- Email: [email protected]
Predicting response to splenectomy in children with immune thrombocytopenic purpura
Journal of Pediatric Surgery (2010) 45, 140–144
www.elsevier.com/locate/jpedsurg
Predicting response to splenectomy in children with immune thrombocytopenic purpura James H. Wood a , David A. Partrick a,c,⁎, Taru Hays b,d , Moritz M. Ziegler a,c a
Department of Surgery, University of Colorado Denver School of Medicine, Aurora, CO 80045, USA Department of Pediatrics, University of Colorado Denver School of Medicine, Aurora, CO 80045, USA c Department of Pediatric Surgery, The Children's Hospital, Aurora, CO 80045, USA d Department of Pediatrics Section of Hematology, University of Colorado Denver School of Medicine and The Children's Hospital, Aurora, CO 80045, USA b
Received 2 October 2009; accepted 6 October 2009
Key words: Immune thrombocytopenia purpura; Splenectomy; Steroid; Predictors
Abstract Purpose: Predicting the response to splenectomy in children with immune thrombocytopenic purpura (ITP) continues to be a clinical challenge. The purpose of this study is to identify preoperative predictors of outcome for splenectomy in children with ITP. Methods: The charts of 19 children who underwent splenectomy for ITP were retrospectively reviewed. Platelet responses to treatment are categorized as complete response (CR, ≥150,000/μL), partial response (PR, ≥50,000 but b150,000/μL), or nonresponse (NR, b50,000/μL). Results: After splenectomy, 13 patients (68%) had CR, 3 (16%) had PR, and 3 (16%) had NR. No correlation existed between CR to splenectomy and any of the following: age, platelet count at diagnosis, last platelet count before splenectomy, platelet count on postoperative day 1, or responses to preoperative intravenous immunoglobulin, WinRho, or Rituximab. However, all 7 patients who had NR to a full course of steroids subsequently had CR to splenectomy. Nonresponse to steroid therapy was directly correlated with CR to splenectomy (P = .01, Fisher's Exact test). Furthermore, postsplenectomy platelet counts were inversely related to peak platelet response to steroids (correlation coefficient = −0.68, P = .01). Conclusions: Preoperative responsiveness to steroid therapy, as measured by peak platelet count, predicts NR to splenectomy for ITP in children, whereas NR to steroid therapy is highly correlated with CR to splenectomy. These findings challenge the widely held notion that steroid responsiveness portends a favorable outcome after splenectomy. © 2010 Elsevier Inc. All rights reserved.
Presented at the 40th Annual Meeting of the American Pediatric Surgical Association, Fajardo, Puerto Rico, May 28-June 1, 2009. ⁎ Corresponding author. Department of Pediatric Surgery, The Children's Hospital, Aurora, CO 80045, USA. Tel.: + 1 720 777 6571; fax: +1 720 777 7271. E-mail address: [email protected] (D.A. Partrick). 0022-3468/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.jpedsurg.2009.10.026
Most children with immune thrombocytopenic purpura (ITP) experience spontaneous resolution of their disease without medical intervention. However, a subset of approximately 20% develops chronic ITP [1], and optimal management of these children has not been clearly defined. The paucity of adequate guidelines for treatment of chronic
Splenectomy in children with immune thrombocytopenic purpura ITP has resulted in diverse treatment strategies, which are more often based on opinion and “conventional wisdom” than on published data [2]. This difficulty is especially evident when one considers the diversity of opinions about the utility of and indications for splenectomy for chronic ITP. In a recent survey, pediatric hematologists were presented with a case simulation in which the patient met the criteria for splenectomy for chronic, refractory ITP. Only 33% of respondents indicated that they would refer the patient for splenectomy. Respondents were asked to explain their reasoning for avoiding splenectomy, and 2 of the top 3 justifications given were the “possible lack of response to splenectomy” and “inability to predict response to splenectomy” findings, which highlight the need for more potent predictors of response to splenectomy in children with chronic ITP. We have, therefore, undertaken a review of splenectomies for ITP at our institution in an attempt to identify preoperative factors that could guide pediatricians and surgeons in determining which children are most likely to have a favorable response to splenectomy.
141
record. Initial follow-up platelet counts were only counted as stable if the response category persisted at the latest available time-point. At our institution, platelet counts are checked at the first follow-up visit, which is generally scheduled 1 month after splenectomy. In some cases, patients who live far from Denver were seen at a primary care provider's office near their home and 30-day platelet counts are unavailable. The average (SD) number of days between splenectomy and postsplenectomy platelet count was 43 (32) days (range, 17140 days), with a trend toward increased intervals for PR and NR patients (P = .02). Final follow-up time was 362 ± 693 days (range, 18-2364 days) with no difference between groups (P = .50). Fisher's Exact test, analysis of variance, and correlation tests were used to calculate significance. All data are presented as mean ± SD, with range in parentheses, unless stated otherwise in the text. The units for all platelet counts are 103/μL.
