The Prognostic Significance of HALP Index for Colon Cancer Patients in a Hispanic-Based Population

Background Survival and recurrence rates following locoregional colon cancer surgical resection are highly variable. Currently used tools to assess patient risk are still imperfect. In the present work, we evaluate, for the first time, the prognostic value of the recently developed HALP (hemoglobin, albumin, lymphocyte, and platelet) index in Hispanic colon cancer patients. Patients and Methods. We conducted a retrospective cohort study in Mexican patients with a nonmetastatic colon cancer diagnosis who underwent surgical resection. We determined the preoperative HALP score optimal cut-off value by using the X-tile software. We plotted survival curves using the Kaplan–Meier method and performed a multivariate Cox regression analysis to explore the association of preoperative HALP score with two primary endpoints: overall survival (OS) and disease-free survival (DFS). Results We included 640 patients (49.8% female). The optimal HALP cut-off value was 15.0. A low HALP index was statistically significantly associated with a higher TNM stage. Low HALP score was statistically significantly associated with shorter median OS in the Kaplan–Meier analysis (73.5 vs. 84.8 months) and in the multivariate Cox regression analysis (HR = 1.942, 95% CI = 1.647–2.875). There was no significant association between the HALP score and DFS. Conclusions Our findings show that the HALP index is an independent factor associated with survival in Hispanic patients, despite recurrence. It seems to reflect both the anatomical extent of the disease and traditionally unaccounted nutritional and inflammatory factors that are significant for prognosis.


Introduction
Colon cancer is a public health concern. At a global level, colorectal carcinoma is currently the third most incidence of cancer and the second most common cause of cancer death. In Mexico, during 2020 alone, this cancer accounted for 14,901 new cases and 7,755 deaths [1].
Surgery is the mainstay of treatment in locoregional colon cancer and continues to be the only curative modality. However, surgery results in a cure only in approximately 50% of cases, as recurrences following surgical resection remain a major problem and are a frequent cause of eventual death [2,3].
Colon cancer is characterized by being heterogeneous. Whether recurrence occurs or not is possibly dependent on a myriad of factors that result in diferent individual risks of recurrence [4]. Some high-risk patients beneft from the addition of adjuvant chemotherapy and strict surveillance, while others do not. Te latter group of patients could be spared from signifcant chemotherapy side efects or bothersome procedures, while the former greatly beneft from these interventions, improving survival rates [4,5]. Terefore, elucidating prognostic factors is an essential task.
Te tumor-node-metastasis (TNM) staging system by the American Joint Committee on Cancer (AJCC) is currently the sturdiest and most frequently used tool for assessing patients' prognosis to guide management in colon cancer [6,7]. However, this tool still has limitations, as individuals within the same stage have highly variable survival rates, and patients' prognoses across diferent stages sometimes overlap [7,8].
Cancer progression and metastasis are not solely dependent on tumor characteristics and anatomic extent [9]. It is well established that both systemic infammation and nutrition play an important role in prognosis [10]. Tese two factors are linked to many tumor characteristics, including proliferation, invasion, metastasis, and recurrence [11]. Various infammatory and nutritional markers have been individually associated with survival outcomes in several types of cancer. Moreover, studies have used a combination of these markers to successfully predict prognosis with the use of a single index; examples include the prognostic nutritional index (PNI), the neutrophil-tolymphocyte ratio (NLR), and the platelet-to-lymphocyte ratio (PLR) [12].
In 2015, Chen et al. developed the novel HALP index, which utilized preoperative hemoglobin, albumin, lymphocyte, and platelet levels to assess gastric cancer patient's prognosis [13]. Shortly after, Jiang et al. studied the prognostic value of this index in colorectal cancer, fnding superb results. Teir study showed that patients with lower HALP scores had an increased risk of death and cancerrelated death, with lower overall survival and cancerspecifc survival. All the aforementioned associations were statistically signifcant and independent of other factors in the multivariate analysis, both in the training and validation sets [14].
Given that the HALP index is an easily reproducible tool, which utilizes widely-available routinary lab values, its adoption in clinical practice could signifcantly beneft developing countries, such as those in Latin America. Unfortunately, so far, this index has not been validated for Hispanic populations, where colon cancer epidemiological and prognostic characteristics are distinct [18]. A formal validation is needed to implement the HALP index in clinical practice.
In the present work, we aim to evaluate the prognostic value of the HALP index for Hispanic colon cancer patients in one of Mexico's largest oncological centers.

