Impact of the COVID‐19 pandemic on the in‐hospital diagnostic pathway of breast and colorectal cancer in the Netherlands: A population‐based study

Abstract Background In the Netherlands, the COVID‐19 pandemic resulted in a temporary halt of population screening for cancer and limited hospital capacity for non‐COVID care. We aimed to investigate the impact of the pandemic on the in‐hospital diagnostic pathway of breast cancer (BC) and colorectal cancer (CRC). Methods 71,159 BC and 48,900 CRC patients were selected from the Netherlands Cancer Registry. Patients, diagnosed between January 2020 and July 2021, were divided into six periods and compared to the average of patients diagnosed in the same periods in 2017–2019. Diagnostic procedures performed were analysed using logistic regression. Lead time of the diagnostic pathway was analysed using Cox regression. Analyses were stratified for cancer type and corrected for age, sex (only CRC), stage and region. Results For BC, less mammograms were performed during the first recovery period in 2020. More PET‐CTs were performed during the first peak, first recovery and third peak period. For CRC, less ultrasounds and more CT scans and MRIs were performed during the first peak. Lead time decreased the most during the first peak by 2 days (BC) and 8 days (CRC). Significantly fewer patients, mainly in lower stages, were diagnosed with BC (−47%) and CRC (−36%) during the first peak. Conclusion Significant impact of the COVID‐19 pandemic was found on the diagnostic pathway, mainly during the first peak. In 2021, care returned to the same standards as before the pandemic. Long‐term effects on patient outcomes are not known yet and will be the subject of future research.


| BACKGROUND
Late 2019, a new infection known as COVID-19 disease was identified in Wuhan, China. 1 The coronavirus spread quickly, became a worldwide problem and impacted cancer care.−4 The infections in the Netherlands started in the south in February 2020.In March 2020, the first measures against the coronavirus were taken in the Netherlands with temporary halt of the population-based screening programs for BC and CRC and societal measures. 5,6eople were advised to visit the general practitioner (GP) only in case of severe complaints and referrals to the hospital were postponed. 5Fewer diagnoses and treatments were performed and surgeries were postponed. 7,8In case of a shortage of capacity, patients were transferred to other hospitals. 6After temporarily halting population screening from mid-March 2020 to mid-May 2020, screening was restarted gradually.Mid 2021, BC screening capacity was 85% and CRC screening had the same capacity in October 2021 as in the years before the pandemic. 9ntil now, studies mainly focused on the first months of the pandemic or focused on treatment, follow-up and the prediction of long-term impact.Few studies focused on the impact of the COVID-19 pandemic on cancer diagnosis.Some studies revealed the impact of the pandemic on CRC 10 or the impact on cancer care in general. 4A study revealed that fewer and delayed referrals possibly influenced the diagnostic pathway in the hospital. 11Another study revealed a postponed start of cancer treatment and patients diagnosed with poorer patient and tumour characteristics during the pandemic. 12o studies are currently published that focus on the inhospital diagnostic pathway.As this may have impact on the treatment and subsequent outcome of these patients, the objective of this study was to determine the impact of the COVID-19 pandemic on the in-hospital diagnostic pathway until start of therapy by analysing the diagnostic procedures performed per patient and the time to start of therapy.This study provides a more complete understanding of cancer care during the pandemic and lessons can be learned to maximise the quality of cancer care during any subsequent pandemic.

| Data collection
This study is a retrospective cohort study, based on data from the Netherlands Cancer Registry (NCR) and Dutch Hospital Data (DHD).The data set consists of data on patient and tumour characteristics from the NCR (e.g.age, sex, type of cancer, stage of disease, region in the Netherlands) and data on the diagnostic procedures from DHD (e.g.diagnostic procedure performed, date of diagnosis, date of first diagnostic procedure, date of start therapy).

| Patients
Patients aged 18 years or older diagnosed with primary BC or CRC during the period from January 2017 to July 2021 were included in this study.Data from DHD and the NCR were probabilistically linked on patient levels using the patient number, date of birth, sex and postal code.Diagnostic procedures more than 6 months prior to the date of diagnosis were removed to avoid including non-cancer-related diagnostic procedures in the analysis.Patients without known hospital diagnostic procedures, an unknown start the date of therapy or a start date of therapy before date of diagnosis were excluded.lower stages, were diagnosed with BC (−47%) and CRC (−36%) during the first peak.

Conclusion:
Significant impact of the COVID-19 pandemic was found on the diagnostic pathway, mainly during the first peak.In 2021, care returned to the same standards as before the pandemic.Long-term effects on patient outcomes are not known yet and will be the subject of future research.

