A comprehensive assessment of intervals in care pathways among cancer patients in a limited resourced setting: results of a prospective patterns of care study in 2022 in Nepal.

Introduction
Access to early detection facilities coupled with access to comprehensive treatment without delay can significantly improve cancer survival and help achieve the United Nations Sustainable Development Goal, target 3.4 to reduce premature mortality from cancer.1 Strengthening the cancer care pathways is key to attaining Universal Health Coverage (to prompt early diagnosis and accessible treatment for cancer),2 3 and remaining aligned to WHO initiatives on breast cancer4 and cervical cancer.5 Delays in diagnosis and treatment accessibility in the majority of cancer patients are the main attributable factor for high cancer mortality in low- and middle-income countries (LMICs).3 6 Longer time to diagnosis and treatment initiation may be detrimental in several ways: requirement of more aggressive management, poorer survival, greater disease-related and treatment-related morbidity, higher healthcare expenditure and poor quality of life.
Delay in cancer care may be ascribed to prolongation of different measurable intervals across the care pathway: ‘prediagnostic interval’ (from onset of symptoms to their first presentation); ‘diagnostic interval’ (from first presentation at primary care to diagnosis confirmation) and ‘treatment interval’ (from being diagnosed to initiation of cancer treatment).7 Prolongation of the prediagnostic interval is known as ‘access delay’ and the same for the diagnostic interval is known as ‘systems delay’. Measuring the delays and understanding the impact of such delays in cancer diagnosis and treatment are essential to designing equitable and affordable cancer diagnosis and care, especially in resource-constrained settings.8
Patterns of care (POC) studies aim to evaluate the compliance of oncology practices to the recommended diagnostic workup and management, and identify patient, provider and system level factors associated with satisfactory utilisation of cancer care. One of the key focuses of POC studies is measuring intervals in the cancer care pathway and the impact of delays on disease outcomes. For our LMIC-focused POC study we selected Nepal, a lower-middle income country in South-east Asia that registered 22 008 new cancer cases, and 14 704 cancer deaths in 2022.1 Many cancer patients abandon investigations and treatment due to high out-of-pocket expenditure as a meagre fraction of expenses is supported by the government.9 A few studies have measured delays in diagnosis and treatment of cervical, gastrointestinal and lung cancer in Nepal;1012 however, a comprehensive assessment ascertaining their impact on stage and survival has not been done.
The present manuscript based on our ongoing POC study in Nepal reports the delays that occur at different phases of diagnosis pathway and assesses its impact on stage at diagnosis across the six common cancers (breast, lung, cervical, stomach, colorectal and oral cancer) at the largest oncology centre in Nepal. Survival data is being collected for future reporting.

Methods
Ethical clearance for the study was obtained from the institutional review board of B.P. Koirala Memorial Cancer Hospital (BPKMCH), national ethics committee at the Nepal Health Research Council (under the Ministry of Health and Population, Nepal) and the institutional ethics committee of International Agency for Research on Cancer / World Health Organization (IARC/WHO).

Study design
An ongoing prospective cohort study (acronym DECAN) recruited adult cancer patients registered at a comprehensive public oncology centre in Nepal. The study design consisted of a baseline face-to-face interview of patients registered at the oncology centre during a period of 6 months (from June to November 2022), followed by abstraction of clinical data, diagnostic test outcomes and treatment details from the medical records on a regular basis.

Study setting
This study was conducted at the BPKMCH hospital, which is located 160 km south-west of the capital (Kathmandu) and is the largest public hospital in the country providing comprehensive cancer care.13 This is among five comprehensive oncology centres in the public sector in Nepal. The 450-bedded hospital provides a wide range of diagnostic, curative and palliative services. In the study year (2022), the hospital registered 6446 new cancer patients referred from all parts of Nepal.

