Platelet and hemoglobin count at diagnosis are associated with survival in African American and Caucasian patients with colorectal cancer

Abstract

Background

African Americans (AAs) compared to Caucasian Americans (CAs) with colorectal cancer (CRC) have lower stage-specific survival CRC patients often present with several hematopathologies (such as thrombocytosis, thrombocytopenia, anemia) at diagnosis, which is associated with poorer survival. However, whether these measures impact the racial disparity in survival is not known.

Methods

The study population was composed of 581 histologically confirmed CRCs at the Medical University of South Carolina (393 CA, 188 AA) diagnosed between 01/01/2000 and 06/30/2013. We used Cox oroportional hazards regression to estimate the association between thrombocytosis, thrombocytopenia, or anemia at diagnosis and risk of death by race. This analysis was adjusted for age, sex, stage and first-line treatment.

Results

In all patients combined, thrombocytosis, thrombocytopenia, and anemia (vs. the normal ranges) were associated with significantly higher risks of death. In the race-specific analyses, AAs (HR 2.51 [95% CI: 1.52–4.15]) vs. CAs (HR 1.15 [95% CI: 0.75–1.75]) with thrombocytosis compared to normal had a higher risk of death (p for difference=0.03).

Conclusions

Abnormal thrombocyte and hemoglobin levels at diagnosis were associated with poorer survival. AAs compared to CAs with elevated platelets at diagnosis had a higher risk of death. Our study is the first to examine the role of race, hematologic measures at diagnosis, and risk of death in colorectal cancer patients. These results suggest that the racial differences in the immune response may contribute to the racial disparity in survival.

Introduction

African Americans (AAs) compared to Caucasian Americans (CAs) with colorectal cancer (CRC) have poorer stage-specific survival. The disparity is particularly pronounced among younger patients^1–6 and continues even after adjustment for confounding factors, such as age, sex, stage, treatment, comorbid conditions, or socioeconomic status. The persistence of the racial difference in survival highlights the need to broaden the search for causes and develop better biomarkers for prediction. Circulating hemaiologic markers (e.g., platelets, hemoglobin) are routinely assessed at diagnosis, exhibit conserved differences by race,^7–9 and are strongly related to patient outcomes,^10–13 but how they may impact the disparity in survival is not known.

Platelets and hemogiooin regulate host defense against pathogens and wound healing.^14–18 In healthy populations, AAs vs. CAs are predisposed to higher circulating platelet counts,^7,19,20 enhanced platelet activation and aggregation capacity^7,21,22 and lower hemoglobin concentrations.^23–27 Not surprisingly, abnormal hematologic measures, denoted by thrombocytosis (platelet counts > 400 ×10/L) and iron-deficient anemia^25–27 are more common in AAs. Interestingly, anemia is frequently associated with thrombocytosis,^28,29 and, more rarely, thrombocytopenia (platelet counts < 150 x 10/L).^29–30 In the context of malignancy, thrombocytosis, thrombocytopenia, and anemia are associated with systemic inflammation, metastatic expansion and poorer outcomes.^{10–13,31–34}

To date, no studies have investigated the association of hematopathologies and survival by race. Therefore, using the Hollings Cancer Center (HCC) hospital-based registry data with linkages to the electronic medical records, we examined the relationship between anemia and abnormal platelet counts on the risk of death in CRC patients. We then examined whether the platelet abnormalities and anemia impact the differences in survival by race.

Materials and Methods

The Medical University of South Carolina (MUSC) Institutional Review Board approved all study activities. The Cancer Registry at Hollings Cancer Center (HCC) at MUSC was used to identify all CRC cases. The registry is part of a state-mandated data system that ascertains all incident cancer cases in South Carolina. The study population comprised a sample of histologically confirmed cases diagnosed between January 1,2000, and June 30, 2013.