2. Results 1. Methods
The study included 5 boys and 14 girls, with an average (SD) age of 12 (3) years (range, 7-17 years), and the average (SD) time from diagnosis to splenectomy was 32 (32) months (range, 3.5-120 months), with no differences between groups (Table 1). Indications for splenectomy included bleeding symptoms refractory to medical therapy (16) and medication avoidance (3). Eighteen of 19 patients underwent planned laparoscopic splenectomy, with 2 conversions to open for small vessel bleeding. Seven patients were transfused with platelets and 3 with packed red blood cells either in the operating room or postoperatively. The average (SD) length of stay was 3.1 (1.5) days (range, 2-8 days), with no readmissions in the first postoperative month and no mortalities. Postsplenectomy, 13 patients had CR (platelet count, 399 ± 173), 3 had PR (64 ± 19), and 3 had NR (20 ± 20). The average platelet count was 8 ± 9 (range, 1-32) at diagnosis, 21 ± 19 (172) at the time of admission for splenectomy, and 133 ± 104 (4359) on postoperative day 1, with no differences between groups. These data are summarized in Table 2. In the CR group, there were 2 relapses. One relapse occurred 15 months after splenectomy with a platelet nadir of 23 but resolved spontaneously with sustained platelets
With approval from the Colorado Multiple Institutional Review Board (protocol #08-0889), we undertook a retrospective review of all electronic medical records available for children who underwent splenectomy at The Children's Hospital between January 2000 and July 2008. We identified 117 records, of which 22 patients had undergone splenectomy for ITP. Of these 22 children, 3 were excluded because of the lack of sufficient data for comparison of postsplenectomy outcomes. In accordance with previously published literature on the subject, we defined responses to each therapeutic modality as follows: platelet count of 150,000/μL or more is considered a complete response (CR), platelet count of 50,000/μL or more but less than 150,000/μL constituted a partial response (PR), and platelet count less than 50,000/μL was defined as no response (NR) [3]. Responses to medical therapy were categorized according to the peak platelet count with treatment. Splenectomy responses were defined by the platelet count on postoperative day 1, at initial follow-up (at least 2 weeks after splenectomy), and at the latest follow-up time-point available in the medical Table 1
Splenectomy complete response (CR), partial response (PR), and no response (NR) groups
Age (y) Sex, n (%) Time to OR (mo)
CR
PR
NR
P
13 ± 3 (7-17) 4 M (44%), 9 F 36 ± 40 (3.5-120)
14 ± 3 (11-17) 0 M (0%), 3 F 21 ± 13 (10-18)
10 ± 4 (7-14) 1 M (50%), 2 F 33 ± 17 (15-48)
.19 .77 .81
Data are mean ± SD (range), unless otherwise specified. M indicates male; F, female; OR, operating room.
142 counts more than 200 ten months later. The other patient relapsed 4 months after splenectomy and required a single course of steroids and rituximab for complete remission. All patients were treated medically before splenectomy. With rare exceptions, medical interventions were pursued sequentially instead of in combination with other medications. Eleven patients received anti–D immunoglobulin (WinRho), 8 were treated with intravenous immunoglobulin (IVIG), and 5 received rituximab before splenectomy. Responses to medical therapy were defined according to the peak platelet count with treatment and plotted in 3 × 3 correlation table with responses to splenectomy. WinRho responses included 1 CR, 1 PR, and 4 NR. There were 1 CR, 3 PR, and 4 NR to IVIG. Also, there was only 1 CR and 4 NR to rituximab. There was no detectable correlation between response to splenectomy and response to WinRho, IVIG, or rituximab (P = .99). Twelve patients were treated with a full course of steroids before splenectomy. Four additional patients were excluded for various reasons, 2 were given steroids only in the immediate preoperative period, 1 was on maintenance steroids before the ITP diagnosis, and 1 relapsed 4 months after splenectomy and achieved durable remission with a short course of medical therapy. Of the 12 included patients, 8 had CR to splenectomy, 1 had PR, and 3 had NR. There was 1 patient in each group who received high-dose steroids. Otherwise, the prednisone dose was 36 ± 5.5 (range, 30-40) mg for 16 ± 7 days with no differences between groups (P = .49). To compare steroid responses between groups, we defined steroid CR, PR, and NR according to the peak platelet response to steroids. Using this definition, there were 4 CR, 1 PR, and 7 NR to steroid treatment. Of the 4 who had CR to steroids, 2 had NR to splenectomy, 1 had PR, and 1 had CR, with an average postsplenectomy platelet count of 72 ± 86 (range, 6-189). On the other hand, all 7 patients with NR to steroids had CR to splenectomy, with an average postsplenectomy count of 453 ± 200 (range, 274-837), P = .01. These findings are summarized in Table 3. Furthermore, when steroid responses were divided between those patients whose counts rose to at least 100 versus those whose counts remained below 100, the relationship between steroid and splenectomy response remains valid (P = .01, Fisher's Exact test exact test). When splenectomy and steroid responses are plotted against one another, there is an inverse relationship between peak platelet count with steroids and the postsplenectomy platelet count, depicted in Fig. 1 (correlation coefficient = −0.68, P = .01).
3. Discussion The first reported splenectomy for ITP was performed in Germany in 1916 [4], and splenectomy continues to be an important treatment for patients with uncontrolled chronic
J.H. Wood et al.
Fig. 1 Postsplenectomy platelet count as a function of peak platelet counts after steroid treatment. Solid black line represents exponential regression. Correlation coefficient = −0.68, P = .01.
