Radiotherapy-Associated Neutropenia and Thrombocytopenia: Analysis of Risk Factors and Development of a Predictive Model

NEUTROPENIA AND thrombocytopenia are common complications of extended-field radiotherapy1 that often cause treatment interruptions and increase the risks of infection or bleeding. Clinically significant myelosuppression may also occur in patients undergoing treatment to smaller radiotherapy fields who have received chemotherapy before or during radiotherapy. Unscheduled interruptions in radiotherapy have been associated with a reduced probability of local control of tumors of the breast,2 uterine cervix,3 lung,4 and upper aerodigestive tract,5,6 often in patients receiving potentially curative treatments. It is likely that tumor cells repopulate rapidly during breaks in treatment and consequently become more difficult to eradicate.7 Because of these findings, radiation oncologists endeavor to keep interruptions in radiotherapy to a minimum, particularly when treatment is administered with curative intent.

In recent years there has been a trend in cancer management towards increased use of combined modality therapy (CMT) for a variety of tumor types. CMT protocols frequently involve concurrent or sequential chemotherapy and radiation therapy. They are usually associated with higher rates of local tumor control, but also carry a higher risk of myelosuppression than single modality therapy. Some protocols prospectively use hematopoietic growth factors to allow intensification in chemotherapy dosage.8 CMT protocols are currently in extensive use for small-cell9 and non–small-cell lung cancer,10 head and neck tumors,11 esophageal carcinoma,12 pancreatic carcinoma,13 colorectal14 and anal15 carcinoma, breast carcinoma,2,8 pediatric malignancies,16 and lymphomas.17 As a consequence of CMT, significant myelosuppression will be encountered with increasing frequency in radiotherapy departments.

It has been shown that granulocyte colony-stimulating factor (G-CSF ) administration can rapidly reverse neutropenia caused by radiotherapy alone18,19 and effectively prevents neutropenia when used prophylactically at reduced dose in radiotherapy patients at high risk of radiation-induced neutropenia.20 G-CSF has also proved effective in treating neutropenia arising in radiotherapy patients who have previously received chemotherapy.19,21 A new platelet growth factor, Mpl ligand, is currently under evaluation in phase I trials.22 It is possible that such an agent could be useful in preventing or treating thrombocytopenia associated with radiotherapy and chemotherapy. An effective platelet growth factor could eliminate delays in radiotherapy due to thrombocytopenia and avoid the well-known risks of platelet transfusions derived from multiple donors.23 

It would be useful to identify prospectively those radiotherapy patients who are at highest risk of neutropenia and thrombocytopenia. High-risk patients could be monitored more closely for myelosuppression and infection or bleeding and could be candidates for clinical trials involving the administration of hematopoietic growth factors in an effort to prevent treatment interruptions and other complications of myelosuppression. In this study, we have reviewed the records of patients treated at Stanford University in the period from 1989 to 1994 who were known to have treatment interruptions associated with thrombocytopenia, neutropenia, or both. A control group of patients treated during the same period was randomly selected. We anticipated that comparison of cases and controls would allow us to identify those factors most strongly correlated with an increased risk of having a treatment interruption due to myelosuppression and that we might be able to develop a model that could prospectively identify patients at high risk. The factors that we studied included previous or concurrent administration of myelosuppressive chemotherapy, the percentage of active marrow irradiated in current and previous radiation treatments, performance status, age, sex, coadministration of potentially myelosuppressive drugs, and the extent of metastasis of malignant disease.

Our primary objective was to determine what factors most strongly influence the likelihood that a patient will have a radiation treatment delay of 2 days or more associated with thrombocytopenia or neutropenia (or both). As a secondary objective, we examined (1) the risk that the patient would have a greater than 10% reduction in the total radiation dose compared with the initial treatment plan; (2) the incidence of severe thrombocytopenia (platelet count, <25 × 10⁹/L); (3) requirements for platelet transfusion; and (4) the risk of having a treatment interruption that exceeded 3 days.

