The impact of the COVID-19 pandemic on the short-term survival of patients with cancer in Northern Portugal.

The COVID-19 pandemic led to potential delays in diagnosis and treatment of cancer patients, which may negatively affect the prognosis of these patients. Our study aimed to quantify the impact of COVID-19 on the short-term survival of cancer patients by comparing a period of 4 months after the outbreak began (2 March 2020) with an equal period from 2019. All cancer cases of the esophagus, stomach, colon and rectum, pancreas, lung, skin-melanoma, breast, cervix, and prostate, from the Portuguese Oncology Institute of Porto (IPO-Porto) and diagnosed between 2 March and 1 July of 2019 (before COVID-19) and 2020 (after COVID-19) were identified. Information regarding sociodemographic, clinical and treatment characteristics were collected from the cancer registry database and clinical files. Vital status was assessed to 31 October of the respective years. Cox proportional hazards regression was used to estimate crude and propensity score-adjusted hazards ratio (HR) and 95% confidence intervals (95% CIs) of death. During follow-up to 31 October, there were 154 (11.8%) deaths observed before COVID-19 and 131 (17.2%) after COVID-19, corresponding to crude and adjusted HRs (95% CI) of 1.51 (1.20-1.91) and 1.10 (0.86-1.40), respectively. Significantly higher adjusted hazards of death were observed for patients with Stage III cancer (HR = 2.37; 95% CI: 1.14-4.94) and those undergoing surgical treatment (HR = 3.97; 95% CI: 1.14-13.77) or receiving radiotherapy (HR = 1.96; 95% CI: 1.96-3.74), while patients who did not receive any treatment had a lower mortality hazards (HR = 0.62; 95% CI: 0.46-0.83). The higher overall short-term mortality observed during the COVID-19 pandemic largely reflects the effects of the epidemic on the case-mix of patients being diagnosed with cancer.

95% confidence interval

Coronavirus disease 2019

hazard ratio

International Statistical Classification of Diseases and Related Health Problems 10th Revision

Portuguese Oncology Institute of Porto

propensity score regression adjustment

severe acute respiratory syndrome coronavirus 2

INTRODUCTION

The novel coronavirus disease 2019 (COVID‐19) first emerged in 2019 in Wuhan, China, and rapidly grew into a global pandemic. As of 22 November 2020, there were nearly 57.8 million confirmed cases and 1.3 million deaths reported. The first COVID‐19 cases in Portugal were confirmed on 2 March 2020, in the northern region, and the first death was confirmed on 16 March. The country entered a lockdown (State of Emergency) on 18 March 2020, which began to be lifted on 2 May 2020. , A wide range of measures to mitigate the spread and morbidity of the virus were adopted to ensure that health systems had the ability to provide high‐quality, accessible and sustainable services. Most health‐care settings implemented minimal services with programmed activity being canceled or suspended, and patients may have delayed routine health procedures or assessment of serious symptoms, due to fear of visiting providers that were also handling suspected COVID‐19 cases. , , , , , Globally, the World Health Organization estimated that 40% of countries reported partial or complete disruptions for cancer treatment.

For patients with cancer, health‐care professionals have had to consider how to balance the delay in diagnosis or treatment against the risk of COVID‐19. This includes mitigating the risks for significant care disruptions associated with the social distancing limits in place and managing the appropriate allocation of available health‐care resources, involved in the diagnosis, treatment and follow‐up of patients with cancer, all of which may impact patient prognosis. , , In fact, a systematic review found that even 4 week delays in cancer treatment are associated with increased mortality. This highlights that cancer treatment delays are a problem, and may have been exacerbated during the COVID‐19 pandemic.

Most research has focused on the direct mortality caused by COVID‐19. Specifically, individuals with underlying health conditions, such as active cancer, have been found to be more vulnerable to complications from COVID‐19, , , as well as higher mortality. , , , Additionally, some studies have estimated the number of excess deaths associated with the pandemic. In the United States of America, over 25% of excess deaths between February and September 2020 were attributable to causes of death other than COVID‐19. Likewise, in Portugal, between March and April 2020, an excess of 2400 to 4000 deaths were estimated, yielding an excess mortality of 3.5‐ to 5‐fold higher than what can be explained by the official number of COVID‐19 deaths reported. Nevertheless, less attention has been paid to the indirect impact of the pandemic on other health conditions, such as cancer, especially considering the potential delays in diagnosis and treatment.

Therefore, our study intends to quantify the impact of the COVID‐19 outbreak on the short‐term survival of patients diagnosed at IPO‐Porto by comparing a period of 4 months after the beginning of the outbreak in Portugal (2 March 2020) with the same period in the previous year with vital status follow‐up to 31 October of the respective year.

METHODS

Setting

The Portuguese Oncology Institute of Porto (IPO‐Porto) belongs to the National Health Service and provides specialized cancer care. IPO‐Porto is one of the largest cancer‐dedicated hospitals in Portugal, admitting patients from all over the country, although mainly from the Northern region. IPO‐Porto provides care to more than 45 000 patients covering the entire cancer continuum.