We abstracted data on demographic characteristics, clinical and pathological variables at diagnosis, treatment received, and patient outcome from the HCC cancer registry. Independent variables obtained included demographic characteristics (age at diagnosis, sex, race). Tumor-related variables obtained from the regisiry included tumor grade (low-grade [i.e., welldifferentiated, moderately differentiated] or high-grade [i.e., poorly differentiated, undifferentiated]); histology (not otherwise specified [NOS], mucin containing cancers); anatomic location of the primary tumor (proximal colon, distal colon, rectum); TNM stage (I, II, III, IV); and all first-line therapies (chemotherapy, surgery, radiation), and smoking status (ever, never). Clinical comorhid conditions (obesity, hypercholesterolemia, hypertension) were abstracted from the MUSC Clinical Data Warehouse (CDW), using ICD codes from the date of the cancer diagnosis. Pretreatment platelet and hemoglobin laboratory data were obtained from the MUSC CDW; we computed the average level using all blood draws that were available within two weeks of diagnosis. The data from the HCC cancer registry and the CDW were linked through the use of an honest broker.

Statistical Analysis

Platelet levels were divided into three categories: low, normal, and high. Thrombocytosis was defined as a median platelet count at diagnosis of greater than 400 × 10⁹/L. Thrombocytopenia was defined as a platelet count of less than 150 × 10⁹/L. Normal-range patients exhibited counts between 150 and 399 × 10⁹/L and served as the reference category in the analyses.³⁵ To define low hemoglobin levels, or anemia, in our population, we used the sexspecific ranges established by the American Society of Hematology.³⁶ For women, anemia was defined as a median hemoglobin concentration of lower than 12.0 g/dL, and for men, the lower threshold was 13.5 g/dL. As a combined measure of abnormal hematologic variables, we created categorical variables with three levels: 0 - no thrombocytosis, no anemia, 1 - one abnormal variable (thrombocytosis or anemia); 2 - both anemia and thrombocytosis. We did the same using thrombocytopenia or anemia, or both compared to patients with neither.

Using chi-square tests, we compared platelet and hemoglobin counts at diagnosis. We used Kaplan-Meier methods to generate median survival time and corresponding 95% confidence intervals for hematologic categories by race. The primary endpoint was overall survival, defined as the time from diagnosis to death from any cause. In the univariate analysis, we used Cox proportional hazards regression (PHR) to model the hazard of death as a function of abnormal platelets (thrombocytosis, thrombocytopenia), anemia and the combined measures (model 1: anemia plus thrombocytosis; model 2: anemia plus thrombocytopenia) in all patients. For each regression analysis, we computed the hazard ratios (HRs) and their 95% confidence intervals (CIs). In the multivariable analysis, we adjusted these models for age, sex, race, stage, and first-line treatment. Treatment was coded as an indicator variable for each treatment type: surgery, chemotherapy, and radiation. We examined the potential for confounding by comorbid conditions (obesity, hypertension, hypercholesterolemia, current smoking) by adding these to the fully adjusted models, one variable at a time.

Next, we examined the association between race and abnormal hematologic variables and death using product interaction terms between race and each of the platelet and hemoglobin measures adjusting for age, sex, stage, and treatment. We assessed the statistical significance of the interaction terms using the likelihood ratio test. We tested the proportional-hazards assumption using visual inspection and the proportional hazards test based on the Schoenfeld residuals. All tests were two-sided, and the comparison-wise type I error rate was controlled at level 0.05. All data analyses were performed using Stata 15.1 (College Station, TX).

Results

A total of 581 patients were included in the analysis (393 CAs,188 AAs;. Univariate associations of demographic and clinical characteristics with race are shown in Table 1. On average, AAs were younger at diagnosis (62 years vs. 64 years, p=0.05), and had more colon tumors vs. rectal (84% vs. 72% p=0.03). AAs, compared to CAs, had a higher prevalence of anemia (84% vs. 72%, p=0.001). We did not observe significant differences in thrombocytosis (16% vs. 12%), thrombocytopenia (7% vs. 6%), stage, or receipt of treatment by race (Table 1).

Table 1:

A comparison of demographics, pathologic characteristics, and hemoglobin and platelet values according to race.

