Clinical and laboratory factors associated with platelettransfusion refractoriness: a case–control study

INAKI ALCORTA, ARTURO PEREIRA AND ANTONI O ORDINAS Service of Haemotherapy and Haemostasis,Hospital Clınic, Barcelona, Spain

Received 29 September 1995; accepted for publication 7 December 1995

Summary. In recent years clinical factors have largelysurpassed alloimmunization as the predominant cause ofplatelet refractoriness. This makes it necessary to properlyidentify and weigh the non-immune factors that have a majorimpact on refractoriness. A case–control study is suitable forsuch an analysis, and to our knowledge has not previouslybeen performed to assess this issue.

Fifty-two refractory patients were compared with 52control patients who were transfused at the same time.Only one transfusion event was analysed per patient. Clinicaland laboratory data were recorded at the time of selectedtransfusion, and their association with refractoriness wasinvestigated by the contingency table method and the Coxstepwise logistic regression.

There were 16 (31%) patients with HLA antibodies in theindex group and only one in the control group. The correctedcount increment in the group of patients refractory dueto HLA antibodies was significantly lower than that in non-

alloimmunized refractory patients [median (range): 48.5(ÿ3560, 4614) and 4058 (ÿ4417, 6886), respectively;U�493, P<0.0001]. In the multivariate analysis, factorsassociated with refractoriness were the presence of HLAantibodies (odds ratio (OR) 50.7; 95% CI 5.5–463); fever(odds ratio 7.2; 95% CI 2.5–21) and BMT because of chronicmyeloid leukaemia (odds ratio 7.3; 95% CI 1.8–30). Thelatter two were the only factors that remained independentlyassociated with refractoriness after excluding alloimmunizedpatients and their controls.

We conclude that HLA antibodies are strongly associatedwith platelet transfusion refractoriness, but account for lessthan a third of these patients. Fever and BMT because ofchronic myeloid leukaemia were the only non-immunefactors independently associated with refractoriness.

Keywords: platelet, transfusion, HLA, antibodies, case–control.

Platelet transfusion has become standard practice in themanagement of thrombocytopenia due to bone marrowfailure. It is usually provided on a prophylactic basis, in orderto maintain the platelet count above a level that is consideredsafe. Platelet count increments or platelet recovery 1 h aftertransfusion are used to monitor response (Slichter, 1990).Unfortunately, 20–60% of patients requiring regular platelettransfusions fail to achieve adequate post-transfusion incre-ments and are considered to be refractory to platelettransfusions (Weiner & Kao, 1992). These patients presenta serious therapeutic challenge.

Prior to the early 1980s, alloimmunization against HLAantigens was the main cause of platelet transfusion refrac-toriness. Thus, in a study by Hogge et al (1983), 86% of theirrefractory patients had lymphocytotoxic antibodies. However,in recent years clinical factors are increasingly being reported

as the predominant cause of platelet refractoriness (Doughtyet al, 1994; Friedberg & Mintz, 1995). This change has severalimportant implications. First, it can be anticipated thatmeasures intended to prevent alloimmunization, such asleucodepletion of blood components, will have a limitedimpact on the prevalence of platelet refractoriness. Second,any effort aimed at preventing or managing non-immunerefractoriness will require the identification of those clinicalfactors that have a major impact on refractoriness.

The complexity and multiplicity of medical problemspresent in myelosuppressed patients requiring platelet trans-fusions makes it difficult to properly identify and weigh theclinical factors that are implicated in refractoriness. Multi-variate statistical methods and a large number of patients areneeded (Slichter, 1990). In order to increase the statisticalpower of previous studies, each transfusion has beenconsidered as an independent event. However, this may leadto overestimation of the influence of characteristics present inpatients who receive more transfusions and ignores thepotential correlation between transfusions. To the contrary,

British Journal of Haematology, 1996, 93, 220–224

220 # 1996 Blackwell Science Ltd

Correspondence: Dr I. Alcorta, Service of Haemotherapy andHaemostasis, Hospital Clınic, Villarroel 170, 08036 Barcelona,Spain.

