Impact of and recovery after Covid-19 on breast cancer diagnosis and care in Johannesburg, South Africa

Abstract

Background

COVID-19 negatively impacted cancer diagnostic and treatment services worldwide. These challenges were further compounded in low- and middle-income countries, where longstanding resource constraints limited the capacity to maintain standard treatment pathways. We aimed to assess the impact of COVID-19 on (i) stage at diagnosis and time intervals from symptom recognition to a pathological confirmed breast cancer (BC) diagnosis and (ii) treatment initiation at two academic hospitals in Johannesburg; Chris Hani Baragwaneth Academic Hospital (Site 1) and Charlotte Maxeke Johannesburg Academic Hospital (Site 2).

Methods

We compared the proportion of BC cases diagnosed at a late stage (stage III/IV) and median time intervals of pre-contact (T₁), diagnostic (T₂) and treatment initiation following diagnosis (T₃) pre, during and post COVID-19 among 2522 women. Using cumulative Kaplan-Meier time-to-event analysis we assessed the proportions of women achieving internationally benchmarked intervals.

Results

Late-stage BC diagnoses increased during COVID-19 but subsequently recovered, with pre-, during-; and post-COVID-19 percentages, respectively, of 50%, 60% and 53% at Site 1 and 63%, 69% and 55% at Site 2. Compared to pre-COVID-19, at Site 1 the T₁ interval increased during and post COVID-19 (from 0.8 (IQR: 0.3-4.0) months to 1.0 (IQR: 0.5-6.0) and 2.0 (IQR: 0.5-6.0) post COVID-19 (P < 0.001), respectively); there were no significant T₁ changes at Site 2. Delays more than 3 months were associated with late-stage diagnosis. T₂ intervals did not differ significantly across COVID-19 periods at both sites. The treatment interval (T₃) at Site 1increased from a pre-COVID-19 median of 1.6 (IQR: 1.0-2.6) to 2.0 (IQR: 1.2–4.3) during COVID-19 and remained at 2.0 (IQR: 1.1-4.0) post COVID-19 (P < 0.001); at Site 2 an increase also occurred during COVID-19 but was not significant. Internationally benchmarked staging and timeliness targets were never achieved.

Conclusions

The already high prevalence of late-stage at diagnosis increased further during the COVID-19 pandemic with evidence of recovery at both sites in the post COVID-19 period. Treatment initiation delays increased and surgery volumes decreased during the pandemic and did not recover to pre-pandemic levels. Our study shows current cancer human and facility resources are insufficient to manage the steadily increasing BC burden in Johannesburg, South Africa.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12885-026-17068-x.

Introduction

Breast cancer (BC) is the highest incidence cancer among women globally [1]. In 2022, there were an estimated 2.3 million newly diagnosed cases and 670 000 BC related deaths worldwide [1, 2]. Compared with high- human development index (HDI) countries, where mortality from BC is steadily declining, regions such as sub-Saharan Africa (SSA) and Asia continue to experience high mortality rates relative to their comparatively lower BC incidence [3–7]. This is mainly due to delayed diagnosis and inadequate BC treatments provided in these regions. South Africa is a middle-income-country where cancer diagnostic and treatment services are provided at no cost to the socioeconomically disadvantaged 85% of the population, who access the public healthcare system [8]. These communities bear comparatively minor out-of-pocket transport and hospital access fees, yet even in urban settings where cancer health care resources are accessible to most community members, women (and men) who access these services are diagnosed mainly with late-stage BC and survival rates are low [9, 10].