ITP or with life-threatening hemorrhage complicating acute ITP [5]. However, despite more than 90 years of experience with splenectomy for ITP, there are still no clearly defined predictors of which patients will potentially benefit from the operation. The gravity of this ambiguity is highlighted by a survey of pediatric hematologists published in 2008 [2]. In this study, physicians were presented with a case scenario of chronic ITP in a child designed to meet qualifications for splenectomy according to the consensus statement of the American Society of Hematologists [6]. Remarkably, only 33% of surveyed pediatric hematologists indicated that they would recommend splenectomy for the child described in the vignette. The top 3 justifications for splenectomy avoidance were (1) long-term risk of sepsis, (2) possible lack of response to splenectomy, and (3) inability to predict response to splenectomy. In our patient population, 68% of patients undergoing splenectomy for ITP had a CR, defined as platelet counts of 150 or greater at initial follow-up with no evidence of recurrence during the full follow-up period. Response rates more than 90% have been reported in some series [7,8]. Although long-term remission rates of at least 70% to 80% are probably more realistic [9], there remains significant skepticism among referring pediatricians about the utility of splenectomy in the treatment of ITP, as demonstrated by 2 recent reviews [10,11] that underscore the perceived inability to predict successful splenectomy. Although several retrospective studies have attempted to identify preoperative predictors of successful splenectomy, few have focused exclusively on ITP in children, and the results of these studies are conflicting For example, a study published by the Intercontinental Childhood ITP study group suggested that older age, male sex, and increased duration of disease before splenectomy may portend improved outcomes in children, although the
Splenectomy in children with immune thrombocytopenic purpura Table 2
143
Platelet counts of splenectomy CR, PR, and NR groups at specific time-points
Diagnosis Presplenectomy POD 1 Postsplenectomy
CR (n = 13)
PR (n = 3)
NR (n = 3)
P
5.6 ± 3 (3-11) 24 ± 22 156 ± 104 (6-359) 399 ± 173 (189-837)
17.7 ± 15 (2-32) 23 ± 13 35 ± 42.5 (4-84) 64 ± 19 (50-86)
2.5 ± 2 (1-4) 7.3 ± 7 126 ± 109 (12-229) 20 ± 20 (6-43)
.08 .43 .19 .00075
POD 1 indicates postoperative day 1. Platelet counts are 103/μL. Data are mean ± SD (range).
authors acknowledge that these are not strong predictors [12]. In contrast, our findings, which are consistent with other pediatric studies [13,15], suggest that there are no differences in age, sex, or time to splenectomy between the CR, PR, or NR to splenectomy groups (Table 2). Likewise, we found that patients in each splenectomy response group had similar platelet counts at diagnosis, upon admission for splenectomy, and on postoperative day 1. These results corroborate the findings of the Intercontinental Childhood ITP study group [12]. The prognostic value of responsiveness to IVIG is an area of ongoing investigation, and the available data seem to support a correlation of favorable responses to IVIG and splenectomy [13-15], although poor response to IVIG may not correlate with failure to respond to splenectomy [15]. Our data reveal no relationship between splenectomy CR and responsiveness to IVIG, WinRho, or rituximab, although a type 2 error is likely, given the relatively small number of patients with these treatments in our study population. The remarkable finding of our study is a highly correlated, inversely proportional relationship between response to steroid treatment and response to splenectomy (Table 3 and Fig. 1). These findings challenge not only our own institutional assumptions but also the published results of some of the few exclusively pediatric studies available [13,16]. For example, Davis et al [13] described a direct correlation between steroid and splenectomy responsiveness. Patients in this study were treated with steroids in the immediate preoperative period, and their median platelet count at admission for splenectomy was 127. Our study, on the other hand, includes only those patients who received therapeutic corticosteroids and excludes patients who received steroids only in preparation for splenectomy. The difference between these 2 studies is highlighted by the fact that our mean platelet count at admission for splenectomy was 21. Table 3
Best response to steroids versus response to splenectomy
Response to splenectomy
Response to steroids CR
PR
NR
CR PR NR
1 1 2
0 0 1
7 0 0
P = .01, Fisher's Exact test.
In contrast, several other studies have failed to find any relationship between steroid responsiveness and surgical outcome [12,14,15,17-19]. Holt et al [14] reviewed 24 patients treated with prednisone for ITP and found no significant association between response to prednisone and response to splenectomy. However, neither the steroid regimen nor the method of categorizing steroid responses is explicitly described in the study. Hemmila et al [15] also found no correlation of steroid response to splenectomy response in 17 patients. However, 94% of patients in this series had either a “good” or “excellent” response to corticosteroids, whereas only 42% of patients in our study had either CR or PR to steroids. The reason for this discrepancy is unclear, but likely betrays a difference in patient populations. Furthermore, with only one steroid failure in the study by Hemmila et al, meaningful conclusions about the predictive value of steroid responsiveness are unlikely. Contributing to the difficulty of interpreting these data is a lack of standardized reporting of platelet response to steroid therapy. For example, platelet responses to steroids are described variously as mean platelet count after steroids [13] or platelet count within 1 month of steroids [15]. We defined responsiveness to medications as peak platelet counts during treatment, which made data interpretation simpler and more readily applicable to the clinical setting. Furthermore, we selected a subset of patients with easily comparable and complete data to evaluate for responses to steroid therapy and splenectomy. Using this approach, we found that a platelet count of 150 or more at any time during steroid treatment was strongly associated with NR or PR to splenectomy and that those patients who had NR to steroids universally had CR to splenectomy (Table 3). The explanation for this phenomenon is unclear, because steroid therapy and splenectomy theoretically have analogous functions: inhibiting antibody production and impairing phagocytosis by splenic macrophages. Given the complex and nonstandardized treatment regimens of ITP patients, it is possible that splenectomy responses were affected by medical intervention not accounted for in our data. Another possibility is that the patients who did not respond to splenectomy have qualitatively or quantitatively different antibodies than those who have CR to splenectomy. Furthermore, in the adult population, there is growing evidence that some patients with ITP have impaired thrombopoiesis leading to decreased platelet production
144 [20]. However, a cohesive explanation is lacking, and there is a need for translational investigation to uncover the mechanistic etiology for divergent responses to steroid treatment and splenectomy. One theoretical shortcoming of this study is that it cannot adequately define the period during which the “peak platelet count” will consistently be encountered, and studies to define this trial period would be very beneficial. This study further highlights the need for prospective analysis of large patient cohorts to fully describe the significance of steroid responsiveness as a predictor of response to splenectomy. However, despite these limitations, the data clearly undermine the long-held belief that steroid responsiveness portends a favorable outcome from splenectomy and suggests that it may be steroid-resistant children who are most likely to have CR to splenectomy.