221Platelet Transfusion Refractoriness

# 1996 Blackwell Science Ltd, British Journal of Haematology 93: 220–224

the repeated sampling method (Bishop et al, 1988) mightoverestimate the impact of factors present in those whoreceive few transfusions. Both methods have the potential ofoveremphasizing the influence of factors that are present in aminority of patients, and that, therefore, have a relativelysmall impact on the overall prevalence of non-immunerefractoriness. In addition, in onco-haematology patients,myelosuppressive regimens, antibiotic protocols and othertherapeutic procedures change with time, as do some clinicalproblems, such as the class of microorganisms responsible forinfections. Platelet production techniques may also changewithin a long study period, making analysis more difficult.

In order to overcome the above difficulties, we haveperformed a case–control study of risk factors for platelettransfusion refractoriness, considering only one transfusionper patient, and in which controls were matched with casesby the date of index transfusion.

MATERIALS AND METHOD

Patients. At our institution, patients in a prophylacticplatelet transfusion programme for thrombocytopenia asso-ciated with bone marrow failure are given 6–8 units ofrandom donor platelet concentrates when their platelet countis 520� 109/l. Response to platelet transfusions is system-atically monitored by 1 h and 16–24 h post-transfusionplatelet counts. Patients with an inadequate 1 h post-transfusion count are given pools of ABO-identical plateletconcentrates which have been stored for 53 d. The numberof platelets in the concentrate is recorded, as well asimmediate pretransfusion and 1 h post-transfusion plateletcounts. The corrected count increment (CCI) is thereforecalculated according to the formula (Weiner & Kao, 1992):CCI� platelet count increment (/�l)� body surface area(m2)/number of platelets transfused (�1011). Patients witha CCI of 57500 on at least two consecutive occasions areconsidered refractory to platelet transfusions. All refractorypatients identified from January 1993 to November 1994were included in this study, and their second transfusion witha low CCI was selected as the case-index transfusion. For eachcase, the random-number table method was used to select acontrol among non-refractory patients who were on aprophylactic platelet transfusion programme for bonemarrow failure during the same period as his or her corre-sponding case, and their nearest transfusion to that of thecase-index was selected as the control-index transfusion.Onlyone transfusion event was analysed per patient and there wasno crossing-over among cases and controls on different timeperiods.

Platelet concentrates. During the first year of the study,platelet concentrates were obtained by the platelet-richplasma method (Walker, 1993). In the second year theywere obtained from the buffy-coat of whole blood units(Pietersz et al, 1985). In both cases platelet concentrates werestored in PL 1240 plastic containers (Baxter Healthcare,Deerfield, Ill., U.S.A.) at 228C on a flat-bed agitator.

Pretransfusion evaluation. Parameters of cases and controlswere recorded at the time of the index transfusion andincluded the following: sex, age, ABO and Rh blood group,

diagnosis, previous febrile non-haemolytic transfusion reac-tions, spleen status (palpable or not), temperature, clinicallyor microbiologically documented infection, antithymocyticglobulin (ATG) treatment (concurrent or in the 15 previousdays), administration of amphotericin B, vancomycin, pent-oxifylline, ranitidine or cyclosporine A, number of systemicantibiotics, bleeding grade, presence of veno-occlusive disease ofthe liver (VOD) or graft-versus-host disease (GVHD), HLAantibodies, and prothrombin ratio. For the purpose of thepresent analysis, temperature was dichotomized as5388C or5388C, age was dichotomized by its median value inthe whole series, and bleeding grade was classified asnone, minor (mucous or cutaneous) or major (intracranial,pulmonary, gastrointestinal or urologic bleeding).

Laboratory investigations. At our institution, all patients on aprophylactic platelet transfusion programme are screened forHLA antibodies at weekly intervals. The antibody screening isperformed against a panel of 18 lymphocytes of selectedphenotype, by using the standard microlymphocytotoxicityassay (Terasaki et al, 1978). Refractory patients withoutdemonstrable lymphocytotoxic antibodies are systematicallyinvestigated for the presence of platelet antibodies by means ofa solid-phase technique (Capture-P, Immucor, Norcross, Ga.,U.S.A.) (Rachel et al, 1985). For the purpose of the presentanalysis, reactivity against 25% or more of the panellymphocytes or a clear-cut monospecific reaction patternwere considered a positive result. In all refractory patients,plasma D-dimer levels were measured in a sample obtainedthe day of transfusion by a latex technique that has a thresholdof 500 ng/ml fibrinogen equivalents (Fibrinosticon R, OrganonTeknica, Boxtel, Holland).