COVID-19 was declared as a pandemic by the World Health Organization (WHO) on March 11, 2020 [11]. In response, South Africa implemented a national lockdown from March 2020 to June 2022. Hospital services were negatively affected as restrictions on patient movement and diversion of resources significantly exacerbated existing healthcare challenges [12–15]. Both public and private sectors experienced a decline in primary health clinic and general practitioner consultations, surgical admissions and delays in service delivery [15]. The pandemic caused widespread disruptions to essential cancer diagnostics, including a 5% reduction in mammography, 23% decline in Pap smear screenings, and a 9% decrease in oncology diagnoses within the private sector [15]. The public health sector was the hardest hit disproportionately affecting socioeconomically disadvantaged populations and potentially worsening existing disparities in BC outcomes [13, 16]. Worldwide, delays in BC screening resulted in more late-stage diagnosis and treatment delays as COVID-19 infection control was prioritised [17, 18]. Treatment practices were also affected during the pandemic with a shift to preoperative therapies such as neoadjuvant chemotherapy and hormonal therapy [19], and reductions in breast surgeries, especially reconstructive procedures [20].

Amidst the peak of the COVID-19 pandemic in 2021, the WHO launched the Global Breast Cancer Initiative (GBCI), defining three implementation pillars aimed at addressing disparities in BC survival [21]. GBCI Pillar 1 targets community health promotion for early detection with the objective that at least 60% of BC cases diagnosed at stage I or II disease. Pillar 2 targets a maximum 60-day diagnostic interval from first health system contact to a pathologically confirmed BC diagnosis. Pillar 3 aims for more than 80% of patients to complete multimodality treatment without abandonment and to initiate treatment within 12 weeks of first health system contact [22]. Moreover, in 2024, the European Society of Breast Cancer Specialists (EUSOMA) updated their quality indicators for BC management, including targeting 80% of patients to initiate first treatment between 6 and 8 weeks from date of diagnostic imaging and associated core biopsy sampling [23].

In Johannesburg, the urban economic heart of South Africa, the University of Witwatersrand-associated academic tertiary hospitals experienced both positive and negative health system changes that added to the COVID-19 impact on public health breast and other cancer management services. These included an additional chemotherapy treatment service established at Chris Hani Baragwanath Academic Hospital (referred to as Site 1 hereafter) in Soweto and a fire at the Charlotte Maxeke Johannesburg Academic Hospital (referred to as Site 2 hereafter) which shut down treatment services for 6 months. COVID-19 and additional health service changes provide a unique opportunity to investigate how these disruptions impact patient and health system time intervals to BC diagnosis, stage at diagnosis and treatment initiation. We compared pre-, during- and post-COVID-19 enrolment periods for (i) stage at diagnosis and (ii) time intervals from symptom recognition to a pathological confirmed BC diagnosis and treatment initiation at Site 1 and Site 2.

Materials and methods

Study setting and data source

Data were collected from the two Johannesburg hospital sites that participated in a longitudinal cohort study assessing determinants of breast cancer primary and secondary outcomes. Enrolment for the study began in 2015 [24, 25]. These sites, major tertiary academic hospitals affiliated with the University of the Witwatersrand, were the only two for which the required data and ethics approval had been obtained. While six hospitals in Gauteng and KwaZulu Natal provinces participated in the study, only the Johannesburg sites extended participant accrual beyond December 2019 and thus data from these two sites were included for this analysis. Site 1, located in Soweto in the south of Johannesburg and Site 2 located in central Johannesburg [26]. Johannesburg is marred by poverty, and persistent inequalities in health care, education, housing, and living conditions [27, 28].

As previously described, Site 1 services an estimated population of three million in and neighbouring Soweto. It runs a weekly walk-in outpatient surgical breast clinic, where patients are referred from primary health clinics or general practitioners in Soweto and via secondary hospitals located within a 65 km radius of the site [25]. Patients referred from primary health clinics are diagnosed and treated at Site 1, whereas those referred via secondary hospitals receive their diagnostic workups before being referred. This weekly walk-in model may influence the dynamics of the diagnostic interval differently from appointment-based systems. Site 1 provides diagnostic imaging, surgical, chemotherapy services and receptor-targeted treatments, but patients received their radiation oncology treatments at the central Site 2 facility [24, 25]. Site 2 serves communities in central and eastern Johannesburg. Similarly to Site 1, most patients are referred from community primary health clinics and general practitioners, and the rest are referred via regional secondary hospitals (39%). Around 90% of patients receive diagnostic workup at Site 2, irrespective of the mode of referral. Site 2 provides comprehensive diagnostic, surgery, chemotherapy and targeted treatments to patients and a centralised radiation oncology service [25].