References [1] Di Paola JA, Buchanan GR. Immune thrombocytopenia. Pediatr Clin North Am 2002;49(5):911-28. [2] Neunert CE, Bright BC, Buchanan GR. Severe refractory immune thrombocytopenic purpura during childhood: a survey of physician management. Pediatr Blood Cancer 2008;5:1513-6. [3] Law C, Marcaccio M, Tam P, et al. High-dose intravenous immune globulin and the response to splenectomy in patients with idiopathic thrombocytopenic purpura. N Engl J Med 1997;336:1494-8. [4] Kaznelson P. Vershwinder der hemmorhagischen diathese bei einem falle von “essentieller thombopenie” (Frank) nack Milzexstripation: splenogene thrombolytische purpura. Wien Klin Wochenschr 1916;29:1451-4. [5] Kliegman RM, Behrman RE, Jenson HB, et al. Nelson Textbook of Pediatrics. Philadelphia (Pa): Saunders; 2007. [6] George JN, Woolf SH, Raskob GE, et al. Idiopathic thrombocytopenic purpura: a practice guideline developed by explicit methods for the American Society of Hematology. Blood 1996;88:3-40. [7] Gadenstatter M, Lamprecht B, Klinger A, et al. Splenectomy versus medical treatment for idiopathic thrombocytopenic purpura. Am J Surg 2002;184:606-9. [8] Watts RG. Idiopathic thrombocytopenic purpura: a 10-year natural history study at the Children's Hospital of Alabama. Clin Pediatr 2004;43:691-702. [9] Mantadakis E, Buchanan GR. Elective splenectomy in children with idiopathic thrombocytopenic purpura. J Pediatr Hematol Oncol 2000;22:148-53. [10] Shad AT, Gonzalez CE, Sandler SG. Treatment of immune thrombocytopenic purpura in children: current concepts. Pediatr Drugs 2005;7:325-36. [11] Psaila B, Bussel JB. Immune thrombocytopenic purpura. Hematol Oncol Clin North Am 2007;21:743-59. [12] Kuhne T, Blanchette V, Buchanan GR, et al. Splenectomy in children with idiopathic thrombocytopenic purpura: a prospective study of 134 children from the Intercontinental Childhood ITP Study Group. Pediatr Blood Cancer 2007;49:829-34. [13] Davis PW, Williams DA, Shamberger RC. Immune thrombocytopenia: surgical therapy and predictors of response. J Pediatr Surg 1991;26:407-13. [14] Holt D, Brown J, Terrill K, et al. Response to intravenous immunoglobulin predicts splenectomy response in children with immune thrombocytopenic purpura. Pediatrics 2003;111:87-90. [15] Hemmila MR, Foley DS, Castle VP, et al. The response to splenectomy in pediatric patients with idiopathic thrombocytopenic
J.H. Wood et al.
[16]
[17] [18] [19]
[20]
purpura who fail high-dose intravenous immune globulin. J Pediatr Surg 2000;35:967-72. Weinblatt ME, Ortega JA. Steroid responsiveness: a predictor of the outcome of splenectomy in children with chronic immune thrombocytopenic purpura. Am J Dis Child 1982;136:1064-6. Wilde RC, Ellis KD, Cooper WM. Splenectomy for chronic idiopathic thrombocytopenic purpura. Arch Surg 1967;95:344-50. Block GE, Evans R, Zajtchuk R. Splenectomy for idiopathic thrombocytopenic purpura. Ann Surg 1966;92:484-9. Akwari OE, Itani KMF, Coleman RE, et al. Splenectomy for primary and recurrent immune thrombocytopenic purpura (ITP). Ann Surg 1987;206:529-41. Stasi RR, Evangelista ML, Stipa E, et al. Idiopathic thrombocytopenic purpura: current concepts in pathophysiology and management. Thromb Haemost 2008;99:4-13.
Discussion Unidentified Speaker: What is your hypothesis for the mechanism of this observation? James H. Wood, MD (Denver, CO): We had several hypotheses, none of which we considered adequate. Dr Hays is one of our coauthors. She is one of the hematologists and we discussed that issue at length. There could be different antibodies in these 2 groups that are interacting. There are data from the adult literature that would suggest that the subset of patients with ITP have as their primary defect a problem with platelet production instead of a problem with destruction. It may just be that we are talking about 2 different types of ITP, although again none of those give a full explanation. John Gosche, MD (Las Vegas, NV): In order to challenge the understanding that we have that responders to steroids are also responders—you should have more than just 12 patients. What are your plans to improve your numbers to see if what you have seen at your institution is comparable to what has been seen at other institutions? Wood, MD: The difficulty with doing a prospective trial, and one of the shortcomings of this study, is that these patients have been preselected because they have failed therapy. If they had not failed their therapy, they never would have had a splenectomy. When we are talking about complete responses to steroids or any other treatment, we are talking about transient responses at best. So it would be very difficult to have a rigorous trial to look at those differences, and we do not have any specific plans yet to study this prospectively, but that is certainly something that needs to be done to more adequately define not only this prognostic value but also the waiting period during which one should expect to see the peak platelet count so they can decide that the patient needs either splenectomy or continued medical therapy.