K E Y W O R D S
breast cancer, colorectal cancer, COVID-19, diagnosis, diagnostic pathway, diagnostic procedures, lead time, population based 2.3 | Definitions

| Periods
The study period between January 2020 and July 2021 was divided into six periods based on the severity of the COVID-19 pandemic for which we used hospitalizations due to COVID-19 in the Netherlands as a proxy.Period A covers weeks 1-11 of 2020 (i.e.pre-COVID), period B weeks 12-20 of 2020 (i.e.first peak), period C weeks 21-41 of 2020 (i.e.first recovery), period D weeks 42-53 of 2020 (i.e.second peak), period E weeks 1-20 of 2021 (i.e.third peak) and period F weeks 21-30 of 2021 (i.e.second recovery).For comparison, data for 2017-2019 were divided accordingly.The second and third peaks are divided into two periods to analyse 2020 and 2021 separately.

| Age categories
Patients were categorized into age categories, based on the age categories for population screening for BC and CRC: 50-75 and 55-75, respectively.Therefore, patients with BC were grouped into ages <50, 50-75 and >75 years and patients with CRC were grouped into ages <55, 55-75 and >75 years.

| Sex
For BC, only females were included.For CRC, males and females were included.

| Stage of disease
Stage of disease was divided into stages 0-4 for BC and 1-4 for CRC, based on the TNM classification (8th edition). 13athological stage was used if available.When pathological stage was unknown, the clinical stage was used.Any remaining unknown stage was classified as stage X.

| Region
Patients were categorized into a region, based on the hospital where a patient was diagnosed.In total, the Netherlands was divided into five regions, 14 these are the north (Friesland, Groningen, Drenthe), the middle east (Overijssel, Flevoland), the middle (Utrecht, Gelderland), the west (Noord-Holland, Zuid-Holland, Zeeland) and the south (Noord-Brabant, Limburg).

| Diagnostic procedures
Per type of cancer, different diagnostic procedures were analysed based on common diagnostic procedures per type of cancer. 15Diagnostic procedures were divided into mammography, ultrasound, PET-CT and CT for BC and ultrasound, endoscopy, CT, MRI and PET-CT for CRC.

| Lead time
The lead time of the diagnostic pathway per patient was calculated as the time between the first diagnostic procedure and the start of therapy.When the date of first diagnostic procedure was missing, the date of pathologically confirmed diagnosis was used as the date of first diagnostic procedure.

| Statistical analysis
Periods during COVID-19 were compared to the same periods in 2017 to 2019.BC and CRC were analysed separately.Patient and tumour characteristics (i.e.type of cancer, age, sex, stage of disease, region, period) were described at the time of diagnosis.Patient characteristics of the study population were investigated using the Chisquared test.An unpaired t-test was performed to analyse the number of weekly diagnosed patients per period.Possible confounders were based on available data and literature and included in the regression analysis.For BC, confounding variables in the regression models were age, stage of disease and region.For CRC, confounding variables were age, sex, stage of disease and region.
The lead time of the diagnostic pathway during the COVID-19 period was compared to the same period before the pandemic (2017-2019) and analysed using Cox Proportional Hazards Regression adjusted for possible confounders.During the first peak, population screening was temporarily halted and the number of infections was not equally spread over the Netherlands.To determine possible differences in the magnitude of factors influencing the lead time during the first peak, Cox regression was performed to compare these factors during the first peak in 2020 to 2017-2019.
Data were analysed using Stata version 17.0.A two-sided p-value of <0.05 was considered statistically significant.

| Breast cancer
BC diagnostic procedures performed per patient are shown in Figure 2A.Adjusted for age, stage and region, the percentage of patients who received mammography was significantly lower during the first recovery (from 86% to 84%) and the odds declined (OR = 0.93 [0.87, 0.99] and p = 0.017) and was significantly higher (from 86% to 87%) during the pre-COVID period (OR = 1.14 [1.05, 1.24] and p = 0.001) and second recovery (OR = 1.10 [1.00, 1.20] and p = 0.039).The percentage of patients who received ultrasound was significantly higher during the pre-COVID period (from 92% to 93%, OR = 1.*Significant difference between periods p < 0.05.**Significant difference between periods p < 0.001. to therapy was found for patients in the age group >75 and stage 4 compared to the other categories.For CRC, the results were the opposite.The decrease in lead time for patients aged >75 and patients in stage 4 was smaller compared to other categories.No notable differences in impact were found between regions in terms of the lead time of the diagnostic pathway for both BC (Table 4) and CRC (Table 5).