Data were collected from the two Johannesburg hospital sites that participated in the South African Breast Cancer and HIV Outcomes (SABCHO) cohort study, initiated in 2015 [29, 30]. Adult women, 18 years and older, newly diagnosed with stage I-IV BC who provided written informed consent to participate in SABCHO were included.

Study participants and data collection

Three discrete, ~ 2-year participant enrolment time periods were selected to compare journeys to diagnosis, defined by histopathology confirmation, and subsequent treatment before, during and after the COVID-19 pandemic. These were: (i) pre-COVID-19 (1 April 2017–31 March 2019), (ii) during COVID-19 (1 April 2020–31 March 2022) during which a fire at Site 2 in April 2021 disrupted treatment services and a new chemotherapy unit at Site 1 commenced services in June 2021; and (iii) post COVID-19 (1 July 2022–30 June 2024). The developments at Site 2 could exacerbate the negative impacts of COVID-19 on service delivery at the hospital. Conversely, the new unit at Site 1 could result in a positive treatment trend, contrary to assumptions about the COVID-19 period. These are important factors to consider when interpreting our findings. The cut-off dates and intervening periods were deliberately selected to minimise temporal spill over between enrolment periods and to ensure that each period represented a distinct pre-COVID-19, COVID-19 and post COVID-19 phase.

The primary outcome variables were the Tumour, Node, Metastasis (TNM) stage at diagnosis (collapsed to stage I/II or III/IV) and three critical time intervals (pre-contact, diagnostic, and treatment intervals). This study adapted the Model of Pathways to Treatment [25] (Fig. 1) to explore the pathways and time frames to symptomatic patient access, BC diagnosis and treatment initiation [31]. In accordance with GBCI targets, these intervals were: self-recalled T₁, the time elapsed from initial symptom recognition to first healthcare system access (pre-contact interval); T₂, the time interval from first health system contact to a pathology confirmed diagnosis (diagnostic interval), and T_3, the time interval from a pathology confirmed diagnosis to initiation of first treatment (treatment interval), which was surgery/chemotherapy/endocrine therapy, assessed from diagnostic and treatment records.

Fig. 1.

An adaptation of Scott’s conceptual model on the journey from symptom recognition to diagnosis and first treatment [31]

During face-to-face baseline interviews conducted at time of diagnosis, participants were asked to recall details of their visits to healthcare facilities and timeframes from symptom recognition to definitive diagnosis. The interviews included obtaining written informed consent for participation in a validated questionnaire and for access to patients’ medical records [29]. The questionnaire included questions to assess knowledge and awareness of BC risks, signs and symptoms, as previously detailed [27]. Participants further provided their self-reported history of clinical breast examination before BC diagnosis. The study procedures and data collection methodology have been comprehensively described in a published protocol [24]. Personal, external and health system factors that contributed to delays greater than 1 month in the diagnostic journey were also recorded. Though data collection tools were in English, trained interviewers were well versed in South Africa’s indigenous African languages to collect BC related information [25]. Dates of histopathological confirmed diagnoses and dates for initiation of surgery, endocrine therapy or chemotherapy as first treatments were identified from patient clinical records entered on the study database.

