Incidence and clinical relevance of abnormal complete blood counts in long-term survivors of childhood cancer

Authors

  • Zsofia B. Long M.D.,

    1. Department of Family Practice & Community Medicine, University of Texas Southwestern Medical Center, Dallas, Texas
    2. After Cancer Experience Program, Children's Medical Center, Dallas, Texas
    Search for more papers by this author
  • Kevin C. Oeffinger M.D.,

    Corresponding author
    1. Department of Family Practice & Community Medicine, University of Texas Southwestern Medical Center, Dallas, Texas
    2. After Cancer Experience Program, Children's Medical Center, Dallas, Texas
    3. Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas, Texas
    4. Center for Cancer and Blood Disorders, Children's Medical Center, Dallas, Texas
    • Department of Pediatrics, Memorial Sloan-Kettering Cancer Center, 1275 York Ave., New York, NY 10021
    Search for more papers by this author
    • Fax: (212) 717-3239

  • Sandra L. Brooks R.D.,

    1. Department of Family Practice & Community Medicine, University of Texas Southwestern Medical Center, Dallas, Texas
    2. After Cancer Experience Program, Children's Medical Center, Dallas, Texas
    Search for more papers by this author
  • Lori Fischbach M.P.H, Ph.D.,

    1. University of Texas School of Public Health, Houston, Texas
    Search for more papers by this author
  • T. Robert Harris Ph.D.,

    1. University of Texas School of Public Health, Houston, Texas
    Search for more papers by this author
  • Debra A. Eshelman M.S.N.,

    1. After Cancer Experience Program, Children's Medical Center, Dallas, Texas
    2. Center for Cancer and Blood Disorders, Children's Medical Center, Dallas, Texas
    Search for more papers by this author
  • Gail E. Tomlinson M.D., Ph.D.,

    1. After Cancer Experience Program, Children's Medical Center, Dallas, Texas
    2. Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas, Texas
    3. Center for Cancer and Blood Disorders, Children's Medical Center, Dallas, Texas
    Search for more papers by this author
  • George R. Buchanan M.D.

    1. After Cancer Experience Program, Children's Medical Center, Dallas, Texas
    2. Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas, Texas
    3. Center for Cancer and Blood Disorders, Children's Medical Center, Dallas, Texas
    Search for more papers by this author

Abstract

BACKGROUND

The purpose of the study was to determine the incidence and clinical significance of abnormal complete blood counts (CBCs) obtained during follow-up of childhood cancer survivors.

METHODS

A retrospective cohort study was conducted on 193 survivors, diagnosed between 1970–1986, who were followed in our center's After Cancer Experience Program and are participants in the Childhood Cancer Survivor Study. Of these patients, 49% were female and 25% were racial/ethnic minorities. The primary outcome was determination of the cumulative percentage of patients having an abnormal CBC by 2 or 3 standard deviations (SDs). Four components of the CBC were examined and employed to define an abnormal CBC: low white blood cell count (WBC), high mean corpuscular volume (MCV), low platelet count, and low hemoglobin concentration. Association of treatment exposures to abnormal values was assessed with a multilevel logistic model.

RESULTS

There were 1297 patient visits during 1401 person-years of follow-up. The mean number of visits per survivor was 6.7 (SD 4.2). The cumulative percentage of subjects with at least one abnormal CBC was 70%. The cumulative percent of subjects with a value abnormal by 2 SD was WBC = 23%, MCV = 37%, platelets = 9%, hemoglobin = 49%. For values abnormal by 3 SD, the frequencies were WBC = 3%, MCV = 20%, platelets = 1%, hemoglobin = 27%. None of the patients developed myelodysplastic syndrome or a secondary leukemia during the follow-up period. Exposure to epipodophyllotoxins was associated with an increased risk of having abnormally high MCV values.

CONCLUSIONS

Mildly abnormal CBC values are common in survivors of childhood cancer. Abnormal values are often of questionable significance but seem to persist over time. Epipodophyllotoxin therapy was found to be associated with increased frequency of high MCV levels. Cancer 2006. © 2006 American Cancer Society.

