- Top of page
- Conflict of Interest Disclosure
- Supporting Information
While splenectomy is an effective therapy for primary immune thrombocytopenia (ITP), possible complications and observed non-complete response (CR) in one-third of patients demonstrate the need for further research into potential pre-surgical predictors of outcomes. Past investigations into platelet sequestration studies, a hypothesized predictive test, have adopted heterogeneous methods and varied widely with regard to power. By studying patients with primary ITP who underwent autologous 111In-labelled platelet sequestration studies at Barts and The London NHS Trust between 1994 and 2008, we evaluated the effectiveness of sequestration site in predicting short, medium, and long-term CR (platelet count >100 × 109/l) to splenectomy through multivariate (gender, age at splenectomy, and mean platelet lifespan) logistic regression modelling. In total, 256 patients with primary ITP underwent scans; 91 (35·5%) proceeded to splenectomy. Logistic regression revealed significant adjusted odds ratios for CR of 7·47 (95% confidence interval [CI], 1·89–29·43) at 1–3 months post-splenectomy, 4·85 (95% CI, 1·04–22·54) at 6–12 months post-splenectomy, and 5·39 (95% CI, 1·34–21·65) at last follow-up (median: 3·8 years [range: 0·5–13·1 years]) in patients with purely or predominantly splenic versus mixed or hepatic sequestration. These findings demonstrate the utility of autologous 111In-labelled platelet sequestration studies as an adjunct predictive instrument prior to splenectomy.
Primary immune thrombocytopenia (ITP) is an autoimmune disease characterized by autoantibody-mediated platelet destruction, T-cell-mediated platelet lysis, and suboptimal platelet production (Ballem et al, 1987; Olsson et al, 2003; McMillan et al, 2004). Primarily an acute (< 1 year) condition in children, its course is predominantly chronic among adults (Provan, 2003). The degree of bleeding in patients with primary ITP is largely, but not solely, dependent on the platelet count, with patients possessing counts below 10 × 109/l at greatest risk of a major haemorrhage (Frederiksen & Schmidt, 1999).
While primary ITP specialists have increasingly adopted a minimally interventional paradigm, therapy is still indicated for patients with symptomatic thrombocytopenia, at high risk of bleeding, or undergoing procedures likely to induce blood loss (Provan et al, 2010). Standard treatment modalities include oral corticosteroids, intravenous immunoglobulin (IVIg), and splenectomy (George et al, 1996).
The spleen serves as a principal site of anti-platelet antibody production and opsonized platelet phagocytosis (Kuwana et al, 2002; Cines et al, 2003). Splenectomy has therefore proved an effective therapy in the management of primary ITP, with a complete response (CR) observed in approximately two-thirds of adults (Kojouri et al, 2004). However, the procedure carries a risk of peri- and post-operative complications, including intrabdominal haemorrhage (Wanachiwanawin et al, 1993), thromboembolic events (Robinette & Fraumeni, 1977), and opportunistic post-splenectomy infections (Bisharat et al, 2001). While the combined risk of these complications in patients with primary ITP is estimated to be low (Schwartz et al, 2003), it provides sufficient impetus to investigate potential pre-operative predictors of response when coupled with a known risk of non-CR in a third of patients and limited data on long-term relapse.
To date, a number of such predictors have been posited, including (i) response to corticosteroid and IVIg therapies, (ii) platelet turnover and lifespan, (iii) patient age, (iv) duration of disease, (v) platelet-bound immunoglobulin, and (vi) site of platelet destruction as determined by radionuclide labelling techniques. In a systematic review of the effectiveness of splenectomy among adult patients with primary ITP, Kojouri et al (2004) reported inconclusive evidence in support of this last variable. Among a series of 15 studies, the authors found the site of platelet destruction to be correlated with post-splenectomy outcome in 6 (Burger et al, 1978; Russo et al, 1987; Lamy et al, 1993; Winde et al, 1996; Najean et al, 1997; Bourgeois et al, 2003) uncorrelated in 8 (Ries, 1977; den Ottolander et al, 1984; Fenaux et al, 1989; Siegel et al, 1989; Naouri et al, 1993; Louwes et al, 1999; Radaelli et al, 2000; Rossi et al, 2002), and inconclusive in one (Ikkala et al, 1978).