Participant selection
Study population included consecutive patients newly registered at BPKMCH (diagnosed either prior to or after registration at the hospital) with a diagnosis of six most common adult cancers in Nepal, namely cancers of breast, cervical, colorectal, lung, oral and stomach during the recruitment period of 6 months. Patients eligible for inclusion included those who had a confirmed diagnosis of any of the above six cancers. Patients diagnosed outside of BPKMCH and/or receiving partial treatment at other hospitals were also included. Patients diagnosed 1 year earlier at the time of recruitment and those registered at BPKMCH with recurrent cancer were excluded. Those who did not provide informed consent or were unable to do so due to very poor general condition or other medical reasons (eg, cognitive or social disorder) were also excluded from participation.

Cohort enrolment and baseline interview
Patients fitting the inclusion criteria were approached for participation in the study. After a trained research associate provided the study information and explained the research purpose, procedures, potential risks and benefits, patients could voluntarily choose to participate or not. Those who agreed to participate signed an informed consent.
Research associates used a semistructured questionnaire to obtain information at baseline on sociodemographic characteristics, date of first recognition of symptoms, date of first visit to any health facility with the symptoms, advice given at the facility, date of undergoing diagnostic investigations, any treatment received outside and date of registration at BPKMCH. Whenever possible, the dates were corroborated with any documents available with the patient or his/her accompanying persons or in the medical records. Pretesting of the semistructured questionnaires for every cancer type was conducted in a representative sample of cancer patients by the research associates prior to the start of the study. Interviews with cancer patients were conducted at the out-patient or in-patient departments of the hospital depending on the logistic convenience. Patients diagnosed with the type of cancer included in the study were verified using the International Classification of Disease 10th Revision as follows: breast cancer (C50), cervical cancer (C53), colorectal cancer (C18-20), lung cancer (C33-34), oral cancer (C00-06), stomach cancer (C16).14 The number of patients to be recruited was decided empirically based on average number of new registrations at BPKMCH per month.

Clinical and histopathological information
Patient’s medical records were used to abstract all clinical and pathological information in a data collection form. Clinical information included any recorded symptoms information, diagnostic procedures such as imaging or laboratory tests. Pathological information was obtained from histopathology, cytopathology and immunohistochemistry reports. Any missing information was obtained from the respective departments of the hospital. Confirmatory diagnosis of cancer was primarily based on the histopathology report. In the absence of a histopathology report, cytology diagnosis or diagnosis based on clinical evaluation supported by radiology and/or endoscopy (usually for too advanced cancers) at BPKMCH was also considered. American Joint Cancer Committee staging system based on size of tumour (T), nodal status (N) and presence or absence of metastasis (M) was used for all cancers except for cervical cancer, for which International Federation of Gynaecology & Obstetrics staging based on clinical, radiological and pathological findings used was followed.15 16

Data management
Baseline information collected at the interview was recorded in the paper-based questionnaire, which was then uploaded to Research Electronic Data Capture database (REDCap) hosted at the IARC/WHO. Patient’s hospital identity number was replaced by a unique serial number created in the REDCap. Patient’s contact details (name, address and phone number) were only collected in papers for the purpose of follow-up telephone calls. Filled out questionnaires and data collection forms were checked for completeness and quality by an investigator before entering into the REDCap software. Regular quality checks of the uploaded data were made by the project investigators together with the research associates. The securely circulated de-identified dataset was available for analysis by project investigators located at BPKMCH and at IARC/WHO.
Quantitative sociodemographic variables were categorised for the ease of interpretation. Patient’s age at diagnosis was categorised into three categories as ‘less than 50 years’, ‘50–69 years’, ‘70 years or older’. The prediagnostic interval was defined as the duration, in days, from the day of recognition of symptom to date of first visit to a health facility. The diagnostic interval was defined as the duration from the date of first visit to health facility to date of confirmatory diagnosis of cancer. The treatment interval was defined as the duration from the date of confirmatory diagnosis of cancer to the date of start of cancer-directed treatment. Stage at diagnosis was categorised as ‘early stage’ for stage I and stage II and as ‘advanced stage’ for stage III and stage IV.