With improved treatment of childhood cancer, a new and growing generation of adolescents and young adults is in need of surveillance for late effects caused by their cancer therapy. Virtually all organ systems can be affected by cytotoxic therapy.1, 2 Childhood cancer survivors are frequently followed in long-term follow-up programs and are periodically screened for late effects of therapy.1, 2 Testing is based on treatment exposures.

The recently published “Children's Oncology Group (COG) Long-Term Follow-Up Guidelines for Survivors of Childhood, Adolescent, and Young Adult Cancers” recommended an annual complete blood count (CBC) with platelet and differential white blood cell (WBC) count for 10 years after diagnosis for patients treated with epipodophyllotoxins and anthracyclines, and for 15 years for patients exposed to alkylating agents.3 The stated rationale for testing is to screen for secondary hematologic alterations, including treatment-related acute myeloid leukemia (t-AML) or myelodysplastic syndrome (MDS).4–8 The optimal duration of such surveillance is not known. No prior studies on childhood cancer survivors specifically examined CBCs for alterations that might suggest evolving marrow injury from cumulative effects of chemotherapy and radiation. Therefore, this study was conducted retrospectively to determine the incidence and clinical significance of abnormal CBCs obtained during the extended follow-up of a cohort of childhood cancer survivors. Because of the small sample size, it was not the intent of this study to determine the incidence of t-AML, but rather to focus on mild to moderate changes in the various parameters of the CBC.

MATERIALS AND METHODS

Between 1970–1986, 483 patients were diagnosed with cancer and were subsequently long-term survivors at Children's Medical Center Dallas, with the following eligibility criteria: 1) diagnosis with one of eight cancer groups (leukemia, Hodgkin disease, non-Hodgkin lymphoma, central nervous system tumor, soft tissue sarcoma, Wilms tumor, neuroblastoma, bone cancer); 2) survival of 5 years or more from date of diagnosis; and 3) age less than 21 at diagnosis. Of this population, 299 (62%) were enrolled in the Childhood Cancer Survivor Study (CCSS). Of the remaining 184 patients, 128 (27%) were lost to follow-up, 52 (11%) refused participation, and 4 (1%) died before enrollment, one of whom developed t-AML. Of the 299 patients enrolled in CCSS, 193 (65%) were seen at least once in a clinical setting in the After Cancer Experience (ACE) Program at Children's Medical Center Dallas or at the University of Texas Southwestern Medical Center at Dallas, TX.9 In this retrospective cohort study, all available laboratory data and clinical information for these 193 participants were reviewed. CBC values that were obtained at least 2 years after completion of cancer treatment were recorded. Patients were generally seen every 1–2 years for routine follow-up. Every available CBC up to August 2002 was included in the analysis.

The demographic information and treatment exposures for the subjects are summarized in Table 1. Anthracyclines included daunorubicin and doxorubicin. Epipodophyllotoxins included etoposide and teniposide. Alkylating agents included BCNU, busulfan, carboplatinum, cis-platinum, cytoxan, ifosfamide, melphalan, nitrogen mustard, procarbazine, and thiotepa. Based on equivalent cumulative doses, tertiles were created for low, medium, and high doses for each composite group. The epipodophyllotoxin tertile doses were less than 983, 983–4108, and more than 4108 mg/m2. Treatment exposures of the leukemia survivors (N = 102) included: anthracycline, 47%; alkylating agent, 58%; epipodophyllotoxin, 26%; and cranial radiotherapy, 30%. Before the recognition of the association of etoposide and t-AML, etoposide was included in the DFW-1 protocol for standard risk acute lymphoblastic leukemia (ALL) therapy at Children's Medical Center Dallas.6

Table 1. Demographics of 193 Childhood Cancer Survivors Whose Blood Counts Were Analyzed in This Study
VariablesN%
  1. NH: non-Hispanic; SD: standard deviation; yr: year.