| DISCUSSION
As a result of the measures taken to control the spread of COVID-19, that is, the temporary halt of the national screening program and the advice to visit the GP only in case of severe complaints, the number of BC and CRC diagnoses decreased.Compared to the years 2017 to 2019, the frequency of hospital diagnostic procedures decreased for both tumour types and there was a shorter time to therapy.
Stage was important to take into account in the model due to the increased proportion of patients diagnosed at a higher stage.An additional forward stepwise regression was performed to reveal the effect of stage in the model.For lead time and several diagnostic procedures, the stage was added to the model as the first and most important variable.The effect of stage on the outcome was strong and therefore included in the final model.
The lower number of BC and CRC diagnoses during the pandemic, mainly in the lower stages, was to be expected due to the halt of the screening program in which low stages are generally detected.Besides, for CRC, a decreasing trend in the number of CRC diagnoses has been present for the past years as a result of the nationwide screening program. 16However, as in other studies, the decrease in number of diagnoses during the first peak of the pandemic was larger compared to the other COVID-19 periods. 8,16here were significant differences in the diagnostic procedures performed per patient.For BC, the percentage of patients who received mammography was significantly lower during the COVID-19 period and the percentage of patients who received PET-CT was significantly higher.The percentage of patients with CRC who received ultrasound was significantly lower during the pandemic and the percentage of patients who received CT and MRI was significantly higher.
The higher proportion of PET-CTs, CTs and MRIs possibly reflects the increased proportion of patients diagnosed at a higher stage, which is associated with more T A B L E 4 Factors influencing the lead times of diagnostic pathway during the period of the first peak for 2017-2019 and 2020 separate for breast cancer (n = 8329 and n = 1472, respectively).

2017-2019 2020
Hazard symptoms.Therefore, different diagnostic procedures were performed compared to preceding periods in which more asymptomatic patients with lower-stage disease were predominant.Stage was important to take into account in the model due to the increased proportion of patients diagnosed at a higher stage.This study showed that lead time of the diagnostic pathway of BC and CRC was significantly shorter during the pandemic, which was consistent with other studies. 10,14,17 shortened time to therapy, probably the result of fewer referred patients with cancer to the hospital and the prioritising of oncologic care, possibly allowed diagnostics to be performed more quickly.The decrease may also be explained by a change from an initial surgical treatment 7 to a hormonal treatment or radiotherapy, which led to an earlier start of therapy.Thirdly, the decrease may be due to the fact that, relatively seen, patients presented with symptoms and a larger tumour burden related to a higher stage, therefore, diagnosis is easier and treatment can start earlier.
Differences in age categories and stage of disease were found.For BC, the largest decrease in time to therapy was seen among the elderly or a higher stage.For CRC, the time to therapy for the elderly or a higher stage was less decreased in comparison with other patients.This may be due to the fact that co-morbid diseases in elderly people have more impact on the treatment options in patients diagnosed with CRC than in patients diagnosed with BC.
A limitation of this study was that the percentage of patients with BC and CRC who received a biopsy and the percentage of patients with BC who received an MRI were not completely available and therefore these diagnostic procedures could not be included in the current analysis.However, this does not affect the results of this study.

| CONCLUSIONS
In conclusion, the COVID-19 pandemic significantly impacted the diagnostic pathway of patients both with BC and CRC.The impact was mainly observed in 2020, particularly during the first peak of COVID infections.There was a drop in number of diagnoses resulting from the temporary halt of population screening.The percentage of patients who received diagnostic procedures T A B L E 5 Factors influencing the lead times of diagnostic pathway during the period of the first peak for 2017-2019 and 2020 separate for colorectal cancer (n = 5885 and n = 1252, respectively).
for early stage tumours decreased (i.e. less mammography for BC, less endoscopy for CRC), and a shortened time to therapy was observed possibly related to the alterations in first therapy.Diagnostics regarding BC and CRC in 2021 were comparable with the pre-COVID period (period A), which means that care returned to the same standards.The long-term effects of these findings on patient outcomes are not known yet and this will be the subject of future research.

T A B L E 1
Patient characteristics of breast cancer patients divided per period (n = 71,159).

3
Time to therapy of diagnostic pathway corrected for confounders: result of the Cox regression analysis.
The median time of the diagnostic pathway of BC is shown in Figure3.The hazard ratios and p-values are shown in Table3.For 21,463 BC patients, the date of diagnosis is used as the date of first diagnostic procedure.After adjustment for age, stage and region, the lead time of the diagnostic pathway of BC significantly decreased during the first peak with 2 days (HR = 1.21 [1.14, 1.28]), first recovery with 1 day (HR = 1.08 [1.04, 1.11]) and second peak with 1 day (HR = 1.14[1.10,1.19]).The median lead time significantly increased during the pre-COVID period with 2 days (HR = 0.83 [0.80, 0.86]) and during the third peak and second recovery with 1 day (both HR = 0.93 [0.89, 0.97]).3.3.2 | Colorectal cancerThe median time of the diagnostic pathway of CRC is shown in Figure3.The hazard ratios and p-values are shown in Table3.For 19,841 CRC patients, the date of diagnosis is used as the date of the first diagnostic procedure.After adjustment for age, sex, stage and region, the lead time of the diagnostic pathway of CRC significantly decreased during the first peak with 8 days (HR = 1.43 [1.35, 1.53]), first recovery with 2 days (HR = 1.11 [1.07, 1.15]) and second peak with 2 days (HR = 1.11 [1.06, 1.16]).