Variable	AA, %	CA, %	P-value*
Age (years) ± SD (n=581)	62 (12)	64 (13)	0.05
Sex
Female (n=305)	56	50	0.19
Stage
I (n=163)	31	27	0.43
II (n=135)	20	25
III (n=142)	24	25
IV (n=141 )	25	24
Colonic Location			0.03
Proximal (n=251)	48	48
Distai (n=141 )	33	24
Rectal (n=124)	18	27
Grade
Low (n=437)	87	88	0.72
High (n=58)	13	12
Histology			0.28
Mucin containing (n=61)	9	11
NOS (n=520)	91	89
Treatment
No treatment (n=23)	5	4	0.44
Surgery (n=315)	56	53
Chemotherapy (n=170)	29	29
Radiation (n=73)	10	14
Obesity			0.04
Yes (n=71)	16	10
Hypertension			<0.0001
Yes (n=411)	85	64
Hypercholesterolemia			0.40
Yes (n=277)	50	46
Current Smoker**			0.01
Yes (n=142)	20	31
Hemoglobin Concentration g/dL ± SD	10.6 (1.85)	11.4 (1.97)	0.0001
Anemia (n=439)	84	72	0.001
Platelet Count k/L ± SD	295.2 (107)	281 (112.6)	0.17
Thrombocytosis (n=77)	16	12	0.31
Thrombocytopenia (n=35)	7	6

^*Significance values were calculated under chi-square tests for categorical variables, or two-sample t-test for age, hemoglobin concentration, and platelet counts.

^**smoklng status was mlsslng for n=63

We found that patients with anemia had lower survival compared to those with normal hemoglobin levels (Table 2). The risk of death was about 40% higher in patients with anemia compared to those in the normal range, even after adjustment for age, sex, race, stage, and treatment. Abnormal platelets at diagnosis were also significantly associated with poorer survival compared to those in the normal range. In the Cox regression models, the adjusted HRs for thrombocytosis (vs. normal) were associated with a 50% higher risk of death, whereas the mortality was 2.5 times greater in those with thrombocytopenia vs. normal values. The combined measures of anemia plus thrombocytosis or thrombocytopenia predicted 2- and 3.5-times higher risk of death, respectively, compared to having neither. Adjustment for age, sex, stage, race, and treatment variables had a marginal impact on the HRs (Table 2). A clinical diagnosis of obesity, hypercholesterolemia, hypertension, or current smoking status had no appreciable effect on the HRs (data not shown).

Table 2:

Median survival for anemia, abnormal platelet counts and risk of death.

	# of deaths/#subjects	Survival (mos) (95% CI)¶	HR† (95% CI)	HR‡ (95% CI)
Anemia#
Normal	54/142	128 (80 -*)	1.0	1.0
Anemia	237/439	53 (43–78)	1.76 (1.31–2.37)	1.44 (1.06–1.98)
Abnormal platelets
Thrombocytopenia (<150 × 109/L)	22/35	26 (7–92)	1.93 (1.24–2.99)	2.50 (1.59–3.94)
Normal (150–399 × 109/L)	220/469	81 (63–107)	1.0	1.0
Thrombocytosis (400 × 109/L)	49/77	26 (17–40)	1.98 (1.45–2.70)	1.54 (1.11–2.12)

^#Anemia defined for women as < 12.0 g/dL, and for men, the lower threshold was 13.5 g/dL.

^¶p-value for the log-rank test for differences anemia (p =0.0002), abnormal platelets (p=0.0001)

^†Univariate model

^‡Multivariable model adjusted for age, sex, race, stage and treatment

^*Upper bound of survival time not reached

Bolded HRs are significant at p < 0.05

In those with anemia compared to those without, we observed significantly lower survival times (Table 2). However, the race-specific HRs indicated that both CAs and AAs with anemia compared to those in the normal range had non significantly higher risks of death (HRs 1.37 [95% CI: 0.96–1.96] and 1.67 [95% CI: 0.90–3.09], respectively), and the interaction between race and anemia was not significant (p=0.58 [Table 3]). Both CAs with thrombocytopenia and AAs with either thrombocytosis or thrombocytopenia had significantly lower survival in the Kaplan-Meier analyses compared to patients with normal values at diagnosis (Table 3). In the Cox regression models, we observed that the HR for AAs with thrombocytosis vs. normal platelets was associated with a 2.51 (95% CI: 1.52–4.15) times higher risk of death, whereas, in CAs, the risk was not significantly elevated, with an HR of 1.15 (95% CI 0.75–1.75); p=0.03 for interaction. On the other hand, the HR for thrombocytopenia was significantly higher in AAs (HR 4.16 [95% CI: 2.11–8.21]) compared to those with normal levels; CAs also exhibited elevated risk of death in the same comparison, with an HR of 1.90 (95% CI: 1.04–3.48); for interaction = 0.24. AAs with both anemia and thrombocytosis (compared to those patients with neither) appeared to be at higher risk of death than CAs in the same comparison, with HR 3.11 (95% CI: 1.49–6.48), vs. CAs, with HR 1.40 (95% CI: 0.82–2.40), and global p-value for interaction = 0.04 (supplemental table). Both CA and AA patients with both anemia and thrombocytopenia vs. those in the normal range had a respectively 3- and 5-fold increased case fatality rate, but the low number of patients in these extreme categories prevented us rrom drawing firm conclusions (supplemental table).

