Although alternatives have been proposed, splenectomy remains the best treatment for patients with chronic autoimmune thrombocytopenic purpura (AITP) (Blanchette et al, 1998). However, only 50–90% of patients respond to splenectomy, and 10% relapse after initial success (Karpatkin, 1997). Moreover, even if the operative mortality rate is now probably low, morbidity can be severe, particularly in older patients who, furthermore, do not respond to splenectomy as well as younger subjects (George et al, 1996). Splenectomized patients are also more at risk of life-threatening infections. Thus, it would be useful to have a means of predicting the response to splenectomy. Recently, Law et al (1997) found a close relationship between the platelet increase after high-dose intravenous immunoglobulin (IVIg) and the outcome of splenectomy. We sought to confirm these findings and evaluate the value of some other preoperative criteria in a larger series of adults with chronic AITP.
The response to high-dose intravenous immunoglobulin (IVIg) was recently reported to be predictive of outcome after splenectomy in patients with autoimmune thrombocytopenic purpura (AITP). We analysed the records of 75 adults with chronic AITP who received IVIg and subsequently underwent splenectomy. There was no significant difference in the response rate to splenectomy according to whether or not patients had responded to IVIg (81% v 67%, P = 0.36). Age, the time from diagnosis to splenectomy, and the response to steroids were also not significantly associated with outcome after splenectomy. These results indicate that the response to IVIg or steroids is not predictive of the efficacy of splenectomy.
PATIENTS AND METHODS
All adults (age >17 years) with AITP followed in three centres who received IVIg and subsequently underwent splenectomy were included. AITP was diagnosed according to standard criteria, i.e. isolated thrombocytopenia, normal or increased megakaryocyte count in an otherwise normal bone marrow aspirate, and absence of other causes of thrombocytopenia. Patients with human immunodeficiency virus infection or systemic lupus erythematosus were excluded. Responses to IVIg, steroids and splenectomy were analysed; a positive response was defined as a platelet count increase to > 50 × 109/l and at least twice the initial value. A response to splenectomy was recorded only if no relapses occurred (in such situations an accessory spleen was not found despite systematic search). Different IVIg treatment schedules were administered: one group of patients received a total IVIg dose of 2 g/kg body weight (b.w.) injected over 2–5 d, and another group received 0.5–1 g/kg b.w. infused in 1 d. In this latter group, IVIg was always reinfused if no platelet response was observed on day 3; thus, all non-responders received a total IVIg dose of 2 g/kg b.w. The response to steroids was assessable in 62 patients; the administered dose of steroids ranged from 1 mg/kg b.w. per day prednisone (per os) to 15 mg/kg b.w. methylprednisolone (i.v.). Patients who responded to steroids subsequently underwent splenectomy if they relapsed when steroids were tapered.
Median values were compared using the Mann-Whitney U test, and categorical data using the Chi2 test with Yates' correction for small groups. A P value of < 0.05 was considered significant.
Seventy-five patients (48 women and 27 men) were included. Their characteristics are summarized in 1Table I. The response rates to IVIg, steroids and splenectomy were 76%, 69% and 77%, respectively. Patients were followed-up for a median of 42 months (range 10–183 months). The time from diagnosis to splenectomy ranged from 3 to 156 months and was > 6 months in 65 (87%) patients.
* Responses to steroids were assessable in only 62 patients.
The rate of response to splenectomy was not associated with age, the length of AITP before surgery, or the previous response to IVIg or steroids (Table II). The positive predictive value of the response to IVIg for the outcome of splenectomy was 81%, and the negative predictive value was 33%. Among the 18 patients who did not respond to IVIg, 12 remitted after splenectomy; their median platelet count was 6 × 109/l (range 3–20 × 109/l) before splenectomy and 230 × 109/l (range 156–818 × 109/l) after splenectomy; the median follow-up after splenectomy was 33 months (range 8–58 months; only one patient was followed for less than a year). The diagnosis of AITP was confirmed retrospectively by bone marrow revaluation in the six patients who did not respond to either IVIg or splenectomy. Among these six patients, five had transiently responded to steroids; three remitted spontaneously 8, 47 and 120 months after splenectomy; two remitted on azathioprine and only one had chronic refractory AITP despite dapsone and danazol therapy.
NS: not significant (P > 0.05).* Responses to steroids were assessable in 62 patients.
The positive predictive value of the response to steroids for the outcome of splenectomy was 81% and the negative predictive value was 26%. Taken together, the positive predictive value of the response to both steroids and IVIg for the outcome of splenectomy was 86% and the negative predictive value was 35%. Interestingly, six of the seven patients who did not respond to either steroids or IVIg subsequently responded to splenectomy.
Several teams have attempted to identify prognostic factors for the response to splenectomy in patients with chronic AITP. Some claimed that a favourable outcome after surgery was more likely in younger patients, those with an initial response to steroids, and those with a shorter interval between diagnosis and splenectomy (Den Ottolander et al, 1984; Rocco & Stein, 1984; Naouri et al, 1993). Like Fenaux et al (1989), we failed to confirm these results. The site of autologous 111In-labelled platelet destruction appears to be an important predictor of splenectomy outcome (Najean et al, 1997). However, despite their high predictive value in particular in patients with predominant splenic platelet destruction, isotope studies are expensive and time consuming, and recently published practice guidelines judged them unnecessary (George et al, 1996).
Law et al (1997) recently reported a high positive predictive value (91%) of the response to IVIg for the outcome of splenectomy in a series of 30 patients. The nine patients who responded poorly to IVIg also responded poorly to splenectomy, giving a negative predictive value of 100%. As far as the positive predictive value is concerned, our results are consistent with those of Law et al (1997). In contrast, the negative predictive value, which would be an important decision-making criterion, is very different (33% in our series and 100% in the report by Law et al, 1997). Thus, in a larger cohort of adults with chronic AITP, we found that the response to IVIg was not a reliable predictor of the outcome of splenectomy. Our results are in keeping with those of Schneider et al (1997) and Fabris et al (1997), who recently reported their experience based on two small cohorts.
These conflicting results are difficult to explain, as we chose the same definition of treatment responses as Law et al (1997), and the lengthy follow-up of our patients prevents any underestimation of the negative predictive value, as relapses are rare more than a year after splenectomy (George et al, 1996; Karpatkin, 1997). Possible explanations include (1) the fact that we studied more patients, all of whom were adults, whereas one-third of the 30 patients studied by Law et al (1997) were <15 years old; (2) the fact that at least two of the patients studied by Law et al (1997) did not have AITP (they were subsequently found to have bone marrow hypoplasia or an abnormal karyotype).
In conclusion, we consider that splenectomy is not contraindicated in patients with chronic AITP who do not respond previously to IVIg and/or steroids.
This work was supported by a grant from Direction de la Recherche Clinique/AP-HP.