow-up and prognostic factors in children. To our knowledge, this is the largest paediatric series and with the longest follow-up reported to date.

Overall response was observed in about 2/3 of patients. More than half of patients achieved a CR with the platelet count returning to normal levels.

The efficacy of treatment was consistent with data previously reported by Wang et al (2005) and Taube et al (2005), although both of these studies defined response as an increase in the platelet count to 30 × 109/l. However, in the series reported by Bennett et al (2006), a platelet count >50 × 109/l was reached by only 31%.

Our OR could be an expression of the natural course of the disease since some patients were treated in its very early stages when refractory to first-line treatments and seriously symptomatic. However, the absence of a correlation between duration of ITP prior to rituximab and sustained response (i.e. the presence of relapse in responding patients with a short disease duration) does not support the hypothesis of a spontaneous remission.

Twenty-one of the 34 (62%) of responders continued to maintain a normal platelet count at a median 20·2 months from rituximab administration. For some of them, the remission is the longest in the literature.

During the long follow-up, no delayed toxicities were registered. Only mild and transient side effects were observed in 18% of patients.

Most patients received four weekly infusions, according to the treatment schedule developed for oncology patients. However, the number of the infusions administered (i.e. the total dose of drug) did not influence the response rate. Our data match those previously reported in children treated with a single infusion (Taube et al, 2005), and adults treated with a single, fixed 100 mg dose (Provan et al, 2007).

These data confirm that further studies are warranted in order to determine the treatment schedule with the best cost:effectiveness ratio.

Previous studies have reported no significant difference in the response rate (Stasi et al, 2001; Cooper et al, 2004; Braendstrup et al, 2005; Bennett et al, 2006; Penalver et al, 2006; Garcia-Chavez et al, 2007) or in the relapse rate, between splenectomized and non-splenectomized patients. These data confirm that rituximab is also a worthwhile therapeutic option in non-splenectomized patients.

None of our responding patients required a further rituximab infusion and all displayed B-cell reconstitution (data not shown). If these data are confirmed in more patients and with a longer follow-up, this therapeutic option should be considered prior to splenectomy in order to defer or even avoid it (Godeau et al, 2006).

In spite of all the limitations because of the number of patients enrolled and the retrospective analysis of data, some factors predictive of complete response, i.e. older age and older age at time of rituximab administration, have been identified. Furthermore, patients that were pubertal at treatment presented a higher probability of sustained response than younger children.

The short time to achieving the cut-off platelet count of 50 × 109/l and the magnitude of the response were significantly associated with remission duration (sustained remission with return to normal platelet count).

As previously reported (Wang et al, 2005; Parodi et al, 2006; Penalver et al, 2006), two patterns of response were observed, confirming that rituximab acts through two or perhaps more mechanisms. If later responses may be related to the blockage of autoantibodies production secondary to the B-cell compartment depletion, other mechanisms may be implicated in earlier responses observed in the first days after the infusion.

An immunomodulant mechanism, i.e. a competitive blockade of FCγ receptors in the reticular endothelial system (the immune-complex decoy hypothesis) has been postulated (Taylor & Lindorfer, 2007). However, in our cohort, earlier responses were significantly associated with both a complete and a sustained response; these long-term remissions cannot be simply explained by this hypothesis. The exact mechanism of action of rituximab remains to be elucidated.

In conclusion, our data provide further evidence of the efficacy and safety of rituximab administration in refractory symptomatic paediatric ITP. Despite the small size of this study, the results suggest that rituximab is more effective in older paediatric patients, and that earlier responses are associated with a higher probability of a long-lasting response.


This work was partially supported by Banca del Piemonte and Regione Piemonte grants awarded to UR. We thank Dr Giuseppe Loffredo for helpful discussion.