TABLE 3:

The risk of death by race (African Americans vs. Caucasian Americans) and categories of hematopathology (anemia, thrombocytopenia, thrombocytosis vs. normal)

			Race†		Hematopathology†
	#dead/#subjects		HRs 95% CI‡		HRs 95% CI‡
Anemia	AAs	CAs	CAs	AAs	CAs	AAs
No anemia	12/30	42/112	1.0	0.98 (0.51–1.89)	1.0	1.0
Anemia	87/158	150/281	1.0	1.19 (0.91–1.56)	1.37 (0.96—1.96)	1.67 (0.90–3.09)
Platelets
Thrombocytopenia	9/12	13/23	1.0	1.74 (0.74–4.10)	1.90 (1.04–3.48)	4.16 (2.11–8.21)
Normal platelets	68/143	152/326	1.0	1.02 (0.76–1.36)	1.0	1.0
Thrombocytosis	21/30	28/47	1.0	2.09 (1.17–3.75)	1.15 (0.75–1.75)	2.51 (1.52–4.15)

^†P for the interaction using the LR test between race and anemia on risk of death= 0.58. The global interaction between categories of platelet measures and race and death, LR test, p=0.07; p for interaction between race x thrombocytopenia using the Wald statistic p=0.24; race x thrombocytosis using the Wald statistic, p=0.03.

^‡HRs for race and categories of hematopathology are adjusted for age, sex, stage, and treatmentHRs that are bolded are significant at p< 0.05.

Discussion

This report shows that both abnormal platelet measures at diagnosis and anemia are directly associated with a higher risk of death, confirming results from earlier studies. This analysis shewed that thrombocytosis led to a higher risk of death among AAs than CAs, especially among individuals with anemia. Overall, our results underscore the importance of considering the joint presence of platelets and hemoglobin measures in assessing risk and incorporating race when assessing the role of hematological measures in CRC outcomes.

Overall, our findings are consistent with several previous observational studies that have demonstrated a higher risk of death in patients with abnormal platelet and hemoglobin variables.^10,11,13,33 A recent meta-analysis¹³ with over 5,000 patients with CRC reported a weighted HR of 1.61 (95% CI: 1.19–2.04) in those with pre-operative thrombocytosis (vs. nonthrombosis) which is consistent with our result of HR 1.54 (95% CI: 1.11–2.12). Moreover, a validation study with over 10,000 patients with CRC undergoing surgical resection³³ found that thrombocytopenia (vs. normal counts) at diagnosis was associated with a 2.51 (95% CI: 1.544.08) times higher risk of 30-day post-surgical mortality; their findings also mirror what we found for overall survival: HR 2.50 (95% CI: 1.59–3.94). Finally, a meta-analysis¹¹ comprising over 3,500 patients with CRC examined the impact of preoperative anemia (vs. normal sex specific hemoglobin values) and found a 1.56 (95% CI: 1.30–1.88) times higher risk of death, also consistent with our cohort, HR 1.44 (95% CI: 1.06–1.98). The similarities of the point estimates of our results with previous research suggest good internal validity for our patient cohort. Our original findings indicated the association that patients with thrombocytosis or thrombocytopenia and anemia compared to patients with no abnormal measures were at 2- and 3.5-times higher risk of death, respectively. These findings are essential to evaluate further in future studies.