COVID‐19 plan at IPO‐Porto

After the rapid spread of the novel coronavirus worldwide in early 2020, IPO‐Porto published a COVID‐19 Contingency Plan on 18 February with a guide on procedures to be followed with any suspected COVID‐19 case. The first confirmed COVID‐19 cases in Portugal were reported on 2 March and, 2 weeks later, IPO‐Porto published recommendations regarding surgeries, outpatient appointments, diagnostic tests and medical procedures, ambulatory care and cancer‐specific emergency service. With Portugal entering the State of Emergency on 18 March, IPO‐Porto reviewed their Contingency Plan and a COVID‐19 Crisis Office was created on 19 March. Beginning in May, in line with the easing of restrictions in Portugal, IPO‐Porto stated that activity levels should gradually and progressively return to normal; recommendations were published for surgeries, outpatient appointments, diagnostic tests and medical procedures, day hospital, hospitalizations, and the clinical research unit.

Study design and data collection

Cases of invasive tumors of the esophagus (code C15 of the International Statistical Classification of Diseases and Related Health Problems 10th Revision), stomach (C16), colon and rectum (C18‐20), pancreas (C25), lung (C34), skin‐melanoma (C43), breast (C50), cervix (C53), and prostate (C61) diagnosed between 1 February and 1 July 2019 and 2020 at IPO‐Porto were identified. The included cancer sites are among the most common solid cancers diagnosed and treated at IPO‐Porto,30 are targeted by organized screening or have low overall survival.

Information on cases' sociodemographic, cancer and treatment (follow‐up to 1 July 2019 or 2020) were collected from the cancer registry database and clinical files between 1 May and 31 August 2020. Additionally, information regarding COVID‐19 diagnosis until 31 October 2020 was obtained for patients diagnosed with cancer in 2020.

Vital status and date of death, if applicable, were assessed through the National Health Service database up to 31 October of the corresponding year of diagnosis in the first week of November 2020. After the identification of deceased cases, information regarding the cause of death was obtained from the cancer registry database and clinical files.

Statistical analyses

For the current study, cancer cases, except for skin‐melanoma, who received the first cancer treatment outside IPO‐Porto were excluded, and only the first primary cancer diagnosed among each patient was considered. Data from 2 March to 1 July were used to evaluate differences in mortality before and after COVID‐19. Data from February were used as “negative controls” to compare mortality during periods without COVID‐19 cases in Portugal.

Survival was defined as the time between the date of the first primary cancer diagnosis and the date of death by any cause, and was calculated using the Kaplan‐Meier estimator. Patients who remained without an event by the end of the study period (31 October 2019 or 2020) were censored. Cox proportional hazards regression analyses were used to compute crude and adjusted for age and stage hazard ratios (HRs) for death by any cause with the corresponding 95% confidence intervals (95% CIs), whenever the number of events was at least four. The time‐scale used in the Cox regression model was survival time. Additional Cox proportional hazards regression analyses were carried out using a propensity score regression adjustment (PSRA) to estimate the effect of the COVID‐19 pandemic on mortality. A logistic regression model was used to estimate propensity scores in which a cancer diagnosis before or after COVID‐19 was regressed on sex, age, cancer site, stage and symptoms. This approach was taken to reduce the number of covariates into a single score to be included as an adjustment variable, for a more efficient control of confounding. Propensity scores in the two groups were graphed using a histogram to evaluate balance. The proportional hazards assumption was evaluated using Schoenfeld residuals.

Stratified analyses were conducted by sex, age, residence, cancer site, stage, symptoms, referral pathway, first treatment and month of diagnosis.

All analyses were performed using STATA 15 (StataCorp, College Station, TX). Results were considered statistically significant for P‐values less than .05 (two‐sided).

RESULTS

The sociodemographic and clinical characteristics of the patients diagnosed with cancer before and after COVID‐19 are presented in Figure 1. There was a greater proportion of females diagnosed in 2020 compared to 2019 (P = .028). Differences in the distribution of cancer cases were observed (P = .001); in particular, fewer cervical and prostate cancers were diagnosed, while more pancreas and lung cancers were identified. After COVID‐19, cases were more often diagnosed at more advanced stages (P < .001) and more often symptomatic (P < .001). Additionally, after COVID‐19, patients were less often referred from another hospital or through cancer screening but more often referred by a doctor or after an appointment at IPO‐Porto (P < .001). A total of nine COVID‐19 diagnoses were confirmed among patients diagnosed with cancer in 2020, with three deaths being observed, until 31 October 2020.