Statistical methods. An unconditional analysis was per-formed (Breslow & Day, 1984). The unadjusted association ofpatients’ characteristics with platelet transfusion refractori-ness was investigated by means of the contingency tablemethod. For each variable, the odds ratio with its 95%confidence interval, and the P value were calculated.Variables showing an odds ratio 42 or a P50:1 wereentered in the stepwise logistic regression model (Cox, 1970).In this model the maximal likelihood method was used forentering or removing terms at each step, and enoughiterations were allowed to achieve the following convergencecriteria: relative improvement in the log likelihood 510ÿ5

and relative change in parameter estimates 510ÿ4. Good-ness-of-fit of data to the logistic model was assessed by theBrown test. The BMDP statistical package was used for allcomputations.

RESULTS

Fifty-two patients refractory to random donor platelettransfusions were identified during the 2-year period. Therewere 21 women and 31 men, with a median age of 40 (range16–71) years. Only 16 (31%) cases had HLA antibodies, andnone of the remaining 36 cases had platelet-specificantibodies. Plasma D-dimer determinations were negative inall refractory patients. The diagnosis of these patients wasallogeneic bone marrow transplantation (BMT) in 19 (13because of chronic myeloid leukaemia), autologous BMT in 18,

acute leukaemia in nine, and other diagnosis in six. 25 caseshad a documented infection at the time of the indextransfusion, and included 21 with bacteraemia (14 of themby coagulase-negative staphylococci), three with pneumoniaand one with enterocolitis. The CCI in the group of patientsrefractory because of HLA antibodies was significantly lowerthan that in non-alloimmunized refractory patients [median(range): 48.5 (ÿ3560, 4614) and 4058 (ÿ4417, 6886),respectively; U � 493, P50:0001].

The control group included 20 women and 32 men, with amedian age of 39.5 (range 15–64) years. One of these controlpatients had lymphocytotoxic antibodies. Platelet-specificantibodies and plasma D-dimer were not investigated in thecontrol group. The diagnosis of control patients wereallogeneic BMT in 15 (five because of chronic myeloidleukaemia), autologous BMT in 18, acute leukaemia in 17,and other diagnosis in two. 14 control patients had adocumented infection at the time of the index transfusion.They included 12 with bacteraemia (nine of them bycoagulase-negative staphylococci), and two with pneumonia.

Table I shows the clinical and analytical data with an oddsratio 42 for their association with platelet transfusionrefractoriness. Allogeneic BMT, when considered irrespectiveof the initial diagnosis, did not show a significant associationwith refractoriness (OR 1.42, 95% CI 0.62–3.23; P40:01).This was also the case for concurrent treatment withamphotericin B, which was present in 28 cases and 24controls (OR 1.36, 95% CI 0.63–2.93; P40:1).

To determine the independent association of factorssummarized in Table I with platelet transfusionrefractoriness,and their adjusted ORs, they were introduced in a stepwiselogistic regression analysis. In this multivariate analysis, onlythe presence of HLA antibodies, BMT for chronic myeloidleukaemia (CML) and fever remained significantly associatedwith refractoriness. Their coefficients, adjusted ORs and 95%CIs of the OR are shown in Table II.

An additional analysis was performed excluding the 16cases with HLA antibodies and their respective controls. Thefactors that remained independently associated with platelet

# 1996 Blackwell Science Ltd, British Journal of Haematology 93: 220–224

222 I. Alcorta, A. Pereira and A. OrdinasTable I. Factors associated with refractoriness by univariate analysis.

No. of patients*

Cases Controls Unadjusted 95% CI ofFactor (n � 52) (n � 52) odds ratio odds ratio P value