Our study is the first to comprehensively evaluate the importance of abnormal platelets, anemia, and combined measures in CRC patients on survival by race. While no studies have evaluated the racial differences in anemia and subsequent risk of death from CRC, in patients with breast cancer³⁷ and thyroid cancer³⁸, anemia at diagnosis was higher in AAs compared to CAs and a marker of poor prognosis. Only one previous investigation has examined the association between thrombocytosis and the risk of death in AA and CA patients.³⁹ In that study, AAs had a three times higher risk of death than patients without thrombocytosis, similar to our results (AAs HR 2.51 [95% CI: 1.52–4.15]). In the same comparisons among CAs, the point estimates were about one half of the risk of AAs, also in-line with what we observed (HR was 1.15 [95% CI: 0.75–1.75]). Larger, more diverse clinical cohorts are needed to clarify the relationship between abnormal hemoglobin and platelet measures at diagnosis and death in AAs vs. CAs, especially for early-onset CRCs, where the survival disparity is the greatest.

Mounting pre-clinical and clinical evidence demonstrates that platelets actively drive disease progression and metastatic expansion.^10,13,18 Platelets are the primary source of TGF-Beta^18,40 in the tumor microenvironment, and together with IL-6 can promote inflammation in the tumor.^41,42 Furthermore, platelets are a leading source of VEGF and other endothelial growth factors,⁴³ which promote angiogenesis and metastasis. Platelets can impair NK and T-cell cytotoxic functions,^18,44,45, which suppress immune surveillance and cytotoxic killing capacity enabling tumors to expand. The presence of anemia is also associated with increased circulating VEGF.⁴⁶ Our results are consistent with the idea that abnormal hemoglobin and platelet function may contribute to this elevated inflammatory milieu and reduced killing capacity, thereby leading to tumor growth and a higher risk of death observed in AA patients.

Previous studies of inheritance may provide some insight into the mechanism underlying racial disparities we observed. For example, the estimates for platelet and hemoglobin level heritability range from 54% to 87%, respectively.^8–9 A few genome-wide studies^47,48 have identified several genes associated with African ancestry which impact red blood cell traits leading to lower circulating hemoglobin levels.^49,50 These studies may partly explain poorer outcomes for AAs, given the clinical data demonstrating that anemia decreases platelet inhibition,⁵¹ leads to enhanced platelet aggregation, and promotes thrombocytosis.²⁸ As AAs have lower circulating hemoglobin levels and superior platelet activation and aggregation capacity,^48,52–56 it follows that thrombocytosis would be more pronounced in AAs compared to CAs, and is consistent with what we report here. On the other hand, AAs with thrombocytopenia at diagnosis were also at an elevated risk of death compared to CAs. The specific mechanisms driving thrombocytosis and thrombocytopenia are likely different, however, and require further investigation. Further, clinical factors such as obesity, hypertension, and diabetes are associated with anemia, platelet counts,^25,57,58 and have a higher prevalence in AAs than CAs.^25,37,59 Although we identified significant differences in the diagnoses of clinical obesity, hypertension, and smoking status by race, adjusting for these factors did not affect point estimates for race, thrombocytopenia, thrombocytosis, or anemia on the risk of death. Finally, our results suggest that antiplatelet medications (such as aspirin⁶⁰) might be efficacious in reducing the racial disparity in survival.

Conclusion

Our study has several strengths, including a large racially diverse population of patients with multiple baseline hematologic values and clinical and clinicopathologic characteristics. However, we also recognize the limitations of our study. Detailed treatment regimen data, which could have confounded or modified the association between race and CRC survival, were not available. The addition of comorbid condition variables to our models did not materially change the HRs for race or hematologic measures; however, we did not have information on the use of medications to treat these conditions, which may have impacted our ability to detect the confounding. We also lacked information on the genetic ancestry of patients making it difficult to understand the contribution to differences in the immune reaction by race. Overall, our results point to the need for detailed studies identifying the immune prognostic signatures and how they differ by race and their potential impact on treatment to help advance understanding of the racial disparity in CRC survival.

Highlights.

• African Americans, compared to Caucasians with colorectal cancer, have worse survival even after adjusting for stage of disease and treatment.

• The risk of death is three times higher in African Americans vs. Caucasians with anemia and thrombocytosis compared to those with normal values.

• Elevated platelet counts at diagnosis may shed light on the racial disparity in colorectal cancer survival.