FIGURE 1

Sociodemographic and clinical characteristics of patients diagnosed with cancer before (n = 1309) and after (n = 763) the onset of COVID‐19 (2 March to 1 July 2019 and 2020, respectively). aPorto Metropolitan Area includes the following municipalities: Arouca, Espinho, Gondomar, Maia, Matosinhos, Oliveira de Azeméis, Paredes, Porto, Póvoa de Varzim, Santa Maria da Feira, Santo Tirso, São João da Madeira, Trofa, Vale de Cambra, Valongo, Vila Nova de Gaia, Vila do Conde. bInternational Statistical Classification of Diseases and Related Health Problems 10th Revision : esophagus, C15; stomach, C16; colon and rectum, C18‐20; pancreas, C25; lung, C34; skin‐melanoma, C43; breast, C50; cervix, C53; prostate, C61 [Color figure can be viewed at wileyonlinelibrary.com]

Figure 2 presents the survival of patients with cancer according to period of diagnosis. During follow‐up to 31 October, 154 (11.8%) and 131 (17.2%) deaths were observed among patients diagnosed before and after COVID‐19, respectively, yielding crude and PSRA HR (95% CI) of 1.51 (1.20‐1.91) and 1.10 (0.86‐1.40), correspondingly. Among patients for whom information on cause of death was available (115 before and 87 after COVID‐19), 97.4% and 98.8% died due to cancer, respectively. Significantly higher hazards were observed among males (HR = 1.66, 95% CI: 1.25‐2.19), and patients aged between 65 and 74 years (HR = 1.86, 95% CI: 1.23‐2.82; Table 1). Stratified analyses by cancer characteristics showed a higher mortality among patients with stomach cancer (HR = 1.96, 95% CI: 1.13‐3.34; Table 2; Supplementary Figure 1), and those diagnosed with Stage III cancer (HR = 2.55, 95% CI: 1.24‐5.22; Table 2; Supplementary Figure 2). No differences in mortality were observed when stratifying results by symptoms at diagnosis (Table 2; Supplementary Figure 3), while patients who were referred from another hospital had a higher mortality (HR = 1.90, 95% CI: 1.35‐2.67; Table 2; Supplementary Figure 4). Patients who received any treatment had a higher hazard of mortality (HR = 1.86, 95% CI: 1.26‐2.74; Table 3; Supplementary Figure 5). This was particularly observed among patients who underwent surgery (HR = 4.79, 95% CI: 1.44‐15.93) or received radiotherapy (HR = 2.53, 95% CI: 1.34‐4.79), which remained statistically significant after PSRA (HR = 3.97, 95% CI: 1.14‐13.77 and HR = 1.96, 95% CI: 1.03‐3.74, respectively). Patients who did not receive any treatment had lower hazards of mortality (HR = 0.70, 95% CI: 0.50‐0.90; Table 3; Supplementary Figure 5), which remained statistically significant after PSRA (HR = 0.62, 95% CI: 0.46‐0.83). Additionally, stratified analyses by month of diagnosis yielded a higher mortality among patients diagnosed in March (HR = 1.67, 95% CI: 1.06‐2.63) and May (HR = 2.04, 95% CI: 1.26‐3.31; Supplementary Table 1; Supplementary Figure 6).

FIGURE 2

Survival (calculated using the Kaplan‐Meier estimator) among patients diagnosed with cancer before (n = 1309, 154 events) and after (n = 763, 131 events) the onset of COVID‐19 (2 March to 1 July 2019 and 2020, respectively) with follow‐up to 31 October 2019 or 2020, respectively. COVID‐19, Coronavirus disease 2019 [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 1

Crude and adjusted hazard ratios and 95% confidence intervals for death before and after the onset of COVID‐19 (2 March to 1 July 2019, and 2 March to 1 July 2020 with follow‐up to 31 October 2019 or 2020, respectively) calculated using Cox regression, according to sociodemographic characteristics

		Total (N)	Proportional variation (%)	Event [N (%)]	Crude HR (95% CI)	Age‐adjusted HR (95% CI) a	Age‐ and stage‐adjusted HR (95% CI) b	Propensity score adjusted HR (95% CI) c
All d	Before	1309		154 (11.8)	1	1	1	1
	After	763	−41.7	131 (17.2)	1.51 (1.20‐1.91)	1.49 (1.18.1.89)	1.15 (0.91‐1.46)	1.10 (0.86‐1.40)
Sex	Males
	Before	691		108 (15.6)	1	1	1	1
	After	364	−47.3	90 (24.7)	1.66 (1.25‐2.19)	1.64 (1.24‐2.18)	1.28 (0.96‐1.71)	1.13 (0.84‐1.53)
	Females
	Before	618		46 (7.4)	1	1	1	1
	After	399	−35.4	41 (10.3)	1.43 (0.94‐2.18)	1.40 (0.92‐2.13)	0.95 (0.62‐1.45)	0.96 (0.62‐1.50)
Age (years)	<55
	Before	309		21 (6.8)	1	1	1	1
	After	191	−38.2	15 (7.8)	1.19 (0.61‐2.30)	1.24 (0.64‐2.41)	0.80 (0.41‐1.57)	0.89 (0.45‐1.77)
	55‐64
	Before	346		40 (11.6)	1	1	1	1
	After	177	−48.8	27 (15.3)	1.41 (0.86‐2.30)	1.41 (0.86‐2.31)	1.07 (0.64‐1.75)	1.07 (0.64‐1.79)
	65‐74
	Before	396		44 (11.1)	1	1	1	1
	After	223	−43.7	46 (20.6)	1.86 (1.23‐2.82)	1.87 (1.23‐2.83)	1.38 (0.90‐2.10)	1.20 (0.77‐1.86)
	>74
	Before	258		49 (19.0)	1	1	1	1
	After	172	−33.3	43 (25.0)	1.37 (0.91‐2.07)	1.37 (0.91‐2.07)	1.13 (0.74‐1.72)	1.08 (0.70‐1.65)
Residence e	Porto Metropolitan Area
	Before	666		91 (13.7)	1	1	1	1
	After	399	−40.1	73 (18.3)	1.38 (1.01‐1.88)	1.37 (1.01‐1.87)	0.98 (0.71‐1.34)	0.95 (0.69‐1.32)
	Outside the Porto Metropolitan Area
	Before	643		63 (9.8)	1	1	1	1
	After	364	−43.4	58 (15.9)	1.70 (1.19‐2.43)	1.68 (1.17‐2.40)	1.43 (1.00‐2.06)	1.32 (0.92‐1.92)