www.elsevier.com/locate/jpedsurg
Predicting response to splenectomy in children with immune thrombocytopenic purpura James H. Wood a , David A. Partrick a,c,⁎, Taru Hays b,d , Moritz M. Ziegler a,c a
Department of Surgery, University of Colorado Denver School of Medicine, Aurora, CO 80045, USA Department of Pediatrics, University of Colorado Denver School of Medicine, Aurora, CO 80045, USA c Department of Pediatric Surgery, The Children's Hospital, Aurora, CO 80045, USA d Department of Pediatrics Section of Hematology, University of Colorado Denver School of Medicine and The Children's Hospital, Aurora, CO 80045, USA b
Received 2 October 2009; accepted 6 October 2009
Key words: Immune thrombocytopenia purpura; Splenectomy; Steroid; Predictors
Abstract Purpose: Predicting the response to splenectomy in children with immune thrombocytopenic purpura (ITP) continues to be a clinical challenge. The purpose of this study is to identify preoperative predictors of outcome for splenectomy in children with ITP. Methods: The charts of 19 children who underwent splenectomy for ITP were retrospectively reviewed. Platelet responses to treatment are categorized as complete response (CR, ≥150,000/μL), partial response (PR, ≥50,000 but b150,000/μL), or nonresponse (NR, b50,000/μL). Results: After splenectomy, 13 patients (68%) had CR, 3 (16%) had PR, and 3 (16%) had NR. No correlation existed between CR to splenectomy and any of the following: age, platelet count at diagnosis, last platelet count before splenectomy, platelet count on postoperative day 1, or responses to preoperative intravenous immunoglobulin, WinRho, or Rituximab. However, all 7 patients who had NR to a full course of steroids subsequently had CR to splenectomy. Nonresponse to steroid therapy was directly correlated with CR to splenectomy (P = .01, Fisher's Exact test). Furthermore, postsplenectomy platelet counts were inversely related to peak platelet response to steroids (correlation coefficient = −0.68, P = .01). Conclusions: Preoperative responsiveness to steroid therapy, as measured by peak platelet count, predicts NR to splenectomy for ITP in children, whereas NR to steroid therapy is highly correlated with CR to splenectomy. These findings challenge the widely held notion that steroid responsiveness portends a favorable outcome after splenectomy. © 2010 Elsevier Inc. All rights reserved.
Presented at the 40th Annual Meeting of the American Pediatric Surgical Association, Fajardo, Puerto Rico, May 28-June 1, 2009. ⁎ Corresponding author. Department of Pediatric Surgery, The Children's Hospital, Aurora, CO 80045, USA. Tel.: + 1 720 777 6571; fax: +1 720 777 7271. E-mail address: [email protected] (D.A. Partrick). 0022-3468/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.jpedsurg.2009.10.026
Most children with immune thrombocytopenic purpura (ITP) experience spontaneous resolution of their disease without medical intervention. However, a subset of approximately 20% develops chronic ITP [1], and optimal management of these children has not been clearly defined. The paucity of adequate guidelines for treatment of chronic
Splenectomy in children with immune thrombocytopenic purpura ITP has resulted in diverse treatment strategies, which are more often based on opinion and “conventional wisdom” than on published data [2]. This difficulty is especially evident when one considers the diversity of opinions about the utility of and indications for splenectomy for chronic ITP. In a recent survey, pediatric hematologists were presented with a case simulation in which the patient met the criteria for splenectomy for chronic, refractory ITP. Only 33% of respondents indicated that they would refer the patient for splenectomy. Respondents were asked to explain their reasoning for avoiding splenectomy, and 2 of the top 3 justifications given were the “possible lack of response to splenectomy” and “inability to predict response to splenectomy” findings, which highlight the need for more potent predictors of response to splenectomy in children with chronic ITP. We have, therefore, undertaken a review of splenectomies for ITP at our institution in an attempt to identify preoperative factors that could guide pediatricians and surgeons in determining which children are most likely to have a favorable response to splenectomy.
141
record. Initial follow-up platelet counts were only counted as stable if the response category persisted at the latest available time-point. At our institution, platelet counts are checked at the first follow-up visit, which is generally scheduled 1 month after splenectomy. In some cases, patients who live far from Denver were seen at a primary care provider's office near their home and 30-day platelet counts are unavailable. The average (SD) number of days between splenectomy and postsplenectomy platelet count was 43 (32) days (range, 17140 days), with a trend toward increased intervals for PR and NR patients (P = .02). Final follow-up time was 362 ± 693 days (range, 18-2364 days) with no difference between groups (P = .50). Fisher's Exact test, analysis of variance, and correlation tests were used to calculate significance. All data are presented as mean ± SD, with range in parentheses, unless stated otherwise in the text. The units for all platelet counts are 103/μL.