The study analyzed radiotherapy treatments that were completed between June 1989 and August 1995. Investigators used a database containing detailed information on all radiation therapy treatments administered at Stanford since the 1950s to identify patients who had unscheduled treatment interruptions of 2 days duration or more (excluding weekends and holidays). Review of the charts of these patients allowed the identification of patients who had treatment interruptions associated with significant neutropenia, thrombocytopenia, or both (the cases). The point at which the decision was made to interrupt radiotherapy because of thrombocytopenia or neutropenia varied between physicians. Whether an individual patient became a case in this study or continued with radiotherapy, with or without platelet transfusion, often depended on the clinical judgment of the radiation oncologist. Blood count data, including differential and platelet counts, were recorded. Although in some instances, few counts were available for the time during which patients were receiving radiation therapy, this was usually because the counts were normal and the attending physicians felt that the patients were at very low risk for radiation-induced myelosuppression. All of the patients had at least one complete blood count performed during treatment.

The database facilitated random selection of a representative set of controls from the entire population of patients who received radiotherapy during the study period, most of whom did not have treatment interruptions due to neutropenia or thrombocytopenia. Randomization was stratified by year. When selecting the controls, we randomly sampled courses of radiotherapy rather than initially listing all patients and then sampling them with equal probability. Thus, we weighted the likelihood of selecting an individual patient greater if that patient had received multiple courses of radiotherapy. If the same patient was selected more than once, the first course selected was to be used. However, no patient was selected twice.

We initially set out to study treatment interruptions associated with thrombocytopenia alone and, having identified our potential cases, chose to use a randomly selected control group that would bring the total number of charts reviewed to approximately 200 (cases and controls). We subsequently decided also to study cases of treatment interruption associated with neutropenia and used the same control group as for the thrombocytopenia cases, because none of the controls had a treatment interruption associated with neutropenia. Total body irradiation or stereotactic radiosurgery patients and patients who received brachytherapy or electron treatments only were excluded. No patient who was identified as a case was also randomly selected as a control. In this investigation, we were primarily interested in patients receiving radiotherapy for malignant disease. We identified 11 cases of treatment interruption associated with thrombocytopenia and 10 associated with neutropenia in patients undergoing radiotherapy for benign diseases, principally extended field treatments for cardiac transplant rejection. These patients were excluded from the analysis and the final number of thrombocytopenia cases and controls was 183.

The charts of both cases and controls were reviewed. Detailed information on the extent of treatment interruption was gathered for each case. All cases had at least one blood count performed during the treatment course that showed at least grade 1 neutropenia and/or thrombocytopenia. In addition, information was collected for both cases and controls on a range of parameters that were considered possible risk factors for neutropenia, thrombocytopenia, or bleeding. These parameters are described fully below. Detailed information concerning each patient's clinical course, medical history, and radiotherapy were found within the Stanford radiotherapy chart. In cases in which details of previous chemotherapy and other clinical variables were not adequately documented in the radiotherapy chart, the patient's medical oncology or general hospital records were also consulted.

In addition to age, sex, and pretreatment weight, information was collected on the following patient and treatment variables for both cases and controls. Patient variables were recorded for the time of the course of radiotherapy under evaluation. If, for example, a patient initially had local disease but subsequently relapsed with metastatic disease and only then received radiotherapy, that patient was considered to have metastatic disease for the purposes of this study.

Karnofsky performance status (KPS).²⁴KPS was contained within the radiotherapy chart as part of the routine pretreatment physical examination for most patients. In those few cases in which KPS was not explicitly recorded, the patient's detailed history and physical examination were used to make an estimate.

Tumor type.Patients with malignant disease were assigned to either a solid tumor or leukemia/lymphoma category. The latter included Hodgkin's disease patients.

Disease extent.The following categories were used: local disease, in which there were no known lymphatic or systemic metastases; regional disease, in which there were loco-regional lymph node metastases but no distant metastases; and metastatic disease, in which distant metastatic spread was confirmed (analyzed by a score of 0, 1, or 2).

Bone marrow involvement.Bone marrow study results were considered positive when aspiration or biopsy contained tumor cells or bone marrow metastasis was apparent on imaging studies (eg, magnetic resonance imaging [MRI]). All other cases were considered to be negative.

