key: cord-1012059-xa4dq8pw authors: Ospina, Aylen Vanessa; Bruges, Ricardo; Mantilla, William; Triana, Iván; Ramos, Pedro; Aruachan, Sandra; Quiroga, Alicia; Munevar, Isabel; Ortiz, Juan; Llinás, Néstor; Pinilla, Paola; Vargas, Henry; Idrobo, Henry; Russi, Andrea; Kopp, Ray Manneh; Rivas, Giovanna; González, Héctor; Santa, Daniel; Insuasty, Jesús; Bernal, Laura; Otero, Jorge; Vargas, Carlos; Pacheco, Javier; Alcalá, Carmen; Jiménez, Paola; Lombana, Milton; Contreras, Fernando; Segovia, Javier; Pino, Luis; Lobatón, José; González, Manuel; Cuello, Javier; Bogoya, Juliana; Barrero, Angela; de Lima Lopes, Gilberto title: Impact of COVID‐19 Infection on Patients with Cancer: Experience in a Latin American Country: The ACHOCC‐19 Study date: 2021-07-01 journal: Oncologist DOI: 10.1002/onco.13861 sha: c0724136d664c550fff787b7106a16b019a5cea4 doc_id: 1012059 cord_uid: xa4dq8pw INTRODUCTION: The ACHOCC‐19 study was performed to characterize COVID‐19 infection in a Colombian oncological population. METHODOLOGY: Analytical cohort study of patients with cancer and COVID‐19 infection in Colombia. From April 1 to October 31, 2020. Demographic and clinical variables related to cancer and COVID‐19 infection were collected. The primary outcome was 30‐day mortality from all causes. The association between the outcome and the prognostic variables was analyzed using logistic regression models and survival analysis with Cox regression. RESULTS: The study included 742 patients; 72% were >51 years. The most prevalent neoplasms were breast (132, 17.77%), colorectal (92, 12.34%), and prostate (81, 10.9%). Two hundred twenty (29.6%) patients were asymptomatic and 96 (26.3%) died. In the bivariate descriptive analysis, higher mortality occurred in patients who were >70 years, patients with lung cancer, ≥2 comorbidities, former smokers, receiving antibiotics, corticosteroids, and anticoagulants, residents of rural areas, low socioeconomic status, and increased acute‐phase reactants. In the logistic regression analysis, higher mortality was associated with Eastern Cooperative Oncology Group performance status (ECOG PS) 3 (odds ratio [OR] 28.67; 95% confidence interval [CI], 8.2–99.6); ECOG PS 4 (OR 20.89; 95% CI, 3.36–129.7); two complications from COVID‐19 (OR 5.3; 95% CI, 1.50–18.1); and cancer in progression (OR 2.08; 95% CI, 1.01–4.27). In the Cox regression analysis, the statistically significant hazard ratios (HR) were metastatic disease (HR 1.58; 95% CI, 1.16–2.16), cancer in progression (HR 1.08; 95% CI, 1.24–2.61) cancer in partial response (HR 0.31; 95% CI, 0.11–0.88), use of steroids (HR 1.44; 95% CI, 1.01–2.06), and use of antibiotics (HR 2.11; 95% CI, 1.47–2.95). CONCLUSION: In our study, patients with cancer have higher mortality due to COVID‐19 infection if they have active cancer, metastatic or progressive cancer, ECOG PS >2, and low socioeconomic status. IMPLICATIONS FOR PRACTICE: This study's findings raise the need to carefully evaluate patients with metastatic cancer, in progression, and with impaired Eastern Cooperative Oncology Group status to define the relevance of cancer treatment during the pandemic, consider the risk/benefit of the interventions, and establish clear and complete communication with the patients and their families about the risk of complications. There is also the importance of offering additional support to patients with low income and residence in rural areas so that they can have more support during cancer treatment. Since the first trimester of 2020, the world population has been facing the greatest health crisis in recent years because of the COVID-19 pandemic. This situation has generated a great public health challenge and has forced the health system to make rapid adaptations to the hospital infrastructure. The first published reports of patients infected by COVID-19 showed that the risk of complications increased with age [1, 2] . Furthermore, predisposing factors for developing respiratory failure included smoking, diabetes, hypertension, and cancer. The initial cohort of patients with cancer from China, described by Liang et al., included 18 patients, which limited the possibility of making global conclusions; however, it showed that patients with cancer had a higher risk of serious events including death (39% vs. 8%; p = .0003), mainly those who received oncological treatment during the month before the infection (75% vs. 43%) [3] . Additional studies showed an increased risk of complications in men, patients with lung cancer, and patients with advanced-stage neoplasms [4, 5] . In June 2020, the results of the TERAVOLT and CCC19 registries identified the following factors associated with mortality in patients with cancer: age, male sex, smoking, number of comorbidities, functional status, and presence of active cancer [6, 7] . The Gustave Roussy Hospital study identified that age older than 70 years, smoking, metastatic cancer, cytotoxic chemotherapy, and an Eastern Cooperative Oncology Group (ECOG) performance status ≥2 are factors that increase the risk of death and the importance of the ECOG scale as an independent predictor of death [8] . In Colombia, the first case diagnosed with COVID-19 was