Study Design.
We conducted a retrospective cohort study in Mexican colon cancer patients without metastatic disease. We explored the association of the HALP index with our two primary endpoints: overall survival (OS) and disease-free survival (DFS). Tis study was approved by the Institutional Review Board (IRB) at Mexico's National Cancer Institute (No. 2021/143). Written informed consent from patients was waived due to the retrospective nature of this study.

Patients.
All patients with a nonmetastatic colon cancer diagnosis who underwent surgical resection at Mexico's National Institute of Cancer, between January 2008 and January 2020, were retrospectively screened for inclusion. We included patients with histologically confrmed stage I-III primary colonic adenocarcinoma who were subjected to radical surgery for the primary tumor. We excluded patients who had been previously treated for cancer, had recurrent or metastatic disease, synchronous malignancies, hereditary colon cancer, underwent other types of surgeries (local excision or palliative surgery), and/or had missing information needed to calculate the HALP index.

Data Collection.
Clinical and pathological variables of all included patients were collected from electronic medical records. Tese included age, gender, body mass index (BMI), date of diagnosis, tumor characteristics (location, diferentiation grade, presence of lymphovascular/perineural invasion, and staging characteristics), surgical margin status, and information regarding complications. Also, to calculate the HALP index, preoperative serum albumin, hemoglobin, lymphocyte, and platelet values were collected for all patients. All patients were staged according to the 7th edition AJCC TNM staging system [6]. Information regarding patients' vital status and disease recurrence was collected in order to assess our primary endpoints (OS and DFS). Patients were followed-up regularly (every two weeks while receiving chemotherapy and every two months otherwise). Follow-up care included clinical examinations, laboratory testing, and imaging (TAC, PET-CT), according to international guidelines. Follow-up for each patient began at the date of diagnosis and continued until (1) loss of followup, (2) death, or (3) last visit before the cut-of date (January 2020).

Statistical Analysis.
HALP index was calculated with the following formula: hemoglobin level (g/L) × albumin level (g/L) × lymphocyte (/L)/platelet count (/L) [13]. Te optimal cut-of value for HALP was determined by using the X-tile software (Version 3.6.1, Yale University, USA) [19]. Te cutof with the lowest p-value calculated from the chi-squared test for OS was selected and patients were classifed as having either low or high HALP. Te optimal cut-of value was frst determined for the entire cohort, and later it was identifed independently within two subgroups (patients who underwent emergency surgery vs. patients who underwent elective surgery), in order to validate our obtained value.
For the analysis, quantitative variables were converted into categories (groups). Age was grouped according to three categories (younger, middle-aged, and senior adults) and BMI was grouped according to nutritional status categories. Frequencies and proportions were reported for all collected variables. Chisquared test was used to detect associations between the collected variables and the dichotomized HALP index.
Te Kaplan-Meier method was used to calculate the median OS and DFS for each variable and plot survival curves. A log-rank test was used to compare survival between groups. Additionally, multivariate Cox regression analysis was performed to identify associations between each variable and survival; hazard ratios (HRs) and the respective 95% confdence intervals (CIs) were calculated. Only statistically signifcant variables in the univariate analysis were included in the subsequent multivariate analyses. All statistical analyses were performed with SPSS software (SPSS 19.0, IBM, Chicago, IL, USA). A two-tailed p value of <0.05 was considered statistically signifcant.

Patient
Characteristics. 893 patient records were screened for inclusion. Ultimately, 640 patients were deemed eligible and included in our fnal sample.

Cut-Of Value Determination.
Te optimal cut-of value in the entire cohort for HALP was set at 15.0, and with this dividing point, patients were classifed as having either a low or high HALP index. Te optimal cut-of points for the two evaluated subgroups, patients who underwent elective surgery vs. those who had emergency surgery, were 15.0 and 14.7, respectively. Given their similarity to the cut-of value obtained for the whole cohort, this value (15.0) was concluded to be a valid cut-of point. X-Tile results and plots are displayed in Figure 1.

3.3.
HALP Association with Clinicopathological Characteristics. HALP index was statistically signifcantly associated with pTNM stage, colon cancer location, BMI, tumor diferentiation degree, and achievement of adequate lymphadenectomy (12 or more resected lymph nodes).