Bone metastasis.Patients were considered to have bone metastasis only when there was evidence of such on radiographs, computed tomography (CT), MRI, or bone scan.

Brain metastasis.Patients were included in this category if CT or MRI showed intracranial tumor.

Concomitant conditions affecting hemostatic function.Information regarding conditions such as idiopathic thrombocytopenia purpura, hypersplenism, or hereditary clotting disorders was sought (analyzed as present/absent).

Thromboembolic disease.Past medical histories of arterial or venous thrombosis or embolism were recorded (analyzed as present/absent).

Serum bilirubin and serum creatinine.Levels of these parameters, whether normal or raised at the start of treatment, were recorded for all patients who had biochemical screening within 1 week of commencing radiotherapy (not analyzed due to a lack of data).

Concomitant medication with potential effects on marrow or clotting function.A list of medications commonly encountered in clinical practice that could affect marrow or clotting function was compared with the list of medicines that the patient was taking at the start of radiotherapy (analyzed as present/absent).

Percentage of active bone marrow previously irradiated.Estimates of this parameter were made by studying treatment records of previous radiation fields and using published data on the distribution of active marrow in adults.25-27 We used a table that provided estimates of the amount of marrow in the common types of radiation fields used at Stanford. Because these estimates were necessarily imprecise, patients were allocated to one of the following categories: cumulative proportion of active bone marrow irradiated less than 20%, 20% to 39%, 40% to 59%, 60% to 79%, or ≥80% (scored as 1 through 5, respectively, for analysis). Allocation of patients to quintiles also facilitated statistical analysis. When a patient had wide-field radiation followed by a small volume boost, only the percentage of marrow in the wide-field treatment was recorded.

Percentage of active bone marrow previously irradiated and percentage of marrow to be irradiated during the radiation therapy course under investigation.This parameter was estimated and analyzed as described above. In general, previous and current radiation fields did not overlap significantly and percentages could be added directly. All charts were reviewed to check for overlapping fields. Marrow in the region of overlap was counted only once in those few cases in which overlap was found.

Prior treatment with chemotherapy with high myelotoxicity potential.This category refers to any severely myelotoxic cytotoxic chemotherapy administered more than 28 days before radiotherapy and was recorded simply as yes or no. In almost all of these cases, the nadir of platelet and neutrophil counts had passed before radiotherapy commenced. A classification of commonly used chemotherapeutic agents into categories of either high or low to moderate myelotoxicity is given in Table 1. Table 1 was based, in part, on reviews of myelosuppression associated with individual drugs28,29 and on our own experience. Almost all patients who received chemotherapy received treatment regimens containing at least one severely myelotoxic drug.

Sequential treatment with chemotherapy with high myelotoxicity potential.This category refers to any severely myelotoxic cytotoxic chemotherapy administered between 28 days and 1 day before the start of radiotherapy and was recorded as yes or no.

Concurrent treatment with chemotherapy with high myelotoxicity potential.This category includes patients receiving chemotherapy within 1 day of starting radiotherapy or at any time during the course of radiation treatment and was recorded as yes or no. The number of cycles of chemotherapy administered during radiotherapy was also recorded but not analyzed.

Number of prior chemotherapy regimens with high myelotoxicity potential.The number and type of different chemotherapy regimens (eg, MOPP and CMF ) administered more than 28 days before radiotherapy were recorded. The number of cycles of each regimen was often difficult to obtain for patients who received chemotherapy outside Stanford and was therefore not collected for every patient. The interval between the last dose of chemotherapy and the start of radiotherapy was recorded but not analyzed separately.

Data were collected on standardized case report forms and all forms were reviewed by a physician before data entry into a database created on a mainframe computer. Summary statistics were provided from that database. Additional analysis was performed using SAS software (SAS Institute Inc, Cary, NC).

The primary analysis for both univariate and multivariate analyses was logistic regression.30 Initially parameters were analyzed as univariate predictors. Those that were significant at the P = .1 level were considered for inclusion in a multivariate model. Criteria for inclusion in the final model was P < .05. In cases in which information on a variable was largely unavailable or if few patients exhibited a characteristic, that variable was excluded from the analysis.