Febrile reactions 27 (51) 17 2.31 1.04–5.14 0.038Splenomegaly 7 1 7.94 0.94–66.97 0.066BMT for CML 13 5 3.13 1.02–9.54 0.038Fever† 29 12 4.20 1.8–9.79 0.0006Major haemorrhage 9 3 3.41 0.86–13.42 0.066Documented infection 25 14 2.51 1.11–5.69 0.026Suspected VOD 10 3 3.89 1–15.07 0.038ATG 5 1 5.43 0.61–48.2 0.2Vancomycin 31 21 2.17 0.99–4.75 0.05Four or more antibiotics 14 5 3.46 1.14–10.47 0.022Blood group not A 36 26 2.25 1–5.01 0.045PT ratio 41.2 18 (51) 8 (51) 2.95 1.14–7.61 0.023HLA antibodies 16 (51) 1 (48) 21.87 2.71–166.55 0.0001

* Figures indicate the number of patients that presented each factor. In parentheses the number of evaluatedpatients if there were missing data.

† Defined as a temperature5388C.

Table II. Logistic regression analysis of factors associated with plateletrefractoriness.

Adjusted 95% CI ofFactor Coefficient odds ratio odds ratio P value

HLA antibodies 3.925 50.7 5.54–463 50.001Fever 1.973 7.19 2.47–21 50.001BMT for CML 1.996 7.36 1.81–29.9 0.002Constant ÿ1.633

Table III. Regression analysis of factors associated with refractorinessafter patients with HLA antibodies and their controls were excluded.

Adjusted 95% CI ofFactor Coefficient odds ratio odds ratio P value

Fever 1.218 3.38 1.07–10.7 0.005BMT for CML 1.122 3.07 0.738–12.8 0.018Infection 1.306 3.69 1.13–12 0.057Splenomegaly 2.034 7.64 0.679–85.9 0.058Constant ÿ1.334

223Platelet Transfusion Refractoriness

# 1996 Blackwell Science Ltd, British Journal of Haematology 93: 220–224

transfusion refractoriness in this subgroup of patients wereBMT for chronic myeloid leukaemia and fever, whereasthe presence of a documented infection and splenomegalyattained a borderline significance (Table III).

DISCUSSION

A case–control study design was used to analyse the factorsimplicated in platelet transfusion refractoriness. This design,by matching the date of transfusion, avoids handicaps due tochanges over time in the case-mix and in clinical and blood-bank practices influencing response to platelet transfusions.Only one transfusion per patient was considered, to avoid biasdue to the influence in the final results of patients with eithermany or very few transfusions. In the present study thefactors independently associated with platelet transfusionrefractoriness included HLA alloimmunization, BMT becauseof CML, and fever.

Alloimmunization to HLA antigens was recognized asthe main cause of platelet transfusion refractoriness in the1980s (Daly et al, 1980; Hogge et al, 1983). HLA antibodiesmediate a rapid clearance of transfused platelets, becauseeven 10 min after transfusion only a minor fraction of thetransfused platelets are in circulation (O’Conell et al, 1988).In our study, not surprisingly, the presence of HLA antibodieswas the single most powerful predictor of platelet refractori-ness. However, only 31% of our refractory patients werealloimmunized. This is in contrast with previously reportedincidence of alloimmunization in refractory patients (Hoggeet al, 1983) and is much lower than the incidence observed byus 10–15 years ago. Other authors have also reported thisapparent decline in the incidence of alloimmunization, thereasons for which are not clear (Doughty et al, 1994).

Bishop et al (1988) were probably the first to observe adeleterious effect of BMT on CCIs. However, only seven outof their 139 patients were BMT recipients. In our seriesallogeneic BMT per se was not associated with refractoriness.Only the subset of patients transplanted because of a CMLshowed an increased risk of refractoriness. Small numberspreclude any stratification of the CML-BMT group, so we canonly speculate on the causes of this increased risk. In themajority of these patients, refractoriness was documented inthe first 30 d after transplant, so an effect of the conditioningregime and/or the underlying disease itself could be enter-tained. In fact, the majority of CML-BMT refractory patientswere considered to be at high risk for GVHD, and received aT-cell-depleted BMT. This protocol was accompanied by ahigher dose of radiotherapy, to prevent graft failure (1300 cGyin four fractions, instead of standard 1200 cGy). Also, therewere five unrelated BMTs in this group, who were con-ditioned with 1300 cGy, thiotepa, cyclophosphamide andATG. Therefore it is possible that CML patients experiencedmore endothelial damage due to this intensive conditioningregime. Alternatively, persistence of hypersplenism in theabsence of a palpable spleen in CML patients cannot be ruledout.