  2. Percentages are based on the total with available data for each variable.

Gender  
 Female9549.2
 Male9850.8
Race  
 White, NH14575.1
 African American, NH2613.5
 Hispanic199.8
 Other31.6
Cancer diagnosis  
 Leukemia10252.9
 Central nervous system tumor4 
 Hodgkin disease115.7
 Non-Hodgkin lymphoma84.2
 Wilms tumor3618.7
 Neuroblastoma199.8
 Soft tissue sarcoma105.2
 Bone tumor31.6
Cancer treatment  
 Alkylating agent10062.5
 Epipodophyllotoxin2616.3
 Anthracycline8050.0
 Radiation, any site8248.8
  Cranial radiotherapy2814.5
  Chest radiotherapy178.8
  Pelvic radiotherapy126.2
  Total body irradiation10.5
Age at cancer diagnosis, y 
 Mean (SD)4.8 (3.9)
 Median3.7
 Range0.0–16.4 
Age at last visit, y 
 Mean (SD)19.5 (5.1)
 Median19.1
 Range7.8–34.5
Interval from diagnosis to last visit, y 
 Mean (SD)14.7 (4.5)
 Median14.5
 Range4.0–28.2

For each blood count the normal range provided on the laboratory report was recorded. For missing normal range values, age- and gender-specific normal range values were substituted. By definition, the normal range encompassed up to plus or minus 2 standard deviations (SDs) from the average for that value in a normal population of the same age and gender. For the purposes of this study, an abnormal blood count was defined as follows: greater than ± 2 SD from the mean value, which corresponds to being outside the normal range for a given laboratory value, and greater than ± 3 SD from the mean, a value being slightly more abnormal. The incidence of abnormal laboratory values was calculated for four blood count measures felt to indicate bone marrow injury, including low WBC count, low hemoglobin concentration, low platelet count, and high erythrocyte mean corpuscular volume (MCV).

Scatterplots were created with the y-axis representing the number of SDs from the mean for each of the blood count measures of interest and the x-axis the time from diagnosis to the visit (Fig. 1a–d). Univariate analysis was performed to assess the association of abnormal CBCs with different cancer treatments, gender, age at diagnosis, and race. A multilevel logistic model was constructed, which incorporated the repeated measures within individuals, to further investigate the relation of cancer treatments, length of time from diagnosis, and total number of visits to the likelihood of obtaining a high MCV, or low hemoglobin, WBC, and platelet count.10 Total chemotherapy doses were divided into tertiles to investigate whether increasing doses of chemotherapy were associated with increased likelihood of having an abnormal laboratory value. The lowest tertile was compared with not having received that particular agent. The multilevel logistic model was used to account for the differences between the numbers of visits among individuals, and therefore patients who were seen more frequently were not overrepresented in the analysis. Additionally, time from diagnosis is a within-subjects variable, allowing response to change over time for each subject. The model was adjusted for age at diagnosis, race, and gender.

Figure 1.

Standard deviation of (a) MCV, (b) hemoglobin, (c) WBC, and (d) platelet values by years from diagnosis for 1297 patient visits. The two horizontal lines represent years from diagnosis, and the area between represents the normal range. MCV, mean corpuscular volume; SD, standard deviation; WBC, white blood cell count.

Data were analyzed with SAS v9.0 (SAS Institute, Cary, NC) and MLWIN software. This study was approved by University of Texas Southwestern Medical Center Institutional Review Board.

RESULTS

The mean number of visits during which at least one component of the CBC was performed was 6.7 (SD = 4.2; range, 1–19) with a mean interval from cancer diagnosis to the last follow-up visit of 14.7 years (SD = 4.5; range, 4.0–28.2). The total follow-up period was 1401 person-years (PY). The mean age at last follow-up visit was 19.5 years (SD = 5.1; range, 7.8–34.5). None of the patients underwent autologous stem cell rescue. None of the subjects developed t-MDS/t-AML or aplastic anemia. Two patients died from nonhematologic causes.

Figure 1 represents scatterplots of SD of MCV (Fig. 1a), hemoglobin (Fig. 1b), WBC (Fig. 1c), and platelet (Fig. 1d) values by years from diagnosis for 1297 patient visits.