Note: Significant associations are bolded.

Abbreviations: 95% CI, 95% confidence interval; COVID‐19, Coronavirus disease 2019; HR, hazard ratio.

Age (continuous).

Age (continuous) and stage (I, II, III, IV, missing).

Propensity score calculated using a logistic regression model with cancer diagnosis before or after COVID‐19 as the dependent variable, and sex, age (continuous), cancer site (esophagus, C15; stomach, C16; colon and rectum, C18‐20; pancreas, C25; lung, C34; skin‐melanoma, C43; breast, C50; cervix, C53; prostate, C61), stage (I, II, III, IV, missing) and symptoms (asymptomatic, symptomatic, missing) as independent variables.

Among patients with information on cause of death (115 before and 87 after COVID‐19), 97.4% and 98.8% died due to cancer.

Porto Metropolitan Area includes the following municipalities: Arouca, Espinho, Gondomar, Maia, Matosinhos, Oliveira de Azeméis, Paredes, Porto, Póvoa de Varzim, Santa Maria da Feira, Santo Tirso, São João da Madeira, Trofa, Vale de Cambra, Valongo, Vila Nova de Gaia, Vila do Conde.

TABLE 2

Crude and adjusted hazard ratios and 95% confidence intervals for death before and after the onset of COVID‐19 (2 March to 1 July 2019, and 2 March to 1 July 2020 with follow‐up to 31 October 2019 or 2020, respectively) calculated using Cox regression, according to cancer characteristics and referral pathway

		Total (N)	Proportional variation (%)	Event [N (%)]	Crude HR (95% CI)	Age‐adjusted HR (95% CI) a	Age‐ and stage‐adjusted HR (95% CI) b	Propensity score adjusted HR (95% CI) c
All	Before	1309		154 (11.8)	1		1	1
	After	763	−41.7	131 (17.2)	1.51 (1.20‐1.91)	1.49 (1.18.1.89)	1.15 (0.91‐1.46)	1.10 (0.86‐1.40)
Cancer sited	Esophagus
	Before	26		7 (26.9)	1	1	1	1
	After	18	−30.8	7 (38.9)	1.59 (0.56‐4.54)	1.51 (0.51‐4.46)	1.31 (0.43‐3.97)	1.37 (0.45‐4.14)
	Stomach
	Before	147		27 (18.4)	1	1	1	1
	After	81	−44.9	25 (30.9)	1.96 (1.13‐3.34)	1.84 (1.06‐3.20)	1.17 (0.66‐2.06)	1.27 (0.70‐2.33)
	Colon and rectum
	Before	180		17 (9.4)	1	1	1	1
	After	113	−37.2	17 (15.0)	1.71 (0.87‐3.35)	1.72 (0.88‐3.37)	1.77 (0.88‐3.57)	1.50 (0.75‐3.01)
	Pancreas
	Before	41		20 (48.8)	1	1	1	1
	After	37	−9.8	15 (40.5)	0.76 (0.39‐1.49)	0.70 (0.36‐1.37)	0.75 (0.38‐1.48)	0.71 (0.34‐1.41)
	Lung
	Before	204		73 (35.8)	1	1	1	1
	After	164	−19.6	58 (35.4)	0.96 (0.67‐1.35)	0.95 (0.67‐1.35)	0.88 (0.62‐1.25)	0.89 (0.63‐1.26)
	Skin‐melanoma
	Before	80		5 (6.2)	1	1	1	1
	After	56	−30.0	4 (7.1)	1.20 (0.32‐4.48)	1.21 (0.32‐4.52)	1.11 (0.25‐4.84)	0.63 (0.12‐3.16)
Stage	I‐II
	Before	723		6 (0.8)	1	1	—	1
	After	359	−50.3	6 (1.7)	2.05 (0.66‐6.35)	2.03 (0.65‐6.30)	—	1.60 (0.48‐5.28)
	III
	Before	245		14 (5.7)	1	1	—	1
	After	113	−53.9	16 (14.2)	2.55 (1.24‐5.22)	2.55 (1.25‐5.24)	—	2.37 (1.14‐4.94)
	IV
	Before	296		122 (41.2)	1	1	—	1
	After	217	−26.7	94 (43.3)	1.04 (0.80‐1.37)	1.05 (0.80‐1.38)	—	0.97 (0.74‐1.28)
Symptoms	Asymptomatic
	Before	502		12 (2.4)	1	1	1	1
	After	208	−58.6	9 (4.3)	1.84 (0.78‐4.37)	1.78 (0.75‐4.22)	1.53 (0.62‐3.78)	1.35 (0.55‐3.33)
	Symptomatic
	Before	705		129 (18.3)	1	1	1	1
	After	501	−28.9	107 (21.4)	1.20 (0.93‐1.56)	1.19 (0.92‐1.54)	0.97 (0.75‐1.26)	1.03 (0.79‐1.34)
Referral pathway	Doctor
	Before	674		75 (11.1)	1	1	1	1
	After	442	−34.4	61 (13.8)	1.27 (0.91‐1.78)	1.27 (0.91‐1.78)	1.05 (0.75‐1.48)	1.02 (0.72‐1.45)
	Another hospital
	Before	434		73 (16.8)	1	1	1	1
	After	208	−52.1	63 (30.3)	1.90 (1.35‐2.67)	1.89 (1.35‐2.66)	1.32 (0.93‐1.87)	1.23 (0.85‐1.76)
	Appointment at IPO‐Porto
	Before	28		5 (17.9)	1	1	1	1
	After	49	+75.0	4 (8.2)	0.43 (0.12‐1.61)	0.43 (0.11‐1.60)	0.56 (0.14‐2.30)	0.36 (0.09‐1.39)