Sample size

Of 3778 women diagnosed with invasive breast malignancies at the sites pre, during and post COVID-19 periods, a total of 2623 (69%) patients consented to participate, 82% pre-COVID-19, some 76% during COVID-19 and 54% post-COVID-19 when substantial increases in patient volumes managed at each site curtailed study enrolment (Supplementary Fig. 1). Of these, 101 participants were excluded from analysis, of whom eighty-one (81) were diagnosed with ductal carcinoma in situ (DCIS), 5 had phyllodes tumours, one case was a breast sarcoma and 13 were lost to follow-up during diagnostic workup. Some 2522 women diagnosed with invasive breast carcinomas, 1641 at Site 1 and 881 at Site 2 were included for analysis.

Statistical analysis

Our study had two primary outcomes of interest: (i) the lengths, in months, of the pre-contact (T₁), diagnostic (T₂) and treatment (T₃) intervals; and (ii) BC stage at diagnosis. For all three time intervals, the calculated means were greater than their corresponding medians, indicating positively skewed distributions. Consequently, medians and inter-quartile ranges (IQRs) were calculated for each enrolment period (pre-COVID-19 COVID-19 post-COVID-19) and by study site. Chi-square tests were used to assess differences in categorical variables across the three enrolment periods, stratified by study site. Kruskal-Wallis tests were used to assess differences in continuous variables across enrolment periods, with statistical significance defined at P < 0.05.

Kaplan-Meier time-to-event analyses were used to estimate the proportion of women: (a) reaching first health system contact within 3 months from symptom recognition (T₁), (b) receiving a pathologically confirmed diagnosed within 2 months of first health system contact (T₂); and (c) initiating first treatment within 2 and 3 months following pathologically confirmed diagnosis (T₃). Log-Rank tests were used to assess differences in time-to-event distributions across enrolment periods and study sites, with statistical significance defined at P < 0.05.

Multinomial logistic regression models were used to examine associations between categorised journey-duration intervals and BC stage at diagnosis, with the shortest duration category specified at the reference outcome. Model coefficients were exponentiated and reported as odds ratios (ORs) with 95% confidence intervals (CIs). All models were adjusted for age at diagnosis. Median spline plots were generated to assess three-monthly variations in median T₁ and T₂ intervals and BC stage at diagnosis.

Simulation-based sensitivity analyses were conducted to evaluate whether the achieved sample sizes were sufficient to detect differences in the T₁, T₂ and T₃ interval distributions across the three enrolment periods, separately for each study site. Simulations reproduced the site-specific sample sizes and approximated the positively skewed interval distributions using log-normal distributions parameterised from the observed medians and interquartile ranges. For each interval and site, 5000 simulated datasets were analysed using the Kruskal-Wallis test at α = 0.05, consistent with the primary analysis. All statistical analyses were performed using Stata software version 18.

Results

Of the 2522 women enrolled, the total number showed a steady increase at the Site 1 site across the three analysis periods. In contrast at Site 2, the fire during COVID-19 caused a 23% decrease in the number of patients; post-COVID-19, the numbers of patients managed at both sites increased by more than 20% when compared with the pre- COVID-19 volumes, with no increase in provider resources (Supplementary Fig. 1). The overall mean (SD) age at diagnosis was 54 [14] years. At Site 1, the mean age was 55, 53, and 54 years pre-, during-, and post-COVID-19 periods respectively (P = 0.378). At Site 2, the corresponding mean ages were 55, 54, and 52 years (P = 0.060). Most women first told a family member of their breast symptoms (86%-96%) with the highest proportion reported during COVID-19 at both Site 1(96%, P < 0.001) and Site 2 (90%, P = 0.018). The majority also visited a primary health facility as their first care seeking contact; 77%-80% at Site (P = 0.639) and 54%-65% at Site 2 (P = 0.193) with no significant differences observe by enrolment period (Table 1). For both sites, late stage at diagnosis proportions increased during COVID-19 with recovery post-pandemic. At Site 1, late-stage diagnosis increased from 50% pre- to 60% during, recovering to 53% post- pandemic (P = 0.002). Similarly, at Site 2, late-stage diagnoses rose from 63% pre- to 69% during the pandemic, then decreased to 55% thereafter (P = 0.006) (Table 1). The majority of patients (> 70%) has hormone-responsive tumours, with no significant difference in sub-type distribution over the three enrolment periods. Initiation of first treatment aligned with South African and International guidelines as previously described [24, 25, 32]. There was a progressive increase in the proportions of neoadjuvant chemotherapy provided at both sites; it steadily increased from the pre-COVID-19 period of 35%, to 43% during the pandemic and 54% post COVID-19 at Site 1 (P < 0.001) and Site 2 experienced a 10% increase from pre COVID-19 (52%), during (56%) and post COVID-19 (62%) of neoadjuvant chemotherapy (P = 0.001) (Table 1). The proportion of women who received no treatment did not change over the three enrolment periods enrolment periods.