confirmed on March 6, 2020, and the number of infected increased exponentially as had been described in the rest of the world [9] . Based on this, the ACHOCC-19 study was proposed to characterize the behavior of COVID-19 infection in our population with cancer, given the absence of conclusive evidence at that time at the global, regional, and local levels. A cohort analytical ambispective study was carried out, with data collected from patients with cancer and COVID-19 infection from the National Cancer and COVID-19 registry of Colombia, ACHOCC-19. It was conducted from April 1 to October 31, 2020, to collect and analyze the clinical characteristics and outcomes of COVID-19 infection in patients with cancer. The first patient was registered on April 8, 2020. The ACHOCC-19 handled anonymized data. The study included adults (more than 18 years of age) with a histologically confirmed diagnosis of cancer (solid tumors), who had been treated and were free of cancer in follow-up and have suspected or confirmed infection by COVID-19 (by polymerase chain reaction [PCR]) immunoglobulin G [IgG] , or immunoglobulin M [IgM] type test or by clinical or suspected case by radiological criteria or who received systemic oncological or surgical treatment with curative intent between April 2020 and October 2020 and had confirmation of COVID-19 (by PCR, IgG, or IgM type test, or are considered a suspected case by radiological criteria. Twenty-two institutions from the main cities of the country participated (Annex 1). The data recording was carried out by 38 oncologists with experience in the health care and research areas in two moments: the first corresponded to the verification of the inclusion criteria and the second to the recording of the outcomes, this way completing a follow-up period of 30 days for all patients. This study was approved by all the ethics committees of the participating institutions. The data were collected in a digital form that included 80 variables, classified as sociodemographic, related to COVID-19, related to cancer, and follow-up variables. The following variables were included as critical, and possibly associated with complications from COVID-19 infection: age, sex, type of residence, active smoking, obesity, number of comorbidities, surgery for cancer management within the last month, cancer status, performance status according to ECOG score, type of treatment for cancer and COVID-19. The type of cancer treatment was subclassified according to the use or not of cytotoxic chemotherapy and the dates of administration of the cancer treatment in relation to the date of diagnosis of the COVID-19 infection. The primary outcome of this study is all-cause mortality within 30 days of COVID-19 diagnosis. Secondary outcomes included the requirement for noninvasive mechanical ventilation and the requirement for invasive mechanical ventilation. © 2021 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. The sample size was calculated through two alternatives, the first with a confidence of 90% and an error of 5%, obtaining a sample of 286 patients, and the second using the information from GLOBOCAN 2018 to estimate the population with cancer in Colombia and taking the incidence information of COVID-19 infection in patients with cancer of 1% reported by Liang et al., which provided a sample size of 381 patients [3, 10] . Because some of the survey questions were optional, we anticipated a nonzero level of absence for some variables. We used mode imputation for categorical variables with an absence rate of 10% or less; variables with a missing rate of more than 10% were not included in the multivariate analysis. A standardized format was made for reporting dates related to COVID-19 diagnosis, COVID-19 death, neoplasia diagnosis, and date of application of the last dose of cancer treatment, which allowed an interaction of date differences for the creation of endtime variables, which were stipulated as measured in days. In the first analysis using descriptive statistics, all sociodemographic variables were included, as were clinical characteristics before and during COVID-19 infection. Additionally, the χ 2 distribution was used to study correlations between different variables and the primary outcome, to determine whether there were independent relationships. Subsequently, the correlations between the study variables and the primary outcome were examined using a logistic regression model to perform an adjusted multivariate data analysis. In the multivariate model, age was defined in an interval variable, and an adjustment for multiple comparisons was made according to the covariate. The demographic risk variables were adjusted for each other and were the adjustment for cancer and COVID-19 treatment covariates. We evaluated the goodness of model using the Hosmer and Lemeshow statistic. The prediction capacity of the model was evaluated through the Omnibus test, and, finally, a summary of the model was made, in which the À2 log-likelihood was used to determine the adjustment of the model to the data. The R squared of Cox and Snell, and the R squared of Nagelkerke were used to calculate the proportion of variance of the dependent