Table 2 summarizes the prevalence of abnormal laboratory results (greater than ± 2 SD and greater than ± 3 SD) per person. The expected percentage of at least one abnormal blood count during follow-up was calculated based on the average number of times each blood count was drawn and the properties of the normal distribution. As some blood counts were drawn more often than others, the expected percentage of abnormal varies. To account for multiple blood draws, and the fact that we were interested in abnormal values in only one direction (high or low), the equation used to determine expected percentage of abnormals >2 SD = 1 − (0.975ˆmean # blood draws), and is 1 − (0.995ˆmean # blood draws) for values greater than 3 SD. The prevalence of having one or more abnormal value of interest > ± 2 SD was 69.9%; 30.1% of the cohort never had an abnormal value > ± 2 SD. The prevalence of having one or more abnormal values of interest > ± 3 SD was 40.4%; 59.6% of the cohort never had an abnormal value > ± 3 SD. The prevalence of abnormal laboratory values > ± 2 SD for each of the blood counts per 1000 PY was 31.4 for low WBC, 45.0 for high MCV, 12.8 for low platelet counts, and 68.5 for low hemoglobin. The prevalence of abnormal laboratory values > ± 3 SD per 1000 PY was 4.3 for low WBC, 24.3 for high MCV, 1.4 for low platelet counts, and 37.1 for low hemoglobin. The mean number of visits (7.8 visits) was higher for those who had at least one abnormal component of the CBC in comparison with those who always had a normal CBC (4.3 visits; P < 0.0001). Similarly, the mean number of years of follow-up were longer for those with an abnormal CBC (8.3 yrs) in comparison with those who always had a normal CBC (4.7 yrs; P < 0.0001). A correlation matrix with the four CBC outcomes by visit was performed. There were weak correlations between high MCV and low platelets (0.15, P < 0.001) and between low WBC and low platelets (0.19, P<0.001).

Table 2. Frequency of Abnormal CBC values (> ± 2 SD and > ± 3 SD) for 193 Childhood Cancer Survivors
Blood count measurePercentage of patients with abnormal values (> ± 2 SD)Expected frequency of abnormal values (> ± 2 SD) in a normal population*Percentage of patients with abnormal values (> ± 3 SD)Expected frequency of abnormal values (> ± 3 SD) in a normal population*
  • CBC: complete blood count; MCV: mean corpuscular volume; SD: standard deviation; WBC: white blood cell count.

  • *

    Calculated based on the average number of times each blood count was measured per person (WBC = 6.7, platelet = 6.6, hemoglobin = 6.7, MCV = 4.8).

Low WBC22.814.5%3.13.3%
High MCV37.311.0%20.12.4%
Low platelet9.314.4%1.03.2%
Low hemoglobin49.214.6%26.93.3%

Estimates of relative risk (RR) and 95% confidence intervals (CI) for the multilevel analysis are provided in Table 3. All independent variables in the table were included in the model. Participants who received epipodophyllotoxins were more likely to have an abnormally high MCV during follow-up. Each dose range increase in epipodophyllotoxin was associated with a 60% increase in risk for high MCV (95% CI = 1.1–2.3). Exposure to an epipodophyllotoxin (or etoposide) was not associated with higher doses of an alkylating agent or craniospinal radiotherapy. In a multilevel logistic regression model including only participants with a history of epipodophyllotoxin exposure, and adjusted for sex, race, and number of visits, a change in the likelihood of a high MCV over time was not found.

Table 3. Multilevel Logistic Model with Relative Risk estimates (RR) and 95% Confidence Intervals (95% CI) of the Likelihood of Having Abnormal (> ± 2 SD) Blood Values for 193 Patients Having 1297 Visits*
Independent VariablesHigh MCVLow HemoglobinLow WBCLow platelet count
RR95% CIPRR95% CIPRR95% CIPRR95% CIP
  • CI: confidence interval; MCV: mean corpuscular volume; NA: not applicable; RR: relative risk; WBC: white blood cell count.

  • *

    Adjusted for age at diagnosis, race, and gender.