Notes: Significant associations are bolded. Number of deaths and hazards ratios for patients with breast, C50; cervix, C53; prostate, C61 are not shown as fewer than four events were observed during follow‐up. Number of deaths and hazards ratios for patients identified by organized or opportunistic screening are not shown as no deaths were observed during follow‐up.

Abbreviations: 95% CI, 95% confidence interval; COVID‐19, Coronavirus disease 2019; HR, hazard ratio.

Age (continuous).

Age (continuous) and stage (I, II, III, IV, missing).

Propensity score calculated using a logistic regression model with cancer diagnosis before or after COVID‐19 as the dependent variable, and sex, age (continuous), cancer site (esophagus, C15; stomach, C16; colon and rectum, C18‐20; pancreas, C25; lung, C34; skin‐melanoma, C43), stage (I, II, III, IV, missing) and symptoms (asymptomatic, symptomatic, missing) as independent variables.

International Statistical Classification of Diseases and Related Health Problems 10th Revision : esophagus, C15; stomach, C16; colon and rectum, C18‐20; pancreas, C25; lung, C34; skin‐melanoma, C43; breast, C50; cervix, C53; prostate, C61.

TABLE 3

Crude and adjusted hazard ratios and 95% confidence intervals for death before and after the onset of COVID‐19 (2 March to 1 July 2019, and 2 March to 1 July 2020 with follow‐up to 31 October 2019 or 2020, respectively) calculated using Cox regression, according to first treatment received

		Total (N)	Proportional variation (%)	Event [N (%)]	Crude HR (95% CI)	Age‐adjusted HR (95% CI) a	Age‐ and stage‐adjusted HR (95% CI) b	Propensity score adjusted HR (95% CI) c
All	Before	1309		154 (11.8)	1		1	1
	After	763	−41.7	131 (17.2)	1.51 (1.20‐1.91)	1.49 (1.18.1.89)	1.15 (0.91‐1.46)	1.10 (0.86‐1.40)
First treatmentd	Surgery
	Before	499		4 (0.80)	1	1	1	1
	After	210	−57.9	8 (3.8)	4.79 (1.44‐15.93)	4.79 (1.44‐15.91)	4.29 (1.22‐15.02)	3.97 (1.14‐13.77)
	Radiotherapy
	Before	202		26 (12.9)	1	1	1	1
	After	49	−75.7	15 (30.6)	2.53 (1.34‐4.79)	2.56 (1.35‐4.85)	1.61 (0.84‐3.09)	1.96 (1.03‐3.74)
	Systemic treatmente
	Before	346		28 (8.1)	1	1	1	1
	After	199	−42.5	23 (11.6)	1.46 (0.84‐2.53)	1.37 (0.79‐2.38)	1.12 (0.64‐1.95)	1.16 (0.66‐2.05)
	Chemotherapy
	Before	212		23 (10.8)	1	1	1	1
	After	121	−42.9	17 (14.0)	1.32 (0.70‐2.47)	1.24 (0.66‐2.33)	1.11 (0.59‐2.08)	1.12 (0.59‐2.15)
	Any (as of 1 July)
	Before	1057		58 (5.5)	1	1	1	1
	After	453	−57.1	46 (10.1)	1.86 (1.26‐2.74)	1.85 (1.26‐2.73)	1.27 (0.86‐1.88)	1.27 (0.85‐1.91)
	None (as of 1 July)f
	Before	252		96 (38.1)	1	1	1	1
	After	309	+22.6	85 (27.5)	0.70 (0.50‐0.90)	0.69 (0.52‐0.93)	0.84 (0.63‐1.13)	0.62 (0.46‐0.83)