Statistical analysis
Analysis included mainly descriptive statistics. Intervals from symptom recognition to various phases till diagnosis were calculated and reported using median and IQR. The Kruskal-Wallis test was used to compare the medians across groups and reported using p-values with values less than 0.05 being considered as a significant difference between groups. The effect of various socioeconomic variables on advanced stage at diagnosis was assessed using logistic regression and was reported as crude and adjusted ORs with the 95% CIs. All analyses were calculated for the overall cohort and separated by cancer types. Statistical software, Stata V.15.1 (StataCorp, LLC, College Station, TX, USA)17 was used for the analysis and the significance level was set at alpha value of 0.05.

Findings

Patient characteristics including diagnostic and treatment details
A total of 1182 patients diagnosed with six common cancers during the recruitment period of 6 months at BPKMCH hospital were included in the analysis after interviewing 1186 eligible patients (diagnosis of cancer could not be confirmed in 4 patients). Patients were well represented from all seven provinces of the country, with the highest number being from Lumbini province (270 patients) and the least from Karnali province (43 patients), which is aligned with the population size and geography of the provinces (online supplemental appendix figure 1). Number (and proportion) of different cancer sites were as follows: breast—280 (23.7%), lung—246 (20.8%), oral cavity—239 (20.2%), cervix—221 (18.7%), colorectum—102 (8.6%) and stomach—94 (7.9%). About 57% (n=673) of the recruited patients were diagnosed at BPKMCH hospital and the remaining (n=509) were diagnosed elsewhere before visiting the hospital (online supplemental appendix figure 2). Stage information was available for 1110 patients, of whom 430 (38.7%) were diagnosed at an early stage. Proportion of early-stage cancers was relatively higher for breast (57.3%) and cervix (65.2%) compared with other types. Cancer-directed treatment was offered to 95.9% (n=1134) of the patients.

Intervals in pathways of care
Figure 1 and online supplemental appendix table 1 represent the intervals across various phases of the diagnostic and treatment pathway by cancer types. The prediagnostic interval ranged from a median of 31 days (IQR=10–107) for colorectal, lung and stomach cancers to 41 days (IQR=17–84 days) for cervical cancer. The overall median was 31 days (IQR=10–63 days). The diagnostic interval varied significantly with breast cancer being the shortest at 34 days (IQR=16–71 days) to lung cancer the longest at 62 days (IQR=25–111), with an overall median of 48 days (IQR=21–104 days). The median of combined prediagnostic and diagnostic interval (from symptom recognition to diagnosis) was shortest for breast cancer (96 days, IQR=47–187 days) and longest for colorectal cancer (142 days, IQR=77–256). Overall, 56.9% of the patients required more than 90 days from symptom recognition to cancer diagnosis with the proportions by cancer sites being 68.6% for colorectal cancer, 60.2% for cervical cancer, 59.6% for gastric cancer, 59.4% for lung cancer, 51.9% for oral cancer and 51.4% for breast cancer.
The median treatment interval (from diagnosis to treatment initiation, 94.6% of cohort participants) was shortest for stomach cancer at 13 days (IQR=7–26 days) and longest for cervical cancer at 39 days (IQR=17–74 days), with an overall median time of 19 days (IQR=7–40). Substantial delays occurred in the total time interval from symptom recognition to treatment initiation with a median duration ranging from 110 days (IQR=63–204) for breast cancer to 171 days (105–311 days) for colorectal cancer. Overall, 63.6% of the patients required more than 120 days from symptom recognition to treatment initiation with the proportions by cancer sites being 72.6% for colorectal cancer, 72.4% for cervical cancer, 67.4% for oral cancer, 64.2% for lung cancer, 59.6% for gastric cancer and 51.1% for breast cancer. The fourth quartile (75th–100th percentile) of patients waited for at least 6 months till the start of treatment irrespective of the diagnosed cancer type with the longest waiting time of more than 10 months by colorectal cancer patients.
Prediagnostic and diagnostic intervals were compared by the stage at diagnosis and by sex separately for each type of cancer (table 1). Among males, no significant difference was observed in the median prediagnostic intervals between early and advanced stage of any of the cancers. The median diagnostic interval exceeded 100 days for colorectal cancer (advanced stage) and stomach cancer (early stage). The median diagnostic interval for lung cancer was also long (exceeded 60 days) in males. In females, there was also no significant difference in the prediagnostic interval between early and late stage of any of the cancers, though the overall median interval was lower for early stage (18 days) compared with late-stage cancers (31 days). The median diagnostic intervals for stomach cancer and colorectal cancer exceeded the same for other cancers in females. Median prediagnostic intervals in females ranged from 6 days (IQR=1–10 days) for early-stage stomach cancer to 41 days (IQR=17–77 days) for late-stage cervical cancer.