HALP and Clinicopathological Characteristics Association
with Survival. Te median follow-up time, OS, and DFS were 46.39, 51.9, and 47.9 months, respectively. In the (univariate) Kaplan-Meier analysis, low HALP score was signifcantly associated with shorter median OS (73.5 vs. 84.8 months; log-rank test p � 0.013; Figure 2), as were advanced pTNM stage, poorly diferentiated tumors, lymphovascular invasion, perineural invasion, presence of cancer obstruction, N positive disease, and positive surgical margins. Multivariate Cox regression analysis established that a low HALP score is an independent factor associated with shorter OS (HR � 1.942, 95% CI � 1.647-2.875; p � 0.031). All variables' association with OS can be reviewed in Table 2.
Tere was no statistically signifcant association, in the univariate ( Figure 3) or multivariate analyses, between HALP score and DFS. Te only statistically signifcant variable associated with DFS in the multivariate analysis was cancer perforation (HR � 1.692, 95% CI � 0.986-2.902; p � 0.046). All variables' association with DFS can be reviewed in Table 3.

Biological and Clinical Signifcance of the HALP Index.
Te HALP index has been found to be a reliable marker for prognosis in cancer patients. It utilizes individuallyvalidated preoperative markers and computes them into a single value, able to identify patients with a higher risk, utilizing the following formula: hemoglobin (g/L) × albumin (g/L) × lymphocytes (/L)/platelets (/L) [13]. Patients with a low HALP score have been shown to have worse outcomes, both in previously published Asian-based studies and in the present, Hispanic-based, study.
Hemoglobin has been widely validated as a prognostic factor for disease progression and survival in a variety of human cancers. Cancer-related anemia (CRA) is present in over 30% of patients at diagnosis. CRA is a consequence of chronic infammation and poor nutritional status. Both tumor cells and tumor-reacting immune cells release proinfammatory cytokines that directly alter the hematopoietic environment and deteriorate nutritional status, further worsening anemia.
Patients with advanced cancer have a more intense and prolonged infammatory and nutritiously impoverished state. Consequently, hemoglobin levels serve as an indicator of the degree of disease progression [20]. Moreover, anemiagenerated hypoxia in the tumor microenvironment increases the tumor's proliferative and metastatic potential. Anemia has been directly linked to treatment resistance and aggressive disease [21].
Since anemia is an independent risk factor for perioperative morbidity and mortality, hemoglobin levels are a commonly assessed value in surgical patients [22]. In the context of colorectal tumor surgical resection, anemia has    been well-validated as a prognostic factor associated with decreased OS and DFS [21]. Serum albumin is one of the most widely used markers to assess nutritional status [23]. Malnutrition in cancer patients derives from cancer's catabolic state and infammationinduced anorexia. It results in metabolic alterations and muscle wasting, which are directly linked to an increased risk of chemotherapy toxicity, postoperative complications, and death [24]. Concomitantly, albumin is a main negative acute phase reactant. Similarly to hemoglobin, its decay refects the systemic infammation produced by cancer [25]. Albumin is also known to stabilize cell growth, promote DNA repair and have antioxidant properties [26]. In vitro, albumin has been shown to suppress tumor proliferation [27]. However, this efect may vary in vivo depending on the stage the tumor is at, so further studies are needed to elucidate the direct efect of albumin on tumor progression. Clinically, albumin levels are associated with the development of surgical complications and have been well-defned as a long-term survival predictor [26]. Particularly, hypoalbuminemia is associated with wound healing and anastomotic complications [28].
Lymphocytes play an important role in the immune response toward malignant cells. Trough both humoral and cellular immune responses, lymphocytes are able to limit tumor growth and metastatic potential. Lymphocyte count and tumor-infltrating lymphocytes are thought to be indicators of the host's ability to mount an efective immune response to cancer [29]. In colorectal cancer, lymphopenia has been found to be associated with decreased chemotherapy efectiveness and lower overall survival [30,31].
In contrast with the aforementioned factors, low platelet count has been consistently related to a better prognosis. A cohort that included 112,231 cancer patients found that a high platelet count was associated with an increased rate of cancerspecifc death [32]. Platelets can release a variety of cytokines, growth factors, and proangiogenic molecules which directly induce tumorigenesis, cancer proliferation, and metastasis. In addition, platelets are able to engulf tumor cells, aiding in cancer immune evasion and further promoting uncontrollable growth and dissemination [33]. Colon cancer patients with elevated platelet counts have shorter OS and DFS [34].
Te HALP index efciently compiles the aforementioned marker values into a joint score, to bolster their independent prognostic efect. Tis way, a single and practical value is put forth with the aim of reaching the required predictive value for implementation in clinical practice.