Fever was found to be associated with poor responseto platelet transfusions by the pioneers of this therapy(Freireich et al, 1963); other authors have also reported this

association (Parker et al, 1974; McFarland et al, 1989). In ourseries, fever was also associated with poor transfusion results.Fever is produced by the action of circulating cytokines, suchas IL-1 and TNF, on the hypothalamus, resulting in thesecretion of prostaglandinE2, which mediates the elevation ofbody temperature (Saper & Breder, 1994). These cytokinesare produced in response to a variety of inflammatory,infectious and neoplastic conditions. Besides other biologicalactions, they are known to promote endothelial cell activa-tion, with expression of adhesion molecules and procoagulantactivity (Waage & Steinshamn, 1993). The endothelialactivation could be responsible for the rapid clearance oftransfused platelets and would explain the association of feverwith platelet transfusion refractoriness.

In contrast with other studies (Kulpa et al, 1981; Bishopet al, 1988), we did not find an association with concurrentamphotericin B administration. Although our study had apotency of 90% or greater to detect a difference in risk due toamphotericin B only if the increase in risk was 4-fold or more,other factors should also be considered. First, amphotericin Bis typically administered to those myelosuppressed patientswith persistent fever or demonstrated fungal infections. It ispossible, therefore, that in previous studies amphotericin hadselected a subpopulation prone to refractoriness due to othercauses. Second, in vivo infusion of amphotericin B does notshorten the life-span of normal platelets, and its effect on themembrane lesion of stored platelets is unlikely to affect theirsurvival in vivo (Murphy, 1992).

Splenomegaly has been identified as a cause of plateletrefractoriness in many studies (Parker et al, 1974; Bishop et al,1988; Hussein et al, 1990). In our series there were fewpatients with splenomegaly, which may account for the lackof statistical significance in the multivariate analysis. Infact, when only patients without HLA antibodies and theircontrols were analysed there was a trend toward increasedrisk of refractoriness in patients with a palpable spleen.However, it should be noted that the low frequency ofsplenomegaly indicates that this factor cannot have asignificant impact on the overall prevalence of refractorinessin myelosuppressed patients. In addition, the association ofsplenomegaly with refractoriness is far from being absolute.Hussein et al (1990) found that 42% of platelet transfusionsadministered to splenomegalic patients produced CCIs 47500,and a large proportion achieved clinically helpful plateletcount increments. The same would apply to DIC, which cancertainly account for platelet refractoriness when it is present(Slichter, 1990). However, it was ruled out by laboratory datain each of our refractory patients, which indicates that DIC isa rather infrequent finding in myelosuppressed patients.

Finally, a finding that may be of interest in themanagement of refractory patients is the different rateof clearance of transfused platelets in alloimmunized andnon-alloimmunized refractory patients. It is possible thatselected non-alloimmunized refractory patients at high riskfor haemorrhage could be maintained at a ‘safe’ plateletcount by transfusing a higher dose of platelets morefrequently (e.g. twice daily). To the contrary, in alloimmu-nized patients it seems virtually impossible to achieve a ‘safe’platelet count with random donor platelet concentrates.

ACKNOWLEDGMENT

This work was partially supported by grants FIS 92/0794 andFIS 94/1098 from the Ministerio de Sanidad y Consumo ofSpain.

REFERENCES

Bishop, J.F., McGrath, F., Wolf, M.M., Matthews, J.P., De Luise, T.,Holdworth, R., Yuen, K., Veale, M., Whiteside, M.G., Cooper, I.A. &Szer, J. (1988) Clinical factors influencing the efficacy of pooledplatelet transfusions. Blood, 71, 383–387.

Breslow, N.E. & Day, N.E. (1984) Statistics of case–control studies.Statistical Methods for Cancer Research (ed. by R. G. Cornell), pp. 91–126. Marcel Dekker, New York.

Cox, D.R. (1970) The Analysis of Binary Data, pp. 90–98. Mathew andCo., London.