Table 1.

Navigation-related features of the journey from patient symptom detection to breast cancer diagnosis by study population group and site

	SITE 1				SITE 2
	Pre-COVID-19 n (%)	During- COVID-19 n (%)	Post- COVID-19 n (%)	P-value	Pre-COVID-19 n (%)	During- COVID-19 n (%)	Post-COVID-19 n (%)	P-value
Number of women:	561(100)	599 (100)	481 (100)		376 (100)	240 (100)	265 (100)
DEMOGRAPHIC AND CLINICAL FACTORS
Age at diagnosis
Average age (SD):	55 (15)	53 (13)	54 (14)	0.378	55 (14)	54 (12)	52 (13)	0.060
Missing: 0
Age in years:				0.755				0.252
< 50 years	225 (40)	253 (42)	200 (42)		156 (41)	96 (40)	124 (47)
≥ 50 years	336 (60)	346 (58)	281 (58)		220 (59)	144 (60)	141 (53)
Missing: 0
BC stage at diagnosis				0.002				0.006
Early-stage (I + II)	279 (50)	239 (40)	228 (47)		140 (37)	75 (31)	119 (45)
Late-stage (III + IV)	282 (50)	360 (60)	253 (53)		236 (63)	165 (68)	146 (55)
Missing: 0
Receptor subtype				0.023				0.172
Luminal A	83 (15)	98 (17)	49 (10)		78 (21)	45 (19)	34 (13)
Luminal B	340 (63)	319 (56)	303 (64)		193 (52)	124 (54)	139 (55)
HER2-Enriched	33 (6)	43 (8)	30 (6)		32 (9)	16 (7)	18 (7)
Triple negative	89 (16)	106 (19)	94 (20)		68 (18)	47 (20)	63 (25)
Missing:78	6	33	5		5	18	11
First treatment initiated a				< 0.001				0.001
Neoadjuvant chemotherapy	197 (35)	256 (43)	261 (54)		195 (52)	134 (56)	164 (62)
Endocrine treatment	59 (11)	92 (15)	52 (11)		7 (2)	3 (1)	9 (3)
Surgery	247 (44)	188 (31)	119 (25)		120 (32)	60 (25)	44 (17)
No treatment/palliative radiation	58 (10)	63 (11)	49 (10)		54 (14)	43 (18)	48 (18)
Missing:0
NAVIGATION-RELATED FEATURES
First person told after symptom recognition				< 0.001				0.018
Family member	484 (86)	575 (96)	450 (95)		323 (86)	218 (90)	230 (87)
Neighbour / friend	18 (3)	6 (1)	6 (1)		18 (5)	5 (2)	12 (4.5)
Private doctor (gp) / pharmacist	13 (2)	3 (0.5)	5 (1)		8 (2)	10 (4)	4 (1.5)
Clinic nurse or doctor	37 (7)	13 (2)	14 (3)		21 (6)	4 (2)	17 (6)
Church and other	7 (1)	0 (0)	0 (0)		6 (1)	0 (0)	2 (1)
Missing: 13	2	2	6			3
Type of first HS contact				0.639				0.193
Primary health care contact	430 (77)	474 (79)	387 (80)		213 (57)	129 (54)	172 (65)
Private doctor (GP) / pharmacy	91 (16)	91 (15)	67 (14)		94 (25)	70 (29)	63 (24)
Traditional and spiritual healers / herbalist / homeopathy	1 (0)	1 (0.2)	0 (0)		1 (0.1)	1 (0.4)	1 (0.3)
Referral hospital	32 (6)	31 (5)	23 (5)		61 (16)	34 (14)	27 (10)
Tertiary hospital breast clinicb	7 (1)	2 (0.3)	3 (0.6)		7 (1.9)	6 (2.5)	2 (0.7)
Missing:1			1
Outcome of first HS contact c				0.523				0.003
Remained within healthcare system	109 (19)	101 (17)	121 (25)		76 (21)	41 (17)	57 (22)
Referred to secondary hospital	186 (34)	176 (30)	137 (29)		178 (48)	124 (53)	137 (52.1)
Referred to tertiary breast clinic	259 (47)	320 (53)	219 (46)		115 (31)	69 (30)	69 (26.2)
Missing: 1			1
Number of HS visits before arrival at tertiary breast clinic				< 0.001				0.235
0–1	254 (45)	314 (54)	211 (44)		121 (32)	72 (30)	69 (26)
2	113 (20)	104 (17)	106 (22)		144 (38)	82 (34)	104 (39)
3	86 (15)	116 (19)	75 (16)		72 (19)	52 (22)	67 (26)
>4	107 (20)	63 (10)	85 (18)		39 (11)	33 (14)	25 (9)
Missing: 8	1	2	4			1
Experienced cumulative delays (> 2 months) within the referral system before diagnosis after 1st HS visit				< 0.001				0.879
Yes (vs. no)	202 (357)	158 (438)	196 (283)		149 (226)	99 (141)	102 (159)
Missing: 8	1	2	4			1