Determinants of risk of advanced stage at diagnosis
We analysed the risk of advanced-stage cancer (stage III/IV, n=666; 56.4%) of the six studied cancer types (table 2). The odds of advanced-stage cancer were significantly higher in males as compared with females (adjusted OR=3.13, 95% CI 2.86 to 3.43). Age was a significant factor, with older patients having an increased odds of advanced-stage cancer as compared with those under 50 years (adjusted OR in 50–69 years old=1.3, 95% CI 1.05 to 1.45; in 70 years or older=1.63, 95% CI 1.38 to 1.94). Place of residence and education levels did not show significant associations with advanced-stage presentation. Among ethnic groups, Dalits (lower socioeconomic class) had an increased adjusted risk compared with higher class Brahmin/Chhetri (adjusted OR=1.80, 95% CI 1.41 to 2.26). Odds of advanced-stage disease by type of cancer are reported in appendix (online supplemental appendix table 2), with higher risk of cervical cancer among women of lower socioeconomic class (compared with high socioeconomic class) and lower risk of the same cancer among women in technical and other professional jobs (compared with those who were housewives).
Across the studied cancer types, lung cancer showed the highest adjusted odds of advanced-stage disease compared with colorectal cancer (used as the reference), with an OR of 5.47 (95% CI 2.98 to 9.14), followed by stomach cancer (adjusted OR=2.67, 95% CI 1.43 to 4.30) in both sexes combined (online supplemental appendix table 3). Breast and cervical cancers had lower odds, with adjusted ORs of 0.36 (95% CI 0.25 to 0.54) and 0.26 (95% CI 0.16 to 0.39), respectively.

Mode of diagnosis and treatment modality by stage
Confirmatory diagnostic methods varied by cancer type and stage as presented in table 3. A biopsy was the most common modality of diagnosis used in 84.0% (362/430) of early-stage and 83.7% (569/680) of advanced-stage cases. Cytology was the confirmatory test in 17.3% (27/156) early-stage cases and 12.9% (15/116) of advanced-stage breast cancers. Radiology (including CT scan) alone or colonoscopy alone was used as the confirmatory diagnosis of 31.4% (11/35) early-stage cases and 40.3% (25/62) advanced-stage colorectal cancers without any histopathology.
Treatment patterns differed by the site of tumour, with surgery combined with radiotherapy and chemotherapy (S+RT+ CT) being the most frequent treatment modality for both early-stage (47.4%) and advanced-stage (48.3%) breast cancers. Combination of RT and CT was used as the treatment modality for management of both early stage (41.0%) and advanced stage (43.2%) cervical cancer patients. Only RT was used to treat 20.9% and 25.7% of early and advanced stages of cervical cancers, respectively. Chemotherapy alone (CT) dominated advanced lung (57.0%) and stomach (38.6%) cancer management. Surgery alone (S) was used in 46.3% of early-stage oral cancers, while advanced stage cancers were treated with multimodal therapy.