Notes: Significant associations are bolded. Number of deaths and hazards ratios for patients who received brachytherapy; chemoradiotheray; immunotherapy or targeted therapy; hormone therapy are not shown as fewer than four events were observed during follow‐up.

Abbreviations: 95% CI, 95% confidence interval; COVID‐19, Coronavirus disease 2019; HR, hazard ratio.

Age (continuous).

Age (continuous) and stage (I, II, III, IV, missing).

Propensity score calculated using a logistic regression model with cancer diagnosis before or after COVID‐19 as the dependent variable, and sex, age (continuous), cancer site (esophagus, C15; stomach, C16; colon and rectum, C18‐20; pancreas, C25; lung, C34; skin‐melanoma, C43; breast, C50; cervix, C53; prostate, C61), stage (I, II, III, IV, missing) and symptoms (asymptomatic, symptomatic, missing) as independent variables.

Four patients received another treatment.

Systemic therapy includes chemotherapy, chemoradiotherapy, immunotherapy, targeted therapy and hormone therapy.

Reasons for receiving no treatment during the follow‐up period: 182 (57 events) and 266 (52 events) awaiting treatment to begin, 22 (18 events) and 17 (14 events) with cancer too advanced for treatment, 24 (21 events) and 22 (19 events) patients who cannot undergo treatment due to current physical condition, 24 (0 events) and 4 (0 events) patients currently under surveillance before and after COVID‐19, respectively.

Considering patients diagnosed in February 2019 and 2020, there was a greater proportion of females diagnosed in 2020 compared to 2019 (P = .021; Supplementary Figure 7). A total of 36 (12.2%) and 36 (13.4%) deaths were observed among those diagnosed in 2019 and 2020, respectively, corresponding to an HR (95% CI) of 1.09 (0.69‐1.74; Supplementary Table 2 and Supplementary Figure 8). No statistically significant differences were observed.

DISCUSSION

The current study shows an increase in the short‐term mortality among patients diagnosed with cancer after the COVID‐19 outbreak compared to the same period in the previous year. This was particularly evident among males, those aged between 65 and 74 years, patients with stomach cancer or Stage III at diagnosis, patients referred from another hospital, as well as those undergoing surgery or radiotherapy until 2 July of the respective year of diagnosis. After PSRA, statistically significant increased hazards continued to be observed for patients diagnosed with Stage III cancer and those undergoing surgical treatment or radiotherapy, while patients who did not receive any treatment during follow‐up had lower mortality hazards.

The overall increased hazards of death among patients diagnosed after COVID‐19 may be explained both by delays in cancer diagnosis as well as changes in the characteristics of patients with cancer regarding their prognosis, both due to adjustments in health‐care settings and in the management of patients with cancer during the pandemic, in addition to bias and residual confounding. In the current study and as previously described in Northern Portugal, a lower proportion of cervical and prostate cancers, and a higher proportion of pancreas and lung cancers were identified since the first confirmed COVID‐19 case in Portugal on 2 March compared to the same period in the previous year. A substantial impact of the pandemic on the number of cancer diagnoses as well as changes in the distribution of cancer sites and patients' characteristics has also been observed in other settings. In the United States of America and in the Netherlands, larger decreases have been observed for cancers of the breast, colorectal, skin cancers (excluding basal cell carcinoma) and hematological cancers. , These variations are influenced by the reduction in the number of asymptomatic cases, often identified through opportunistic or organized screening, , at earlier stages, and as such with potentially better prognosis, as well as by the proportional increase of cancers with a usually poorer survival (pancreas and lung cancers). These changes occurred as a result of patient, doctor and health‐care factors. In particular, IPO‐Porto had to adapt their activity to ensure necessary services while reducing the risk of infection by COVID‐19 among both patients and health‐care professionals. Moreover, many medical appointments, namely cancer screening, were canceled, postponed or replaced by telehealth in many health‐care settings due to the COVID‐19 pandemic. Additionally, patients themselves may have delayed routine health procedures or the assessment of mild symptoms, due to fear of visiting health‐care providers that were also handling suspected COVID‐19 cases. Nevertheless, patients with well recognized symptoms will be more likely to visit health‐care services, while vague cancer symptoms may be dismissed by patients. In line with these observations, we also found that cases were more often symptomatic, and referred to IPO‐Porto by a doctor or after an appointment at IPO‐Porto, which may have led to a delayed diagnosis. Additionally, although patients were less often referred through cancer screening or from another hospital, after COVID‐19 the latter had a higher mortality.