Key
Results. In our study, the HALP index was found to be independently associated with overall survival. Having a low HALP score nearly doubled the hazard of death (HR � 1.942, 95% CI � 1.647-2.875; p � 0.031). Te HALP index represented the third strongest variable associated with OS, right after the well-defned and widely implemented variables: TNM stage (Stage II vs. Stage III) and surgical margin status (R0 vs. R1/R2).
We did not fnd a statistically signifcant association between the HALP index and DFS. Tis could be attributable to the fact that DFS is largely dependent on other variables that were not included in our analysis, such as receiving adequate adjuvant chemotherapy. It should be noted that a statistically signifcant association was not found between the TNM stage and DFS either, which gives us further reason to believe that other factors, which were not considered in our analysis, played an important role in determining DFS.
It should be borne in mind that the association identifed here between a high HALP score and longer OS suggests that the HALP index is able to predict long-term survival, despite recurrence.
Our sample size was fairly large and the included patients had clinical characteristics similar to those included in previous colon cancer studies undertaken in Mexico and   Journal of Oncology 7 other Latin-American countries [35][36][37]. Surgical quality indicators, such as successful lymph node resection and microscopically negative surgical margins, were most often achieved. All this supports the generalizability of results to other Hispanic colon cancer populations where quality standards of treatment are met. Te optimal HALP index cut-of point for our population was lower than that reported in previous studies. Using the X-tile software, we determined that 15.0 was the optimal cut-of value, while previous study cut-of points ranged from 22.2 to 56.8 [13,15]. It should be noted that those studies were undertaken in Asian populations. Te diference in our cut-of point may be purely incidental, but it could also stem from biological and clinical diferences between Asian and Hispanic populations.
In order to validate our cut-of point of 15.0, we independently calculated the optimal cut-of point in two subgroups (patients who underwent elective surgery vs. those who underwent emergency surgery). In these calculations, we obtained nearly identical cut-of values (15.0 and 14.7, respectively), which confrms the validity of this cut-of point in our overall population. Moreover, even though the nutritional and infammatory status of patients undergoing elective and emergency surgery may difer signifcantly, our results support the adoption of the same cut-of point for both populations.
Previous studies have reported female gender [12,13,15,17] and older age [11][12][13]15] to be associated with lower HALP scores. Reasons for this association include the fact that female patients tend to have lower hemoglobin levels, while older patients tend to have reduced levels of both hemoglobin and albumin [13,38,39]. Nevertheless, this association was not seen in our population. In contrast, left colon cancers in our cohort were associated with lower HALP scores, an association that has not been reported elsewhere. Just as in previous reports, our study found the TNM stage to be associated with the HALP score [12,13,[15][16][17]. Tis is consistent with the belief that computed values for HALP calculation are correlated with the degree of disease progression.

Study Limitations.
Our study has some potential limitations. It was a single-center study undertaken in a thirdlevel cancer hospital in Mexico City, which might restrict, to some degree, the generalizability of results to other countries and populations. Additionally, the retrospective nature of this study increases the risk of selection bias. Further assessment of the HALP index with our cut-of point in diferent Hispanic-based populations should be undertaken.

Future Remarks.
Since the HALP index is constructed through modifable lab values, it is reasonable to believe that preoperative interventions could be undertaken in patients with a low HALP to improve outcomes. Eforts to improve some of the individual HALP-included values have already been undertaken [20,26]. Knowledge gathered on the prognostic value of the HALP index in our population might indicate that potential preoperative interventions could

Conclusion
Overall, our fndings suggest that the HALP index is a viable independent preoperative predictor of survival for Hispanic patients with TNM stages I-III who are undergoing primary tumor surgical resection. Tese study results give us reason to believe that the HALP index refects both the anatomical extent of the disease, measured through the TNM staging system, and traditionally unaccounted nutritional and infammatory factors that play a signifcant role in prognosis. Te wide availability of the routinary lab values needed to compute this score, along with the very practical nature of its implementation, make its adoption in clinical practice feasible. We provide a cut-of value that can be used in similar populations to assess prognosis and guide management.

Data Availability
All datasets used for the development of this work are available from the corresponding author upon reasonable request.