Abbreviations: GP – general practitioner, HS – health system, BC – breast cancer

^aFirst treatment initiated excluded 315 participants; i. participants who had not received treatment at the time of the study (n= 270), ii. participants who had pathology confirmed BC diagnosis post treatment (n= 14)

^bparticipants who arrived at the tertiary breast clinic via a self-referral

^c27 participants who arrived at the tertiary breast clinic as first contact with the health care system were omitted

At Site 1, the proportion of women who delayed by more than 1 month to the health system increased by close to 10% during the three enrolment periods; from 42% pre-COVID-19, 46% during COVID-19 and 51% post COVID-19 (P = 0.029) (Supplementary Table 1). Believing their breast symptoms were not serious was the main reason for delays, more so during COVID-19 (83%) when compared to pre (81%) and post (75%) at Site 1 (P = 0.039) and during COVID-19 (89%) compared to pre (80%) and post-COVID-19 (79%) period at Site 2 (P < 0.001). Once in the health system, for all enrolment periods, more than 60% of women required 1–2 referral visits to reach the tertiary hospitals and 27% and 40% required more than 2 months from first health system access to a pathologically confirmed BC diagnosis (Table 1). More women reported a shorter number of visits (0–1 months) before arrival at Site 1 (P < 0.001) during the COVID-19 pandemic than other enrolment periods at the site.

Lengths of the pre-contact (T₁) and diagnostic (T₂) intervals

Site specific median (IQR) time intervals for T_1, T₂, and T₃ are presented in Fig. 2 and the proportion of woman achieving GBCI and EUSOMA targets in Fig. 3. At Site 1 the median length of the pre-contact interval (T₁) significantly increased from 0.8 (IQR: 0.3-4.0) months pre-COVID-19 to 1.0 (IQR: 0.5-6.0) during COVID-19 and 2.0 (IQR: 0.5-6.0) post COVID-19 (P < 0.001). At Site 2 the T₁ interval did not differ significantly across enrolment periods. As shown in Figs. 3 A and B, fewer than 70% of women accessed the health system within 3 months of symptom detection during any of the enrolment periods (Site 1 58%-64% and Site 2 45%-58%).