There were 45 patients with malignant disease who experienced treatment interruption associated with thrombocytopenia (thrombocytopenia cases). These were compared with 138 randomly selected controls who had also received radiotherapy for malignant disease. Although the controls experienced no significant treatment delays, 2 control patients had platelet counts during radiotherapy that were low enough to require platelet transfusions and 2 other control patients required platelet transfusions within 2 months of completing radiation therapy. Baseline characteristics of both cases and controls are shown in Table 2. Treatment factors for thrombocytopenia cases and controls are shown Table 3. Univariate analysis was used to search for significant differences between cases and controls (noted for P ≤ .1, P ≤ .05, P ≤ .01, and P < .001 in Tables 2 and 3).

Those variables that were significant at the P = .1 level in the univariate model were entered into a multivariate model and results are shown in Table 4. The percentage of active marrow irradiated (P < .001) and the concurrent administration of severely myelotoxic chemotherapy (P < .001 ) were the factors most highly predictive for the interruption of radiotherapy by thrombocytopenia. The odds ratio (OR) increased by a factor of 45.5 if the patient had concurrent severely myelotoxic chemotherapy. The OR increased by 4.1-fold for each 20% increment of the percentage of marrow irradiated. Brain metastasis was also a significant factor (OR, 7.3; P = .01).

The multivariate analysis model (Table 4) was used to create a regression score for each patient. The total regression score for each individual patient was calculated by adding scores from each of the following parameters. (1) The total cumulative percentage of bone marrow irradiated was given a score of 1.4 for each increment of 20% (eg, a patient with 30% of marrow irradiated received a score of 2.8). (2) The presence of brain metastasis was given a score of 2.0. (3) Concurrent severely myelotoxic chemotherapy was given a score of 3.8.

Patients were arbitrarily categorized into low-risk, moderate-risk, and high-risk groups based on scores of less than 1.5, 1.5 to 4.5, or greater than 4.5. The distribution of patients between these different categories is shown in Table 5. According to this criterion, only 4 (2%) of the control patients would have been considered at high risk for thrombocytopenia versus 22 (49%) of those patients who were actually cases. Only 3 (7%) of the cases were categorized as low risk, compared with 74 (55%) of the controls.

Based on these scores, the administration of concurrent severely myelotoxic chemotherapy was found to cause a risk of treatment interruption for thrombocytopenia, which is similar to the risk from irradiation of 40% to 60% of active marrow or the risk in a patient with brain metastasis undergoing radiation of 20% to 40% of active marrow.

Analysis of secondary outcome measures showed that 12 (26.7%) of the patients with treatment interruption due to thrombocytopenia had at least one platelet count performed that was less than 25 × 10⁹/L and that 44 (98%) had a treatment interruption of ≥3 days. Only 1 (<1%) of the control patients (who had no significant treatment interruptions) had a platelet count of less than 25 × 10⁹/L. Reductions in administered radiation dose by greater than 10% of the planned dose occurred in 23 (51%) of the thrombocytopenia cases, compared with 15 (11%) of the controls. A radiation dose reduction of greater than 10% occurred in 11 of 26 patients (42%) receiving treatment with curative intent who had treatment interruptions associated with thrombocytopenia as compared with 7 (7%) of the controls. Twenty-two (49%) of thrombocytopenia cases required a reduction in daily fraction size. Platelet transfusions were administered to 7 (16%) cases (all of whom also had neutropenia) and 2 (1.4%) controls during the radiation therapy course. Two of the above cases and 9 other cases also required platelet transfusions within 2 months of finishing radiation therapy, compared with only 1 of the above and 2 other control patients that required platelets within 2 months of the completion of therapy. Platelet transfusions within 2 months of completing radiation therapy occurred most frequently in patients that had had both thrombocytopenia and neutropenia during treatment (10 patients) compared with thrombocytopenia alone (1 patient) or neutropenia alone (no patients).