  • **

    Chemotherapy dosing was categorized as none, lowest tertile, middle tertile, and highest tertile.

Anthracycline dose**0.70.5–1.0.071.31.0–1.8.100.80.5–1.2.270.90.5–1.8.86
Epipodophyllotoxin dose**1.61.1–2.3.020.90.6–1.3.501.00.6–1.7.910.90.4–2.1.81
Alkylating agent dose**1.51.0–2.2.081.00.7–1.4.861.20.8–1.9.431.90.9–4.1.09
Radiation, yes vs. no1.90.9–4.1.080.40.2–0.8.0060.70.3–1.5.360.20.0–0.9.04
Time since diagnosis, y0.90.9–1.0.040.90.8–0.9< .00011.00.9–1.1.811.11.0–1.2.16
Total number of visits1.00.9–1.11.01.11.0–1.2.011.01.0–1.1.391.10.9–1.3.30

Anthracyclines and alkylating agents were not significantly associated with any of the abnormal CBC values. Participants who had radiation treatment were less likely to have low hemoglobin concentrations (RR = 0.4; 95% CI = 0.2–0.8) and low platelet counts (RR = 0.2; 95% CI = 0.0–0.9). Increasing time from diagnosis to each follow-up visit was associated with a lower likelihood of having a high MCV value (RR = 0.9; 95% CI = 0.9–1.0). Increasing time from diagnosis to follow-up was also associated with a lower likelihood of having a low hemoglobin concentration (RR = 0.9; 95% CI = 0.9–1.0). Increased number of follow-up visits was related to increased risk for low hemoglobin (RR = 1.1; 95% CI = 1.0–1.2). Black race was associated with increased risk of low hemoglobin (RR = 7.2; 95% CI = 2.8–18.2) and low WBC count (RR = 4.2; 95% CI = 1.4–12.9). Of note, no difference was found in males and females in the frequency of abnormal hemoglobin concentration. Seventeen (9%) survivors had a low hemoglobin value and a low MCV at the same visit.

Because leukemia survivors represented a large proportion of the participants (53%), the analysis was performed with and without them. The findings were not significantly different for leukemia survivors versus the rest of the participants. Thus, all survivors were grouped for the final analysis. Of the 11 subjects with a markedly abnormal laboratory value (WBC ≥30,000/mm3, WBC ≤2,500/mm3, MCV ≥110 μm3, hemoglobin ≤9 g/dL, or platelets ≤100,000/m3), nine had comorbid health conditions including the following: current Epstein–Barr viral infection, iron deficiency, Down syndrome with cyanotic heart disease, end-stage renal disease, portal hypertension secondary to chronic hepatitis B, and seizure disorder requiring divalproex sodium therapy. As with the leukemia subjects, the prevalence rates and multilevel logistic models were performed with and without these nine subjects. Because there were no significant differences in the findings, they were included.

DISCUSSION

This study is the first to report blood count results obtained over an extended period of follow-up in childhood cancer survivors. Among 193 survivors followed for an average of 15 years after cancer diagnosis, 70% had at least one abnormality of the four CBC measures examined. Thus, abnormal CBCs were common in this population, possibly reflecting bone marrow injury. Assuming a normal distribution and accounting for the average number of times each blood count was measured per person, the expected frequency of an abnormally high or low value (i.e., > ± 2 SD from the mean) is between 11 and 15%. In this cohort the actual frequency of low WBC was 23%, high MCV was 37%, and low hemoglobin was 49% compared with the expected values.