The decrease in the overall mortality hazard estimates after adjusting for potential confounders that was observed in the present analysis supports the hypothesis that the pre‐post pandemic differences are largely explained by changes in the characteristics of the cases being diagnosed in each period. The significant mortality hazards that persist after adjustment for various prognosis indicators may reflect residual or uncontrolled differences in the survival of patients diagnosed in the two periods. This may be due to the fact that the marked decrease in the absolute number of cancer cases is likely to translate into a larger proportion of cases with a worse prognosis in 2020, even within each apparently homogeneous stratum. To overcome this, we also used a PSRA approach to consider several variables that have been described to differ between patients diagnosed with cancer before and after COVID‐19, and probably contributed for the observed differences in mortality. Nevertheless, after the inclusion of the PSRA, patients with Stage III cancer and those undergoing surgical treatment or radiotherapy were found to have a significantly higher mortality, while patients who did not receive any treatment had a lower mortality hazards. Apart from that, the available clinical information from each case does not allow us to disentangle the contribution of potential delays in access to care, which are likely to have occurred in at least a subset of the cases included here, or the contribution to the change in the case‐mix of patients.

Treatment disruptions and modifications have been reported as a result of the pandemic, which may have also impacted the prognosis of these patients. In Canada, over half of patients with lung cancer receiving treatment between March and May 2020 underwent at least one change in their cancer treatment plan. In the United States of America, nearly half of patients with breast cancer encountered treatment delays. A study using data from the Public Health England National Cancer Registration Service also found that modest delays in surgery for cancer will impact patient survival. In the current study, we found that patients undergoing surgery or receiving radiotherapy had higher hazards of mortality during the COVID‐19 pandemic. Additionally, we have observed no evidence of delays between diagnosis and first appointment, multidisciplinary tumor board meeting or first treatment in a previous analysis, although a greater proportion of patients diagnosed after COVID‐19 did not receive a first treatment until 2 July compared to before COVID‐19. Nevertheless, we may hypothesize that patients who underwent surgery or radiotherapy as a first cancer treatment after COVID‐19 had a worse prognosis, while patients who did not receive a first treatment during the follow‐period had a better prognosis. In particular, patients awaiting to receive the first treatment were more often female and diagnosed with a cancer with better survival (skin melanoma and breast cancer, and Stage I or II). These variables, as well as age and presence of symptoms, were included in the propensity score adjustment, and higher survival continued to be observed after COVID‐19. Furthermore, we conducted stratified analyses by first treatment received and month of diagnosis (data not shown); however, no consistent variation over time appeared to be observed and much larger samples are needed for assessing changes in how health care was provided throughout the pandemic.

Patients with cancer have been shown to be at an increased risk of severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) infection, as well as more likely to present a COVID‐19 phenotype characterized by more severe disease and increased risk of death, although with differences according to cancer type. In our study, the number of patients who had a SARS‐CoV‐2 infection diagnosis was relatively small (nine confirmed), and therefore the occurrence of COVID‐19 among the few patients in our cohorts could not meaningfully account for the nearly 50% higher death hazard in patients diagnosed in 2020. Furthermore, most of the deaths observed in our study were due to cancer, and there were only three deaths among the patients with a confirmed COVID‐19 diagnosis.

Strengths and limitations

The current study compared two cohorts of patients diagnosed with cancer over a 4‐month period before and after COVID‐19 to evaluate the impact of the pandemic on short‐term mortality. We also included a comparison between patients diagnosed in February, periods without COVID‐19 cases in Portugal, and this “negative control” supports the validity of our findings.

We opted to exclude patients who received a first treatment outside IPO‐Porto from our analyses, to ensure the completeness of data collected as well as the comparability between the two periods, since there is a usual lag between the incidence of cancer cases and their registration. Cancer cases diagnosed in 2019 and treated elsewhere are more likely to have potentially received additional care at IPO‐Porto until the present than those diagnosed in 2020, and as such included in the cancer registry. We have previously analyzed two interrupted time series to compare the variation in the number of incident cancer cases diagnosed before COVID‐19 and after, as well as the comparison of data from the month of February, and found that the observed decrease in the number of cancer cases after 2 March 2020 was unlikely to be due to delayed registration. Additionally, the collection of data for the current study occurred between 1 May and 31 August 2020 to ensure that clinical and treatment information was as complete as possible, and vital status was obtained through manual linkage with the National Health Service database, which is continuously updated. We had few missing data overall, with the exception of cause of death, which was only used to describe the cause of death and not included in further analyses. Therefore, although the current study was implemented quickly in response to the COVID‐19 pandemic, we believe that there are no significant differences in the completeness of data, or that this would affect the results of the present study in a meaningful manner.