A total of 63 patients with malignant disease had treatment interruptions associated with grade 3 or 4 neutropenia (neutropenia cases). These were compared with the 138 randomly selected controls as described above. Thirty-seven (59%) of the 63 neutropenia cases had treatment interruptions that were also associated with thrombocytopenia. Baseline characteristics of neutropenia cases and those patients that also had thrombocytopenia are shown in Table 2. Treatment factors for these cases are shown in Table 3. Baseline characteristics and treatment factors for patients who had treatment interruptions due to neutropenia as well as both neutropenia and thrombocytopenia have been compared with controls and P values are shown in Tables 2 and 3 for those comparisons that were different with P ≤ .1.

Multivariate analysis (Table 4) confirmed that the most important variables for the occurrence of a treatment interruption associated with neutropenia or with both thrombocytopenia and neutropenia were the concurrent administration of severely myelotoxic chemotherapy (ORs, 42.1 and 48.6, respectively; P < .001) and the total cumulative percentage of marrow irradiated (OR increased by 3.3 and 3.9, respectively, for each 20% of bone marrow irradiated; P < .001). As for the thrombocytopenia cases, a regression score was calculated for each case from these two groups using the results of the multivariate analysis. Cases and controls were allocated to arbitrarily defined groups categorized as high, intermediate, or low risk on the basis of these regression scores. Results of this analysis are shown in Table 5.

Fifty-five patients had a treatment interruption of ≥3 days associated with neutropenia. Twenty-two neutropenia cases (35%) and 11 neutropenia cases (24%) treated with curative intent had reductions in the total administered dose of greater than 10% compared with the planned dose, and 25 patients required a reduction in daily fraction size. Thirty-seven patients with both thrombocytopenia and neutropenia had treatment interruptions of ≥3 days in duration. Twenty of these patients (54%) and 10 patients (46%) in this group that were treated with curative intent had reductions in the total administered dose of greater than 10% compared with the planned dose, and 20 patients required a reduction in the daily fraction size.

A pretreatment assessment of the risk that a particular patient will develop significant neutropenia or thrombocytopenia during radiotherapy is not simply an academic exercise. We now have the ability to intervene, treating established neutropenia with G-CSF²¹ or preventing it with the use of prophylactic G-CSF.20 In addition, since the discovery of the Mpl ligand,31-33 there is now the prospect of effective treatment or prevention of chemotherapy and radiotherapy-induced thrombocytopenia with a megakaryocyte colony-stimulating factor. To effectively target the use of such an agent in radiotherapy patients, an estimate of the risk of significant myelosuppression for each individual would be invaluable. In this study, we hoped to provide numerical data to quantify the risks involved and to identify factors or combinations of factors that were most highly predictive for the occurrence of significant neutropenia and thrombocytopenia during radiotherapy.

Radiotherapy and chemotherapy cause different types of bone marrow injury. Severe marrow injury occurs only within the irradiated volume during radiotherapy, whereas unirradiated marrow remains unaffected. Temporary functional ablation of marrow within the radiation field may occur even with relatively low radiation doses. If the irradiated volume is small, there will be no significant effect on peripheral blood counts. Repopulation of the marrow cavity and restoration of active hematopoiesis may take several years after exposure to moderate radiation doses.34 With higher radiation doses, irreversible damage to the medullary stroma may be associated with permanent marrow aplasia. By contrast, myelotoxic chemotherapy causes injury to all of the bone marrow, but recovery of peripheral blood counts usually occurs within weeks. With many agents there appears to be minimal long-term damage to hematopoietic tissues, but some agents35,36 can cause prolonged depletion of marrow stem cells, compromising recovery from future bone marrow injury by reducing bone marrow reserve. The interaction between these types of marrow injury is complex. In this study, we have sought to determine the relative importance of chemotherapy and radiation-induced injury in patients who experienced significant hematopoietic toxicity during radiotherapy.