Discussion
We performed a comprehensive assessment of different intervals in the diagnostic pathway of common cancers in Nepal in compliance with definitions stipulated by the Aarhus statement aimed at improving the reporting of studies in cancer early diagnosis.18 This study provided an opportunity to compare the intervals (and delays) across cancer types and the way these intervals relate with age, socioeconomic factors and stage at diagnosis. Baseline information is directly collected from cancer patients and corroborated with their medical records to ensure completeness and high quality of the data. The results highlight strikingly long duration (median ranges from 110 to 171 days) from recognition of symptoms to start of treatment irrespective of the diagnosed cancer type. The fourth quartile (75th–100th percentile) of patients waited for more than 6 months till the start of treatment irrespective of the studied cancer types. This study identified that patients at an older age and those belonging to low socioeconomic status are particularly vulnerable to late diagnosis. These delays may reflect barriers in symptom recognition, healthcare access or referral systems.19 Our results emphasise the clear need for interventions targeting the following three steps highlighted in the WHO Guide to Cancer Early Diagnosis, and in doing so, improve early cancer diagnosis in Nepal.5 These steps are improving community awareness and access to primary care to reduce prediagnostic delay, expanding facilities for cancer diagnosis to reduce the diagnostic delay and strengthening referral and treatment facilities to reduce treatment delay. As a priority for a limited-resourced country like Nepal, the Firstline healthcare providers (nurses, medical officers and general practitioners) need to be trained to be able to recognise early symptoms of common cancers, perform appropriate clinical examination to detect the signs of cancer and refer patients with high suspicion to proper facilities without delay.
All three intervals estimated by us (prediagnostic, diagnostic and treatment) have been reported to be significantly longer in the LMICs compared with the high-income countries, accounting largely for the poor cancer outcomes in the latter.20 A systematic review of a large number of studies from LMICs reported pooled median prediagnostic interval for breast cancer to be 58 days (95% CI 35 to 92), colorectal cancers 90 days (95% CI 43 to 279), stomach cancer 38 days (95% CI 30 to 46), cervical cancer 79 days (95% CI 14 to 119), oral cancer 55 days (95% CI 35 to 90) and lung cancer 33 days (95% CI 28 to 59).20 The intervals for the same cancers were lower in our study. However, a major limitation of estimating prediagnostic interval is that it is dependent on the patient’s ability to recognise the early symptoms and is subject to recall bias. This may explain the difference.
The same systematic review estimated pooled diagnostic intervals for various cancers in LMICs, though the number of studies was limited. The median intervals were 55 days (95% CI 26 to 93) for breast cancer, 61 days (95% CI 30 to 91) for colorectal cancer, 39 days (95% CI 32 to 46) for stomach cancer and 30 days (95% CI 16 to 44) for oral cancer.20 Except for breast cancer, the intervals for all other cancers were longer in Nepal. The long diagnostic intervals reported for cancers requiring specialised investigations (endoscopy, CT scan) are indicative of limited access to these tests in the country. Pooled median treatment intervals reported for LMICs in the systematic review were 28 days (95% CI 22 to 40) for breast cancer, 19 days (95% CI 10 to 30) for colorectal cancer, 82 days (95% CI 14 to 150) for gastric cancer, 71 days (95% CI 63 to 109) for cervical cancer and 55 days (95% CI 20 to 103) for oral cancer. A significant observation in our study is that more than half of the cancers were diagnosed at BPKMCH itself, indicating lack of access to diagnostic facilities in the periphery. Decentralisation of diagnostic facilities to primary and secondary levels of care will reduce the load on the already overwhelmed oncology centres and reduce diagnostic delay.
Delays in all the intervals have a measurable impact on post-treatment survival. Using large datasets from the USA, a study estimated a relative 24% decline in disease-specific survival per month of delay in treatment initiation for breast cancer, and a 3.1%–4.6% absolute decline in overall survival with delays of 90 days.21 Diagnostic delay resulting in increased mortality was observed for head and neck cancers in a systematic review (relative risk of mortality 1.34; 95% CI 1.12 to 1.61).22 The impact of delays on prognosis can be worse for more aggressive cancers like lung and stomach, where time-sensitive interventions are crucial.