Detailed patient‐level data were obtained, which allowed us to understand the impact of the COVID‐19 pandemic on the short‐term mortality of patients with cancer. However, uncontrolled differences in the prognosis of patients diagnosed in the two periods may contribute for some of the differences observed. To partially overcome this, we also conducted a PSRA analysis and some hazards remained statistically significant. Additionally, a more comprehensive assessment of the potential effects of the pandemic on cancer survival will require a longer follow‐up, and a continued monitoring of the survival among patients diagnosed over the next months. This may include patients whose diagnosis or access to treatment may have been delayed, possibly to a different extent as the pandemic evolves.

Our study includes data from a single‐center located in Northern Portugal, which may impair the generalizability of our results to other settings. Nonetheless, IPO‐Porto is one of the largest cancer‐dedicated hospitals in Portugal, receiving patients from any part of the country, with different sociodemographic backgrounds and our study includes patients presenting a large spectrum of cancer sites.

CONCLUSION

The results presented here describe the impact of the COVID‐19 outbreak on the short‐term mortality of patients with cancer at an oncology center in Northern Portugal. The higher overall mortality observed should be cautiously interpreted due not only to the differing case‐mix of patients diagnosed with cancer during the COVID‐19 pandemic, but also the impact of government, health‐care services and patient responses. The first months of the COVID‐19 pandemic led to changes in the timely diagnosis, treatment and follow‐up of cancer cases, which will inevitably impact the prognosis of patients with cancer. Additionally, considering the backlog of cancer cases likely presenting with late, nonoperable disease needing assessment along with the expected volume of new cancer cases and the ongoing reallocation of resources, health‐care systems will require a strategic prioritization of patients to mitigate deaths attributable to the COVID‐19 pandemic.

CONFLICT OF INTEREST

The authors declared no potential conflicts of interest.

ETHICS STATEMENT

The data contained no unique personal identifiers and were extracted from the cancer registry database and clinical files. Hence, patient written informed consent was waived. The study was approved by the Ethics Committee of the Portuguese Institute of Oncology of Porto (Ref. CES IPO: 164/020).

Supplementary Table 1 Crude and adjusted hazard ratios and 95% confidence intervals for death before and after the onset of COVID‐19 (2 March to July 1, 2019, and 2 March to July 1, 2020 with follow‐up to October 31, 2019 or 2020, respectively) calculated using Cox regression, according to month of cancer diagnosis.

Supplementary Table 2. Crude and adjusted hazard ratios and 95% confidence intervals for mortality among patients diagnosed with cancer in February 2019 and February 2020 with follow‐up to October 31, 2019 or 2020, respectively, calculated using Cox regression, according to sociodemographic and cancer characteristics.

Supplementary Figure 1. Survival (calculated using the Kaplan‐Meier estimator) among patients with cancer, according to period of diagnosis (2 March to July 1, 2019 vs 2 March to July 1, 2020) by cancer site1 with follow‐up to October 31, 2019 or 2020, as applicable.

Supplementary Figure 2. Survival (calculated using the Kaplan‐Meier estimator) among patients with cancer, according to period of diagnosis (2 March to July 1, 2019 vs 2 March to July 1, 2020) by stage at diagnosis, with follow‐up to October 31, 2019 or 2020, as applicable.

Supplementary Figure 3. Survival (calculated using the Kaplan‐Meier estimator) among patients with cancer, according to period of diagnosis (2 March to July 1, 2019 vs 2 March to July 1, 2020) by symptoms at diagnosis (until July 2, 2019 or 2020, as applicable), with follow‐up to October 31, 2019 or 2020, as applicable.

Supplementary Figure 4. Survival (calculated using the Kaplan‐Meier estimator) among patients with cancer, according to period of diagnosis (2 March to July 1, 2019 vs 2 March to July 1, 2020) by referral pathway (until July 2, 2019 or 2020, as applicable), with follow‐up to October 31, 2019 or 2020, as applicable.

Supplementary Figure 5. Survival (calculated using the Kaplan‐Meier estimator) among patients with cancer, according to period of diagnosis (2 March to July 1, 2019 vs 2 March to July 1, 2020) by first treatment received (until July 2, 2019 or 2020, as applicable), with follow‐up to October 31, 2019 or 2020, as applicable.

Supplementary Figure 6. Survival (calculated using the Kaplan‐Meier estimator) among patients with cancer, according to period of diagnosis (2 March to July 1, 2019 vs 2 March to July 1, 2020) by month of diagnosis, with follow‐up to October 31, 2019 or 2020, as applicable.

Supplementary Figure 7. Sociodemographic and clinical characteristics of cancer patients diagnosed in February 2019 and 2020.

Supplementary Figure 8. Survival (calculated using the Kaplan‐Meier estimator) among cancer patients diagnosed in February 2019 and 2020 with follow‐up to October 31, 2019 or 2020, as applicable.

Click here for additional data file.