variable (30-day mortality) explained by the predictor variables (independent). The final variables for which the model was adjusted had a classification table by blocks to calculate the global percentage correctly predicted and the correctly predicted percentage for deaths and nondeaths. Survival analysis was performed using a Cox regression model; survival times were calculated using the date of death and the date of symptom onset, reported in days. The model was adjusted for the variables age, sex, area of residence, cancer status, obesity, smoking, number of comorbidities, cytotoxic and noncytotoxic treatment or no cancer treatment, cancer treatment during the last year, metastatic disease, number of complications due to COVID-19, and treatment for COVID-19. Descriptive analysis was performed in Stata/SE edition 16.0 (Stat Corp., College Station, TX), and all correlation and multivariate analyzes were performed in two parallel programs: R version 4.0.3 (R Foundation, Vienna) and IBM SPSS Statistics 25 (IBM, Armonk, NY). In a subsequent subanalysis, we developed a neural network, in which we identified the covariates that had the greatest sample weight for the predictive models, measuring them against the primary outcome of all-cause mortality within 30 days after the COVID-19 diagnosis. Data optimization was performed weekly in the registry to avoid loss of follow-up, alteration, or failures in data collection. Data with p values less than .05 were considered statistically significant, both in the logistic regression model and in the survival analysis. The sponsors of the study had no role in study design, data collection, analysis, interpretation of data, or writing of the article. Of the 783 patients collected in the ACHOCC-19, 41 patients were excluded because of duplication of information. All records were started during the COVID-19 pandemic and 100% had a 30-day follow-up report (interquartile range, 0-30 days). Regarding the demographic, clinical, tumor, and socioeconomic characteristics of the population (Table 1) , 534 (72%) patients were older than 51 years and 200 (27%) older than 70 years. Four hundred three (54%) were women. As risk factors for complications from COVID-19, 18 (2.42%) had active smoking, 87 (11.7%) had more than two comorbidities, 37 (14%) had an ECOG performance status >2, and 295 (38.7%) had metastatic neoplasia. The area of residence was urban for 677 (91%) patients and 354 (47.7%) belonged to the low-income socioeconomic level. The most prevalent neoplasms were breast cancer (n = 132, 17.77%), colorectal cancer (n = 92, 12.34%), and prostate cancer (n = 81, 10.9%). Regarding cancer treatment, 352 (63.9%) patients were in active treatment, of whom 213 (28.67%) received systemic treatment with palliative intention. Eighty-one patients (10.98%) had undergone surgical treatment in the month prior to COVID-19 infection. Of those who were receiving active cancer treatment, 205 (27.6%) had a partial response or stable disease and 147 (36.38%) had progressing cancer. A total of 454 (61.18%) patients had received cancer therapy within the previous 12 months. According to the type of treatment, 209 (28.18%) were receiving cytotoxic therapy and 144 (19.5%) were under noncytotoxic treatment. Regarding the clinical presentation of COVID-19 infection, 220 (29.6%) patients were asymptomatic throughout the course of the disease, 229 (Table 2 ). In the bivariate descriptive analysis, higher mortality was observed in patients with more than 70 years of age, residence in a rural area, low socioeconomic status, lung cancer, presence of more than two comorbidities, history of smoking, high acute-phase reactants at the time of diagnosis of infection, and treatment for COVID-19 infection with antibiotics, corticosteroids, or anticoagulants. In a descriptive analysis between the use of antibiotics, corticosteroids, and anticoagulation compared with the clinical presentation of the disease at diagnosis, it was evidenced these three drugs were administered to patients with the worst prognosis. One hundred two (37.78%) patients with severe disease with mechanical ventilation and 129 (47.78%) patients with moderate illness with hospitalization received antibiotics, whereas anticoagulation was administered to 58 (35.58%) patients with severe disease with mechanical ventilation and to 78 (47.85%) patients with moderate disease with hospitalization. Ninety-two (34.33%) patients with severe disease with mechanical ventilation and 145 (54.10%) patients with moderate disease with hospitalization received corticosteroids. The use of these three types of drugs in patients with mild disease or without hospitalization was significatively lower (<15%). Some possible prognostic variables had a failure rate of more than 10% and, therefore, were not included in the multivariate model. The associations between prognostic variables and 30-day all-cause mortality are shown in Table 3 . The goodness of adjustment reported in our models had a statistical value of 10.3 with a significance of 0.24, which indicates that the model has a good adjustment. The accuracy of the model to classify patients who died in 30 days was 92.7%. Several