Clinical and public health implications
Reducing delays in the pathway of care for patients with symptoms suggestive of cancers needs to be a key component of cancer control policies for LMICs including Nepal. Efforts to address the observed delays in cancer diagnosis and treatment require a concerted effort from stakeholders at the patient, provider and health system levels. This multitiered approach can significantly improve early diagnosis and timely management, ultimately enhancing outcomes. At the patient level, the focus must be on empowering individuals to recognise symptoms and seek care promptly that includes strategies such as awareness campaigns, investment in formal education, addressing sociocultural barriers and facilitating access to healthcare services.5 23 Promoting the use of primary care for initial contact, investment in accessible diagnostic technologies and capacity building are vital to minimise intervals to ‘seeking and receiving’ care.5 24 One of the proven methods for reducing diagnostic delay is to train primary care providers to be able to recognise the early signs and symptoms of common cancers and also to provide patient-centred communication to reduce fear and misunderstanding about the diagnostic process in the line of the National Institute for Health and Care Excellence guidelines in the UK.25 Preventing ‘financial toxicity’ from treatment of advanced stage cancer by addressing the social determinants of health is of paramount importance in a country like Nepal where a large proportion of patients attending public health services have no formal education and lack formal employment.26 Decision makers should take actions to strengthen the referral system, invest in efficient and need-based appropriate diagnostic infrastructure development and develop robust data collection systems.24 27
One of the best practices to promote early diagnosis is to set predefined quality standards for diagnostic interval (usually 60 days) and treatment interval (usually 60 days).28 These standards can be monitored by conducting hospital-based POC studies at a regular interval. POC studies are also critical to investigate disparities in treatment quality, access and outcomes including survival, especially for stage-specific therapies, which would further inform policy and practice improvements, thus guiding equitable healthcare delivery.29
Our study has several limitations. first, our use of convenient sampling method for patients’ recruitment might have introduced selection bias with uneven number of patients recruited by cancer types. However, the distribution of participants by cancer type aligns with the real-world cancer burden in the country with common cancers like breast and lung being better represented as compared with less common stomach cancer.30 Second, our reliance on the patient’s memory for the dates such as date of symptom recognition or date of first visit to health facility, which are not always recorded on the patients’ medical records to verify might have created recall bias. The interviewer made efforts to minimise the recall bias by allowing patient sufficient time to remember if that day was at the beginning, mid or end of the month and the date was marked as 5th, 15th or 25th of the month. Similarly, our inclusion criteria involved those cases diagnosed within 12 months at the time of interview assuming that patients were likely to remember the recent dates. One important limitation is that our study was oncology centre-based, thus missing the larger picture of delays occurring in cancer patients who do not reach the hospitals for treatment. Additionally, the proportion of literate population in this study cohort (ie, 55%) is under-representative of the national average of 76%, whereas various broadly classified ethnic groups are proportionally represented to that of the nation’s scenario.31
Stage information was missing in 6% (72 out of 1182) of the study patients. We did multiple imputation for the missing data, which yielded no changes in the study results, this indicates that the missing data likely did not introduce significant bias or influence the robustness of the findings (online supplemental appendix 4). There were some losses of contacts with the study participants to complete any missing information at the baseline interview. Interviewers made efforts to contact them through repeated phone calls as well as contacting them through local healthcare facilities that resulted in baseline information fully completed.
In conclusion, our POC study has highlighted the unacceptable delays in accessing diagnostic and treatment services for patients suffering from the common cancers in Nepal. Integrated multidisciplinary efforts to bridge gaps across prediagnostic and diagnostic intervals, particularly for cancers with pronounced delays, can significantly improve cancer outcomes and reduce bottlenecks in cancer care pathways in Nepal and in other LMICs.

Supplementary material
10.1136/bmjgh-2025-019297online supplemental file 1