Only a few studies have investigated peripheral blood count abnormalities of otherwise healthy survivors during follow-up. One analysis of 73 survivors of childhood Ewing sarcoma reported a higher incidence of t-AML in those treated with higher dose chemotherapy (including large doses of alkylating agents, topoisomerase II inhibitors, and granulocyte colony-stimulating factor [G-CSF]) in comparison with those treated with lower dose chemotherapy.11 Patients receiving larger doses of chemotherapy had significantly greater MCV values and lower mean platelet counts for the entire length of the 40-month observation period. A study involving 1939 Hodgkin disease survivors highlights a similar correlation between decreased platelet counts in the course of the follow-up period and an increased risk for t-AML.12

Injury to hematopoietic stem cells is thought to be in part caused by telomere shortening as a consequence of chemotherapy, with repeated cycles of hematopoietic regeneration.13 Occult bone marrow dysfunction may occur in the face of a normal complete blood count. For instance, in a study involving six children who developed t-AML after treatment for neuroblastoma, the diagnosis was found incidentally in routine bone marrow testing in four of the patients.14 Other studies also describe in vitro evidence of treatment-induced bone marrow damage in asymptomatic patients.15–17

Because of the small sample size, it was not the intent of this study to determine the incidence of t-AML, but rather to focus on mild to moderate changes in the various parameters of the CBC. In long-term follow-up of childhood cancer survivors, blood count abnormalities were a common finding. Whether or not this is indicative of clinically significant marrow dysfunction that will worsen with time is not known. Further, it is not known if this population with previous myelotoxic therapy will face an increased risk of MDS or other marrow abnormalities, such as anemia of chronic disease, as they enter their later decades of life. There was only a slight decrease in the frequency of abnormal MCVs and hemoglobin values over time. Therefore, it is possible that bone marrow damage after cancer treatment is permanent.

In this study, increasing epipodophyllotoxin doses were found to correlate with increased likelihood of obtaining high MCV values. For each dose range increase in an epipodophyllotoxin, there was a 60% increase in the likelihood of having an abnormally high MCV value. Notably, the likelihood of the high MCV did not appear to change with increasing interval from the cancer diagnosis. The population of participants who had an epipodophyllotoxin were predominantly standard risk ALL survivors who were not treated with high doses of an alkylating agent or other therapies that may affect marrow recovery. However, recognizing the limitation of this small biased sample, it is important to further study the long-term changes in marrow function after exposure to an epipodophyllotoxin.

Black race was associated with lower hemoglobin levels, likely due in part to the increased incidence of thalassemia trait in this group of patients, and also with an increased frequency of low WBC count. Both of these findings are consistent with known ethnic differences in hemoglobin concentrations and WBC counts in the normal population.18, 19 Because of the retrospective nature of this study, whether or not African Americans with a low MCV had been previously evaluated for thalassemia could not be determined. Interestingly, patients who received radiation therapy were less likely to have low hemoglobin concentrations and low platelet counts. This association may reflect less intensive chemotherapy in these patients.

Markedly abnormal CBC values were infrequent. Generally, an underlying condition was present that could explain the markedly abnormal blood count measure. For example, iron deficiency anemia or end-stage renal disease was found among patients with markedly low hemoglobin concentrations. Low platelet counts were found in patients with Down syndrome with cyanotic heart disease and in a patient with idiopathic thrombocytopenic purpura.

There are several limitations of this study. First, this was a retrospective chart review with a relatively small sample size. Second, the cohort was biased by selection of those individuals who are followed in the ACE program, possibly resulting in an overestimate of the frequency of abnormalities. Third, the percent of the population who were exposed to an epipodophyllotoxin may have been higher than other children with cancer treated in that era. As noted in Materials and Methods, etoposide was used for standard risk ALL therapy in the early to mid-1980s until its association with t-AML was recognized. Finally, potential underlying comorbid conditions such as infection or anemia due to unrelated causes could have confounded the findings.

In conclusion, mildly abnormal CBCs are common in childhood cancer survivors. Marked abnormalities were infrequent and were generally explained by an underlying condition. Abnormalities of the four CBC components measured seemed to persist over a prolonged portion of the follow-up period. Generally, abnormal CBCs obtained during long-term follow-up are not clinically significant, and do not appear to reflect preleukemic changes. This study provides evidence for the common observation of abnormal blood counts during long-term screening and may contribute to the optimization of follow-up guidelines.

Acknowledgements

The authors thank Smita Bhatia, M.D., M.P.H., and Naomi Winick, M.D., for invaluable comments in reviewing the article.

Ancillary