2. Results 1. Methods
The study included 5 boys and 14 girls, with an average (SD) age of 12 (3) years (range, 7-17 years), and the average (SD) time from diagnosis to splenectomy was 32 (32) months (range, 3.5-120 months), with no differences between groups (Table 1). Indications for splenectomy included bleeding symptoms refractory to medical therapy (16) and medication avoidance (3). Eighteen of 19 patients underwent planned laparoscopic splenectomy, with 2 conversions to open for small vessel bleeding. Seven patients were transfused with platelets and 3 with packed red blood cells either in the operating room or postoperatively. The average (SD) length of stay was 3.1 (1.5) days (range, 2-8 days), with no readmissions in the first postoperative month and no mortalities. Postsplenectomy, 13 patients had CR (platelet count, 399 ± 173), 3 had PR (64 ± 19), and 3 had NR (20 ± 20). The average platelet count was 8 ± 9 (range, 1-32) at diagnosis, 21 ± 19 (172) at the time of admission for splenectomy, and 133 ± 104 (4359) on postoperative day 1, with no differences between groups. These data are summarized in Table 2. In the CR group, there were 2 relapses. One relapse occurred 15 months after splenectomy with a platelet nadir of 23 but resolved spontaneously with sustained platelets
With approval from the Colorado Multiple Institutional Review Board (protocol #08-0889), we undertook a retrospective review of all electronic medical records available for children who underwent splenectomy at The Children's Hospital between January 2000 and July 2008. We identified 117 records, of which 22 patients had undergone splenectomy for ITP. Of these 22 children, 3 were excluded because of the lack of sufficient data for comparison of postsplenectomy outcomes. In accordance with previously published literature on the subject, we defined responses to each therapeutic modality as follows: platelet count of 150,000/μL or more is considered a complete response (CR), platelet count of 50,000/μL or more but less than 150,000/μL constituted a partial response (PR), and platelet count less than 50,000/μL was defined as no response (NR) [3]. Responses to medical therapy were categorized according to the peak platelet count with treatment. Splenectomy responses were defined by the platelet count on postoperative day 1, at initial follow-up (at least 2 weeks after splenectomy), and at the latest follow-up time-point available in the medical Table 1
Splenectomy complete response (CR), partial response (PR), and no response (NR) groups
Age (y) Sex, n (%) Time to OR (mo)
CR
PR
NR
P
13 ± 3 (7-17) 4 M (44%), 9 F 36 ± 40 (3.5-120)
14 ± 3 (11-17) 0 M (0%), 3 F 21 ± 13 (10-18)
10 ± 4 (7-14) 1 M (50%), 2 F 33 ± 17 (15-48)
.19 .77 .81
Data are mean ± SD (range), unless otherwise specified. M indicates male; F, female; OR, operating room.
142 counts more than 200 ten months later. The other patient relapsed 4 months after splenectomy and required a single course of steroids and rituximab for complete remission. All patients were treated medically before splenectomy. With rare exceptions, medical interventions were pursued sequentially instead of in combination with other medications. Eleven patients received anti–D immunoglobulin (WinRho), 8 were treated with intravenous immunoglobulin (IVIG), and 5 received rituximab before splenectomy. Responses to medical therapy were defined according to the peak platelet count with treatment and plotted in 3 × 3 correlation table with responses to splenectomy. WinRho responses included 1 CR, 1 PR, and 4 NR. There were 1 CR, 3 PR, and 4 NR to IVIG. Also, there was only 1 CR and 4 NR to rituximab. There was no detectable correlation between response to splenectomy and response to WinRho, IVIG, or rituximab (P = .99). Twelve patients were treated with a full course of steroids before splenectomy. Four additional patients were excluded for various reasons, 2 were given steroids only in the immediate preoperative period, 1 was on maintenance steroids before the ITP diagnosis, and 1 relapsed 4 months after splenectomy and achieved durable remission with a short course of medical therapy. Of the 12 included patients, 8 had CR to splenectomy, 1 had PR, and 3 had NR. There was 1 patient in each group who received high-dose steroids. Otherwise, the prednisone dose was 36 ± 5.5 (range, 30-40) mg for 16 ± 7 days with no differences between groups (P = .49). To compare steroid responses between groups, we defined steroid CR, PR, and NR according to the peak platelet response to steroids. Using this definition, there were 4 CR, 1 PR, and 7 NR to steroid treatment. Of the 4 who had CR to steroids, 2 had NR to splenectomy, 1 had PR, and 1 had CR, with an average postsplenectomy platelet count of 72 ± 86 (range, 6-189). On the other hand, all 7 patients with NR to steroids had CR to splenectomy, with an average postsplenectomy count of 453 ± 200 (range, 274-837), P = .01. These findings are summarized in Table 3. Furthermore, when steroid responses were divided between those patients whose counts rose to at least 100 versus those whose counts remained below 100, the relationship between steroid and splenectomy response remains valid (P = .01, Fisher's Exact test exact test). When splenectomy and steroid responses are plotted against one another, there is an inverse relationship between peak platelet count with steroids and the postsplenectomy platelet count, depicted in Fig. 1 (correlation coefficient = −0.68, P = .01).
3. Discussion The first reported splenectomy for ITP was performed in Germany in 1916 [4], and splenectomy continues to be an important treatment for patients with uncontrolled chronic
J.H. Wood et al.
Fig. 1 Postsplenectomy platelet count as a function of peak platelet counts after steroid treatment. Solid black line represents exponential regression. Correlation coefficient = −0.68, P = .01.