However, the methodologies of these studies were notably heterogeneous with regard to the isotopic label used, sequestration classification scheme, patient inclusion criteria, and both the definition of and period for response (Tables SI–IV). Moreover, 6 (67·8%) uncorrelated or inconclusive studies had fewer than 50 evaluable patients (Ries, 1977; Ikkala et al, 1978; Siegel et al, 1989; Naouri et al, 1993; Louwes et al, 1999; Rossi et al, 2002) undergoing both a platelet sequestration study and a splenectomy. The power of these investigations to detect an association between sequestration site and response to splenectomy was therefore limited. These shortcomings underscore the need for an additional suitably powered study, which incorporates currently standardized isotopic labelling techniques and consensus criteria for response evaluation.
Barts and The London NHS Trust is the principal ITP referral centre for adults in the United Kingdom (UK), and autologous 111In-labelled platelet sequestration studies have been routinely practiced here as part of the planning process prior to splenectomy since the mid-1990s. Using the cohort of patients with primary ITP who have undergone such studies, the objectives of our investigation were two-fold: (i) to evaluate the prognostic utility of platelet sequestration site as a predictor of short, medium, and long-term response to splenectomy through multivariate logistic regression modelling and (ii) to compare descriptively the long-term outcomes of splenectomy with alternate therapeutic approaches in patients with primary ITP.
- Top of page
- Conflict of Interest Disclosure
- Supporting Information
The results of our study revealed an increased likelihood of CR to splenectomy in patients exhibiting a purely or predominantly splenic pattern of platelet sequestration as shown through autologous 111In-labelled scanning. At last follow-up, patient age at splenectomy also emerged as an independent predictor of response, implicating a 5% increased odds of CR per year difference in younger candidates for surgery. This finding is consistent with a majority of past investigations (Kojouri et al, 2004).
The significance of platelet sequestration pattern as pre-surgical predictor of short, medium, and long-term CRs was not affected by the incorporation of primary ITP duration as an additional covariate in subgroup analyses (Table III). Of note, the association also did not appear to be tied to a difference in MPLS, which was similar between the cohorts (median- purely or predominantly splenic: 31·1 h [range: 7·2–271·0 h] vs. mixed or hepatic: 25·9 h [range: 6·8–407·0 h], P = 0·491). These data suggest that the rate of platelet destruction was independent of both the predominant location of platelet destruction and the likelihood of response to splenectomy in patients with primary ITP, a disparate conclusion than that reached by Siegel et al (1989).
Long-term outcomes, as measured by platelet count and ITP-specific treatment status, among patients indicated but not opting for splenectomy (the comparative, non-splenectomized cohort) were notably poor in comparison with splenectomized patients with purely or predominant platelet sequestration. However, they corresponded to those of the mixed or hepatic, splenectomized cohort with respect to ITP-specific treatment status. At last follow-up, 41·2% (28 of 68 patients) of the comparative, non-splenectomized cohort was on treatment as compared with 30·0% (6 of 20 patients) of splenectomized patients with mixed or hepatic sequestration (Table II). When excluding patients taking solely low-dose corticosteroids, the difference between these proportions narrowed: 30·0% (20 of 68 patients) vs. 25·0% (5 of 20 patients), respectively.
These results highlight splenectomy as an effective treatment for younger patients with primary ITP exhibiting a purely or predominantly splenic platelet sequestration. Short, medium, and long-term success was observed in 85–90% of all patients with such sequestration while reliance on treatment was limited to 4·3% (3 of 69 patients) a median of 3·8 years (range: 0·5–13·1 years) following surgery.
More controversially, these data further support the adoption of a cautious approach to splenectomy for mixed or hepatic patterned patients. While wide 95% CIs prevent precise classification of the ORs of non-CR in patients with mixed or hepatic versus purely or predominantly splenic sequestration, our findings illustrate the existence of a significant difference between the cohorts. Coupled with increased recognition of the physiological importance of the spleen, potential long-term vascular complications following splenectomy for haematological disorders (Crary & Buchanan, 2009), and arguably similar, clinically relevant long-term outcomes among patients indicated but not opting for splenectomy, this association raises the question as to whether patients with mixed or hepatic platelet sequestration may fare comparably without splenectomy. Undeniably, platelet counts in the comparative, control cohort were significantly lower than those of the mixed or hepatic, splenectomized cohort (median: 58 × 109/l [range: 2–358 × 109/l] vs. 199 × 109/l [range: 2–577 × 109/l], P < 0·001). However, most patients (57 of 68, 83·8%) in the former cohort had counts above 10 × 109/l (Fig 4), a conventional threshold for increased risk of major bleeding events (Provan & Newland, 2002). Furthermore, no fatal haemorrhages were observed in this group over a median of 2·7 years (range: 0·1–11·4 years).