Radiotherapy departments treat patients with an extremely wide range of malignancies. Each neoplasm has a different tendency to involve distant sites. Treatment strategies range from small-field radiotherapy alone to intensive combinations of wide-field radiation therapy and combination chemotherapy. Any retrospective investigation attempting to categorize the widely varied treatments administered to such a diverse population of patients requires a great deal of simplification. In this study, radiation treatments were categorized by increasing volumes of marrow irradiated in increments of 20%. Radiation dose was not considered because bone marrow is exquisitely radiation sensitive and in all cases the bone marrow dose was sufficiently high to cause severe bone marrow injury within the irradiated volume during the treatment course. Chemotherapy was considered in terms of the presence of one or more severely myelotoxic drugs in a particular regimen and by the number of different regimens used, rather than by the number of cycles of each regimen. Because this information was readily obtainable, we could include the maximum number of patients in the study and thereby arrive at some general conclusions. Information on the number of cycles of each chemotherapy regimen, the dose intensity achieved for each patient, and the incidence of myelotoxicity during chemotherapy might have been useful in refining our model. However, in the setting of a retrospective study, this information could not readily be collected for the majority of patients.

Our results indicate that the two strongest risk factors for treatment interruptions during radiotherapy associated with neutropenia or thrombocytopenia (or both) are the concurrent administration of chemotherapy containing one or more severely myelotoxic drugs and increasing volumes of active bone marrow in the radiation field. Chemotherapeutic drugs commonly concurrently administered during radiation therapy included cisplatin, methotrexate, 5-fluorouracil, vincristine, cyclophosphamide, adriamycin, and VP-16. In patients receiving small-field radiation therapy, the chemotherapy is probably primarily responsible for the myelosuppression and the relative contribution of the radiation therapy to the myelosuppression increases with increasing amounts of bone marrow in the field. The fact that the presence of brain metastases was a stronger risk factor for thrombocytopenia than the presence of either bone or bone marrow metastases seemed initially to be a surprising result. However, this observation may reflect in part the fact that brain metastases often occur relatively late in the course of malignant disease in patients who are also likely to have received prior chemotherapy and/or radiation therapy. However, brain metastases was still a predictor even after adjusting for the percent of bone marrow irradiated in the multivariate analysis. The increased risk of treatment interruption associated with neutropenia in leukemia/lymphoma patients compared with solid tumor patients is not surprising given the frequent use of intense chemotherapy regimens and/or wide-field radiotherapy in these patients. As would be expected, radiation treatment interruption due to myelosuppression was more common in patients with advanced disease compared with local disease and increased in incidence with increasing numbers of prior chemotherapy regimens. Other patient characteristics, such as Karnofsky status, sex, weight, age, and administration of nonchemotherapy drugs with a potential for marrow toxicity had no discernible impact on the risk of a treatment interruption occurring during radiotherapy.

On the basis of the regression analysis it was possible to identify three categories of patients who had different probabilities of having an interruption in radiotherapy due to thrombocytopenia, neutropenia, and both thrombocytopenia and neutropenia. For these groups, the high-risk group contained a large proportion of cases (49%, 59%, and 68%, respectively) and a small proportion of controls (2%, 7%, and 7%, respectively). Similarly, the low-risk groups contained a small proportion of cases (7% of thrombocytopenia cases, 13% of neutropenia cases, and 8% of cases with both thrombocytopenia and neutropenia), but a large proportion of controls (55% of thrombocytopenia controls and 60% of both neutropenia and thrombocytopenia and neutropenia controls). The large intermediate groups contained similar proportions of cases and controls for all three groups, with the largest intermediate group for both cases and controls occurring in the thrombocytopenia group. Using these criteria, it may be possible to identify approximately half of the radiotherapy patients who are likely to develop problems with treatment interruption due to thrombocytopenia or neutropenia before treatment commences.

We plan to use this model in a prospective study in which radiotherapy patients will be assigned prospectively to high-risk, intermediate-risk, or low-risk groups for thrombocytopenia according to criteria derived from the multiple linear regression analysis. Such a study will allow us to test the hypothesis that the data presented here can be used to prospectively estimate the risk that a particular patient will develop significant thrombocytopenia during radiotherapy. Patients who fulfill the criteria for the high-risk category would be potential candidates for future clinical trials involving a platelet growth factor.

The authors thank Drs W. Sheridan and J.F. LaBrecque for their thoughtful review of the manuscript, J. Ramback and J. Tod for assistance with the database and data entry, E. Chen for assistance with data collection, Dr J. Poen for assisting with the review of the data collection forms, and S. Clarke for secretarial services in preparing this manuscript.