ITP or with life-threatening hemorrhage complicating acute ITP [5]. However, despite more than 90 years of experience with splenectomy for ITP, there are still no clearly defined predictors of which patients will potentially benefit from the operation. The gravity of this ambiguity is highlighted by a survey of pediatric hematologists published in 2008 [2]. In this study, physicians were presented with a case scenario of chronic ITP in a child designed to meet qualifications for splenectomy according to the consensus statement of the American Society of Hematologists [6]. Remarkably, only 33% of surveyed pediatric hematologists indicated that they would recommend splenectomy for the child described in the vignette. The top 3 justifications for splenectomy avoidance were (1) long-term risk of sepsis, (2) possible lack of response to splenectomy, and (3) inability to predict response to splenectomy. In our patient population, 68% of patients undergoing splenectomy for ITP had a CR, defined as platelet counts of 150 or greater at initial follow-up with no evidence of recurrence during the full follow-up period. Response rates more than 90% have been reported in some series [7,8]. Although long-term remission rates of at least 70% to 80% are probably more realistic [9], there remains significant skepticism among referring pediatricians about the utility of splenectomy in the treatment of ITP, as demonstrated by 2 recent reviews [10,11] that underscore the perceived inability to predict successful splenectomy. Although several retrospective studies have attempted to identify preoperative predictors of successful splenectomy, few have focused exclusively on ITP in children, and the results of these studies are conflicting For example, a study published by the Intercontinental Childhood ITP study group suggested that older age, male sex, and increased duration of disease before splenectomy may portend improved outcomes in children, although the
Splenectomy in children with immune thrombocytopenic purpura Table 2
143
Platelet counts of splenectomy CR, PR, and NR groups at specific time-points
Diagnosis Presplenectomy POD 1 Postsplenectomy
CR (n = 13)
PR (n = 3)
NR (n = 3)
P
5.6 ± 3 (3-11) 24 ± 22 156 ± 104 (6-359) 399 ± 173 (189-837)
17.7 ± 15 (2-32) 23 ± 13 35 ± 42.5 (4-84) 64 ± 19 (50-86)
2.5 ± 2 (1-4) 7.3 ± 7 126 ± 109 (12-229) 20 ± 20 (6-43)
.08 .43 .19 .00075
POD 1 indicates postoperative day 1. Platelet counts are 103/μL. Data are mean ± SD (range).
authors acknowledge that these are not strong predictors [12]. In contrast, our findings, which are consistent with other pediatric studies [13,15], suggest that there are no differences in age, sex, or time to splenectomy between the CR, PR, or NR to splenectomy groups (Table 2). Likewise, we found that patients in each splenectomy response group had similar platelet counts at diagnosis, upon admission for splenectomy, and on postoperative day 1. These results corroborate the findings of the Intercontinental Childhood ITP study group [12]. The prognostic value of responsiveness to IVIG is an area of ongoing investigation, and the available data seem to support a correlation of favorable responses to IVIG and splenectomy [13-15], although poor response to IVIG may not correlate with failure to respond to splenectomy [15]. Our data reveal no relationship between splenectomy CR and responsiveness to IVIG, WinRho, or rituximab, although a type 2 error is likely, given the relatively small number of patients with these treatments in our study population. The remarkable finding of our study is a highly correlated, inversely proportional relationship between response to steroid treatment and response to splenectomy (Table 3 and Fig. 1). These findings challenge not only our own institutional assumptions but also the published results of some of the few exclusively pediatric studies available [13,16]. For example, Davis et al [13] described a direct correlation between steroid and splenectomy responsiveness. Patients in this study were treated with steroids in the immediate preoperative period, and their median platelet count at admission for splenectomy was 127. Our study, on the other hand, includes only those patients who received therapeutic corticosteroids and excludes patients who received steroids only in preparation for splenectomy. The difference between these 2 studies is highlighted by the fact that our mean platelet count at admission for splenectomy was 21. Table 3
Best response to steroids versus response to splenectomy
Response to splenectomy
Response to steroids CR
PR
NR
CR PR NR
1 1 2
0 0 1
7 0 0
P = .01, Fisher's Exact test.
In contrast, several other studies have failed to find any relationship between steroid responsiveness and surgical outcome [12,14,15,17-19]. Holt et al [14] reviewed 24 patients treated with prednisone for ITP and found no significant association between response to prednisone and response to splenectomy. However, neither the steroid regimen nor the method of categorizing steroid responses is explicitly described in the study. Hemmila et al [15] also found no correlation of steroid response to splenectomy response in 17 patients. However, 94% of patients in this series had either a “good” or “excellent” response to corticosteroids, whereas only 42% of patients in our study had either CR or PR to steroids. The reason for this discrepancy is unclear, but likely betrays a difference in patient populations. Furthermore, with only one steroid failure in the study by Hemmila et al, meaningful conclusions about the predictive value of steroid responsiveness are unlikely. Contributing to the difficulty of interpreting these data is a lack of standardized reporting of platelet response to steroid therapy. For example, platelet responses to steroids are described variously as mean platelet count after steroids [13] or platelet count within 1 month of steroids [15]. We defined responsiveness to medications as peak platelet counts during treatment, which made data interpretation simpler and more readily applicable to the clinical setting. Furthermore, we selected a subset of patients with easily comparable and complete data to evaluate for responses to steroid therapy and splenectomy. Using this approach, we found that a platelet count of 150 or more at any time during steroid treatment was strongly associated with NR or PR to splenectomy and that those patients who had NR to steroids universally had CR to splenectomy (Table 3). The explanation for this phenomenon is unclear, because steroid therapy and splenectomy theoretically have analogous functions: inhibiting antibody production and impairing phagocytosis by splenic macrophages. Given the complex and nonstandardized treatment regimens of ITP patients, it is possible that splenectomy responses were affected by medical intervention not accounted for in our data. Another possibility is that the patients who did not respond to splenectomy have qualitatively or quantitatively different antibodies than those who have CR to splenectomy. Furthermore, in the adult population, there is growing evidence that some patients with ITP have impaired thrombopoiesis leading to decreased platelet production
144 [20]. However, a cohesive explanation is lacking, and there is a need for translational investigation to uncover the mechanistic etiology for divergent responses to steroid treatment and splenectomy. One theoretical shortcoming of this study is that it cannot adequately define the period during which the “peak platelet count” will consistently be encountered, and studies to define this trial period would be very beneficial. This study further highlights the need for prospective analysis of large patient cohorts to fully describe the significance of steroid responsiveness as a predictor of response to splenectomy. However, despite these limitations, the data clearly undermine the long-held belief that steroid responsiveness portends a favorable outcome from splenectomy and suggests that it may be steroid-resistant children who are most likely to have CR to splenectomy.