Three methodological strengths make our study an important addition to the existing literature on the utility of platelet sequestration studies in patients with primary ITP. First, the use of 111In-tropolone may have contributed to more accurate test results than previous investigations using either 51Cr or 111In-oxine. 111In-tropolone has been reported by some research groups to exhibit better platelet labelling efficiency and imaging properties than 51Cr (Robertson et al, 1981). It is additionally more soluble than 111In-oxine in aqueous media, sparing possible platelet impairment from ethyl alcohol, used on occasion with the latter radionuclide label (Dewanjee et al, 1982).
Second, the labelling of autologous platelets enabled accurate adjustment for MPLS in our logistic regression models, preventing potentially spurious findings that may have resulted from alloimunization (Stratton et al, 1989). It was this substitution of homologous with autologous platelets by Ballem et al (1987) that helped illustrated the existence of suboptimal levels of platelet production among a segment of primary ITP population.
Third, the development of a cohort of patients indicated but not opting for splenectomy represented a novel construct by which to compare long-term outcomes in splenectomized and non-splenectomized patients. Although past investigations have aptly documented a high success rate for splenectomy in patients with primary ITP, they have left unanswered the question of whether equivalent outcomes were possible via an alternate treatment paradigm.
Four limitations to the study should similarly be noted. First, while data were collected on 91 patients undergoing splenectomy following an autologous 111In-labelled platelet sequestration study, the results of the scan were not blinded to patients or their haematologists. As a result, only 20 (20·4%) patients with mixed or hepatic sequestration elected to undergo surgery. The significant difference in the proportion of patients with mixed or hepatic versus purely or predominantly splenic sequestration proceeding to splenectomy (P < 0·001; Fig 2) raises the possibility of selection bias, namely that mixed or hepatic patterned patients who opted for splenectomy were worse off than those who did not. Though plausible, it is similarly possible for haematologists to have encouraged surgery among mixed or hepatic patterned patients deemed more likely to succeed. To address this issue, post-hoc analyses were conducted on a subset of patients for whom pre-surgery platelet counts and post-diagnosis duration were available. Neither variable differed significantly (P = 0·59 and P = 0·83, respectively). Furthermore, mixed or hepatic patterned patients who underwent splenectomy were on average eleven years (P = 0·01) younger than those who did not, suggesting that non-adjusted results would have likely been biassed toward, rather than away from, the null.
Second, though advocated by the International Working Group on ITP (Rodeghiero et al, 2009), the use of 100 × 109/l as a platelet count threshold for CR may have discounted clinically relevant responses to splenectomy. Ideally, effectiveness would have been assessed by therapeutic impact on not only platelet count but also the bleeding events and health-related qualify of life. However, sufficient data were not available to gauge the effect of splenectomy on the latter two variables, an inherent limitation of retrospective investigations. Despite this shortcoming, the relative distributions of platelet counts among the purely or predominantly and mixed or hepatic, splenectomized cohorts at each of the assessment intervals suggest that a greater proportion of patients within the latter cohort continued to have platelet counts placing them at increased risk of major haemorrhage (i.e. < 10 × 109/l) following surgery (Figs 3 and 4).
Third, the algorithm used to capture MPLS yielded skewed results for patients with near-normal platelet lifespans, as it has been shown to perform poorly in situations of uniform destruction (i.e. normal MPLS) (Maisey et al, 1998). These errant results at the normal end of the MPLS spectrum may have been compounded by the limited time points at which blood samples were taken (0·5, 3, 24, and 48 h post-injection), which may have hindered accurate regression of platelet associated radioactivity on time. However, as the vast majority of study patients, 94·5% (225 of 238 patients for whom MPLS was evaluable), exhibited a markedly reduced MPLS of < 4 d, it is unlikely that the multivariate models were appreciably biassed.
Fourth, reasons, supplied by haematologists, for patients not undergoing splenectomy were subjective and susceptible to possible recall bias. Use of these reasons nevertheless likely resulted in an improved picture of probable long-term outcomes had patients in the splenectomized cohort not opted for surgery than evaluation of results from the non-splenectomized cohort as a whole.
In conclusion, while this investigation demonstrates utility in including autologous 111In-labelled platelet sequestration studies as an adjunct predictive tool prior to splenectomy in primary ITP, the aforementioned limitations illustrate that considerable work is still needed on this topic. Potentially useful next steps include a meta-analysis of observational studies and the development of an international prospective cohort study, the latter of which would permit critical appraisal of proposed refinements to the scanning protocol and the incorporation data from autologous 111In-labelled platelet sequestration studies, which were beyond the scope of this investigation (e.g. platelet production levels).