Daly, P.A., Schiffer, C.A., Aisner, J. & Wiernik, P.H. (1980) Platelettransfusion therapy: one-hour posttransfusion increments arevaluable in predicting the need for HLA-matched preparations.Journal of the American Medical Association, 243, 435–438.

Doughty, H.A., Murphy, M.F., Metcalfe, P., Rohatiner, A.Z.S., Lister,T.A. & Waters, A.H. (1994) Relative importance of immune andnon-immune causes of platelet refractoriness. Vox Sanguinis, 66,200–205.

Freireich, E.J., Kliman, A., Gaydos, L.A., Mantel, N. & Frei, E. (1963)Response to repeated platelet transfusions from the same donor.Annals of Internal Medicine, 59, 277–287.

Friedberg, R.C. & Mintz, P.D. (1995) Causes of refractoriness toplatelet transfusion. Current Opinion in Hematology, 2, 493–498.

Hogge, D.E., Dutcher, J.P., Aisner, J. & Schiffer, C.A. (1983)Lymphocytotoxic antibody is a predictor of response to randomdonor platelet transfusion. American Journal of Hematology, 14,363–369.

Hussein, M.A., Lee, E.J. & Schiffer, C.A. (1990) Platelet administeredto patients with splenomegaly. Transfusion, 30, 508–510.

Kulpa, J., Zaroulis, C.G., Good, R.A. & Kutti, J. (1981) Altered platelet

function and circulation induced by amphotericin B in leukemicpatients after platelet transfusion. Transfusion, 21, 74–76.

McFarland, J.G., Anderson, A.J. & Slichter, S.J. (1989) Factorsinfluencing the transfusion response to HLA-selected apheresisdonor platelets in patients refractory to random platelet con-centrates. British Journal of Haematology, 73, 380–386.

Murphy, S. (1992) Amphotericin B and platelet transfusion.Transfusion, 32, 7–8.

O‘Connell, B., Lee, E.J. & Schiffer, C.A. (1988) The value of 10-minuteposttransfusion platelet counts. Transfusion, 28, 66–67.

Parker, R.D., Yamamoto, L.A. & Miller, W.R. (1974) Interactioneffects analysis of platelet transfusion data. Transfusion, 14,567–573.

Pietersz, R.N.I., Loos, J.A. & Reesink, H.W. (1985) Platelet concen-trates stored in plasma for 72 hours at 228C prepared frombuffycoats of citrate-phosphate-dextrose blood collectedin a quadruple-bag saline-adenine-glucose-mannitol system. VoxSanguinis, 49, 81–85.

Rachel, J.M., Sinor, L.T., Tawfik, O.W., Summers, M.L. & Beck, W.L.B.(1985) A solid-phase red cell adherence test for platelet cross-matching. Medical Laboratory Sciences, 42, 194–195.

Saper, C.B. & Breder, C.D. (1994) The neurologic basis of fever. NewEngland Journal of Medicine, 330, 1880–1886.

Slichter, S.J. (1990) Mechanisms and management of plateletrefractoriness. Transfusion Medicine in the 1990’s (ed. by S. T. Nance),pp. 95–179. American Association of Blood Banks, Arlington, Va.

Terasaki, P.L., Bernocco, D., Parks, M.S., Ozturk, G. & Imachi, Y.(1978) Microdoplet testing for HLA-A, -B, -C, and -D antigens.American Journal of Clinical Pathology, 69, 103–107.

Waage, A. & Steinshamn, S. (1993) Cytokine mediators of septicinfections in the normal and granulocytopenic host. EuropeanJournal of Haematology, 50, 243–249.

Walker, R.H. (1993) (ed.) Technical Manual, 11th edn, pp. 730–731.American Association of Blood Banks, Bethesda, Md.

Weiner, R.S. & Kao, K.J. (1992) Clinical and laboratory diagnosis ofthe refractory state. Clinical Decisions in Platelet Therapy (ed. by S. R.Kurtz and D. B. Brubaker), pp. 73–85. American Association ofBlood Banks, Bethesda, Md.

# 1996 Blackwell Science Ltd, British Journal of Haematology 93: 220–224

224 I. Alcorta, A. Pereira and A. Ordinas