References [1] Di Paola JA, Buchanan GR. Immune thrombocytopenia. Pediatr Clin North Am 2002;49(5):911-28. [2] Neunert CE, Bright BC, Buchanan GR. Severe refractory immune thrombocytopenic purpura during childhood: a survey of physician management. Pediatr Blood Cancer 2008;5:1513-6. [3] Law C, Marcaccio M, Tam P, et al. High-dose intravenous immune globulin and the response to splenectomy in patients with idiopathic thrombocytopenic purpura. N Engl J Med 1997;336:1494-8. [4] Kaznelson P. Vershwinder der hemmorhagischen diathese bei einem falle von “essentieller thombopenie” (Frank) nack Milzexstripation: splenogene thrombolytische purpura. Wien Klin Wochenschr 1916;29:1451-4. [5] Kliegman RM, Behrman RE, Jenson HB, et al. Nelson Textbook of Pediatrics. Philadelphia (Pa): Saunders; 2007. [6] George JN, Woolf SH, Raskob GE, et al. Idiopathic thrombocytopenic purpura: a practice guideline developed by explicit methods for the American Society of Hematology. Blood 1996;88:3-40. [7] Gadenstatter M, Lamprecht B, Klinger A, et al. Splenectomy versus medical treatment for idiopathic thrombocytopenic purpura. Am J Surg 2002;184:606-9. [8] Watts RG. Idiopathic thrombocytopenic purpura: a 10-year natural history study at the Children's Hospital of Alabama. Clin Pediatr 2004;43:691-702. [9] Mantadakis E, Buchanan GR. Elective splenectomy in children with idiopathic thrombocytopenic purpura. J Pediatr Hematol Oncol 2000;22:148-53. [10] Shad AT, Gonzalez CE, Sandler SG. Treatment of immune thrombocytopenic purpura in children: current concepts. Pediatr Drugs 2005;7:325-36. [11] Psaila B, Bussel JB. Immune thrombocytopenic purpura. Hematol Oncol Clin North Am 2007;21:743-59. [12] Kuhne T, Blanchette V, Buchanan GR, et al. Splenectomy in children with idiopathic thrombocytopenic purpura: a prospective study of 134 children from the Intercontinental Childhood ITP Study Group. Pediatr Blood Cancer 2007;49:829-34. [13] Davis PW, Williams DA, Shamberger RC. Immune thrombocytopenia: surgical therapy and predictors of response. J Pediatr Surg 1991;26:407-13. [14] Holt D, Brown J, Terrill K, et al. Response to intravenous immunoglobulin predicts splenectomy response in children with immune thrombocytopenic purpura. Pediatrics 2003;111:87-90. [15] Hemmila MR, Foley DS, Castle VP, et al. The response to splenectomy in pediatric patients with idiopathic thrombocytopenic
J.H. Wood et al.
[16]
[17] [18] [19]
[20]
purpura who fail high-dose intravenous immune globulin. J Pediatr Surg 2000;35:967-72. Weinblatt ME, Ortega JA. Steroid responsiveness: a predictor of the outcome of splenectomy in children with chronic immune thrombocytopenic purpura. Am J Dis Child 1982;136:1064-6. Wilde RC, Ellis KD, Cooper WM. Splenectomy for chronic idiopathic thrombocytopenic purpura. Arch Surg 1967;95:344-50. Block GE, Evans R, Zajtchuk R. Splenectomy for idiopathic thrombocytopenic purpura. Ann Surg 1966;92:484-9. Akwari OE, Itani KMF, Coleman RE, et al. Splenectomy for primary and recurrent immune thrombocytopenic purpura (ITP). Ann Surg 1987;206:529-41. Stasi RR, Evangelista ML, Stipa E, et al. Idiopathic thrombocytopenic purpura: current concepts in pathophysiology and management. Thromb Haemost 2008;99:4-13.
Discussion Unidentified Speaker: What is your hypothesis for the mechanism of this observation? James H. Wood, MD (Denver, CO): We had several hypotheses, none of which we considered adequate. Dr Hays is one of our coauthors. She is one of the hematologists and we discussed that issue at length. There could be different antibodies in these 2 groups that are interacting. There are data from the adult literature that would suggest that the subset of patients with ITP have as their primary defect a problem with platelet production instead of a problem with destruction. It may just be that we are talking about 2 different types of ITP, although again none of those give a full explanation. John Gosche, MD (Las Vegas, NV): In order to challenge the understanding that we have that responders to steroids are also responders—you should have more than just 12 patients. What are your plans to improve your numbers to see if what you have seen at your institution is comparable to what has been seen at other institutions? Wood, MD: The difficulty with doing a prospective trial, and one of the shortcomings of this study, is that these patients have been preselected because they have failed therapy. If they had not failed their therapy, they never would have had a splenectomy. When we are talking about complete responses to steroids or any other treatment, we are talking about transient responses at best. So it would be very difficult to have a rigorous trial to look at those differences, and we do not have any specific plans yet to study this prospectively, but that is certainly something that needs to be done to more adequately define not only this prognostic value but also the waiting period during which one should expect to see the peak platelet count so they can decide that the patient needs either splenectomy or continued medical therapy.