clinically relevant prognostic variables associated with an increase in 30-day all-cause mortality were identified after partial adjustment in our multivariate model: cancer status (disease progressing vs. stable disease, in partial response, remission or without evidence of disease), ECOG performance status (2 vs. 0; 3 or 4 vs. 0), and the presence of two complications from COVID-19 infection. Variables that behaved as protective factors were also identified, including residence in an urban area and absence of complications from COVID-19. Age, nutritional status, smoking, sex, type and time of administration of cancer treatment, and recent surgery did not reach statistically significant values to associate them with 30-day mortality in the multivariate logistic regression analysis. In a descriptive analysis of the variables "Hospital admission" and "Use of cancer treatment," 369 (58.9%) patients required hospital admission, of whom 44% were actively receiving cancer therapy. The frequency of mechanical ventilation was higher as the decade of age increased from 50 years, with a slight decrease after 70 years; a high frequency of invasive ventilatory support was found in the group aged 31-40. Adjusted for the gender variable, a higher frequency of invasive ventilation was evidenced in men. The result of the Cox regression as a 30-day survival analysis had the following statistically significant hazard ratios: cancer status, metastatic disease, and use of antibiotics and corticosteroids ( Table 4) . The survival plots for the statistically significant variables in the Cox model are found in Figures 1-4 . Figure 5 shows the variables adjusted in the logistic regression model with adjusted odds ratios (ORs) and 95% confidence intervals. In a deep learning analysis using a neural network, it was found that the variables with the highest predictive weight for 30-day mortality from all causes after the diagnosis of COVID-19 were ECOG status 3-4, number of complications due to COVID-19, and cancer in progression. On the other hand, variables such as living in an urban area and cancer in partial response showed little prediction of mortality, congruent with an OR lower than 1. The other variables did not have an important predictive weight in the neural network. Given the absence of epidemiologically significant data on the impact of COVID-19 infection in patients with cancer, the ACHOCC-19 study was generated from the Asociaci on Colombiana de Hematología y Oncología. The objective of this project was to collect local information on the behavior of the infection to generate recommendations applicable to our population. The participation of 38 researchers was obtained from 21 hospital institutions nationwide in 10 cities. The information collection was based on prognostic factors associated with mortality in patients with cancer that were reported in the initial series from China [3] . In addition, characteristics such as the area of residence and socioeconomic level were included, considering that in developing countries, these factors influence access to the health system and clinical outcomes. In Colombia, the first peak of the pandemic occurred between July and August 2020. During the previous months, the health system generated a gradual increase in intensive care unit beds and adaptation of hospitalization areas [11] . During the highest phase of infections, there was no collapse of the health system or a shortage of resources and supplies that impeded patient care. At the time of our cohort analysis, we did not find any series of patients with cancer and COVID-19 infection published in Latin America. The available results of the CCC19 cohort were reviewed, which shows important similarities to ours [7] . In our study, we detected that patients with cancer are at higher risk of mortality due to COVID-19 infection if they have active cancer, present metastatic or progressive neoplasia, and/or impaired functional status with an ECOG status higher than 2. Diagnostic PCR was performed in 97% of those infected, and only half of the patients had a follow-up test because of the emergence of evidence that did not support this requirement [12] . The outcome of acute-phase reactants was available in a small percentage of the population, and although it was associated with worse outcomes, this cannot be adequately interpreted. Our patients had a higher 30-day mortality rate (26%) when compared with the information available from the CCC19 cohort (13%) [7] and from a cohort of similar size from China (1.4%) [13] . These data also differ from those of the New York region, where the Mount Sinai Hospital analysis reported a mortality of 11% [14] . A series of 218 patients from the Montefiore health system showed a similar case fatality rate of 28%, although the authors recognized a selection bias toward more severe cases [15] . We consider that the high mortality in our cohort is related to the higher frequency of patients with progressive metastatic disease (36.38%) who were under palliative systemic cancer treatment (28.67%). Patients with progressing cancer died at a numerically higher rate; however, the frequency of mechanical ventilation for this group was not proportional, because many of these patients were not candidates for advanced respiratory support because of their oncological condition. On the contrary, patients with stable or responding neoplasia were the ones who received the most mechanical ventilation, which we interpreted as having their underlying oncological disease controlled. When comparing with the available series, we conclude that the geographical location does not generate differences in the clinical outcomes of COVID-19 infection, because the results of our cohort in Latin America are similar to the North American and European cohorts [6] [7] [8] 16] . For our population, it was evidenced that socioeconomic factors such as residing in rural areas and having a low level of economic income are associated with higher mortality, which we relate to difficulties in accessing the health system, presenting advanced neoplasms in progression, and presenting uncontrolled comorbidities, all generating a synergistic effect (syndemic). The absence of an association of 30-day mortality with recent surgery and administration of systemic cancer treatment during the last year suggests that approaches with curative intent, including surgical resection, and neo/adjuvant and maintenance systemic treatments, do not seem to increase the risk and do not justify stopping treatment for these patients during the pandemic. Similar findings were reported in the CCC19 cohort [7] . In the survival analysis, the administration of corticosteroids, antibiotics, and anticoagulants behaved as a risk factor for mortality; however, in the analysis stratified by severity of the COVID-19 infection, it was evidenced that the patients who received them had the worst prognosis because they had more severe clinical conditions with a higher risk of mortality and the risk cannot be associated only with the administration of the medications by itself. However, it cannot be ruled out that in patients with poor functional status and/or progressing neoplasia with severe COVID 19 disease the use of corticosteroids may behave as a factor that makes it difficult to control another associated infectious process. For this reason, we consider that this variable should be studied in greater depth. On the other hand, it is possible that the group of patients who were asymptomatic or had mild disease who did not receive any of these treatments will be responsible for the best prognosis of the patients without in-hospital treatment for COVID-19 infection. Most of the patients included were symptomatic (70%); therefore, selection bias cannot be ruled out. However, a significant percentage of asymptomatic patients (29.6%) was included, because screening had started in some institutions in the country. This sample will allow us to perform a subsequent analysis and define the characteristics of patients with cancer who remained asymptomatic and without complications during COVID-19 infection, for which there are no published data. Our findings raise the need to carefully evaluate patients with metastatic cancer, in progression, and with impaired ECOG status to define the relevance of cancer treatment during the pandemic, consider the risk/benefit of the interventions, and establish clear and complete communication with the patients and their families about the risk of complications. There is also the importance of offering additional support to patients with low income and residence in rural areas so that they can have more support during cancer treatment. Republic of Korea. Korea Disease Control and Prevention Agency Case-fatality rate and characteristics of patients dying in relation to COVID-19 in Italy Cancer patients in SARS-CoV-2 infection: A nationwide analysis in China SARS-CoV-2 Transmission in patients with cancer at a tertiary care hospital in Wuhan, China Clinical characteristics of COVID-19-infected cancer patients: A retrospective case study in three hospitals within Wuhan, China COVID-19 in patients with thoracic malignancies (TERAVOLT): First results of an international, registry-based, cohort study Clinical impact of COVID-19 on patients with cancer (CCC19): A cohort study Determinants of the outcomes of patients with cancer infected with SARS-CoV-2: Results from the Gustave Roussy cohort Cancer today. Web site of the International Agency for Research on Cancer Documentos Técnicos covid-19 Clinical characteristics of Coronavirus Disease 2019 in China Do patients with cancer have a poorer prognosis of COVID-19? An experience in New York City Case fatality rate of cancer patients with COVID-19 in a New York hospital system COVID-19 in cancer patients: Clinical characteristics and outcomean analysis of the LEOSS registry This study was completely sponsored by the Asociaci on Colombiana de Hematología y Oncología. Additionally, special thanks are given to those who do not meet the criteria of authors but collaborated with review and data collection: Ricardo Duarte, Mauricio Camelo, Sandra Franco, Javier Cuello, Carlos Ortiz, Andres Borda, Guido Gonz alez, Abraham Hernandez, Angelica Jayk, Fernando Oviedo, Marco Torregroza, Rocio Mariño, Denisse Rubio, Mariana Borras, and Marc Edy Pierre.