Fig. 2.

Median duration (IQR) (in months) of the pre contact (T₁), diagnostic (T₂) and treatment (T₃) intervals by the study population group and study site

Fig. 3.

Proportion of women reaching an event T₁ (A and B), T₂ (C and D) and T₃ (E and F) by the study population group (pre-, during and post-COVID-19) and by study site. A Proportion HS contact since symptom recognition (T1), by enrolment period, Site 1. B Proportion HS contact since symptom recognition (T₁), by enrolment period, Site 2. C Proportion diagnosed with BC since HS contact (T₂), by enrolment period, Site 1. D Proportion diagnosed with BC since HS contact (T₂), by enrolment period, Site 2. E Proportion initiated treatment since BC diagnosis (T₃), by enrolment period, Site 1. F Proportion initiated treatment since BC diagnosis (T₃), by enrolment period, Site 2

Surgical treatments received

As shown in table 2, during the pandemic the proportion of women who received surgery decreased drastically at Site 2 (50%) and even more markedly at both sites post the pandemic. This trend occurred both for patients with early- and late-stage disease. Because centralised radiation oncology services were constrained during all three enrolment periods, the proportion of women who received mastectomy over breast conserving surgery was high at both sites (75%-90%) irrespective of stage at diagnosis and period of enrolment. For those who did receive surgery there was no difference in trends by receptor subtype both for women with early- and late- stage disease.

Table 2.

Surgical treatments received by site, study period, stage at diagnosis and receptor subtype

	SITE 1				SITE 2
	Pre-COVID-19 n (%)	During- COVID-19 n (%)	Post- COVID-19 n (%)	P-value	Pre-COVID-19 n (%)	During- COVID-19 n (%)	Post-COVID-19 n (%)	P-value
Proportion of women received surgery	384 (68)	382 (64)	233 (48)	< 0.001	228 (61)	74(31)	29 (11)	< 0.001
Missing: 0
Type of surgery received				< 0.001				0.998
Mastectomy	330 (86)	285 (75)	209 (90)		181 (79)	59 (80)	23 (79)
Breast conserving surgery	54(14)	97 (25)	24 (10)		47 (21)	15 (20)	6 (21)
Missing: 0
Proportion who received surgery by BC stage at diagnosis *				0.003				0.453
Early stage (I + II)	246 (88)	199 (83)	137(60)		122(87)	42(56)	19(16)
Late stage (III + IV)	138 (49)	183(51)	96(38)		106(45)	32(19)	10(7)
Missing: 0
Proportion who received surgery by BC stage at diagnosis and type of surgery *
Early stage (I + II):				< 0.001				0.974
Mastectomy	199 (81)	125 (63)	117 (85)		82 (67)	29 (69)	13 (68)
Breast conserving surgery	47 (19)	74 (37)	20 (15)		40 (33)	13 (31)	6 (32)
Late stage (III + IV):				0.014				0.705
Mastectomy	131 (95)	160 (87)	92 (96)		99 (93)	30 (94)	10 (100)
Breast conserving surgery	7 (5)	23 (13)	(4 (4)		8 (7)	2 (6)	0 (0)
Missing: 0
Proportion received surgery by receptor subtype *				0.306				0.733
Luminal A	49 (59)	68 (69)	33 (67)		58 (74)	19 (42)	7 (21)
Luminal B	243 (71)	209 (66)	147 (49)		112 (58)	37 (30)	18 (13)
HER2-Enriched	22 (25)	21 (49)	10 (33)		14 (44)	3 (19)	1 (6)
Triple negative	59 (66)	64 (60)	42 (45)		42 (62)	12 (26)	2 (3)
Missing:38	11	20	1		2	3	1
Proportion who received surgery by BC stage at diagnosis and receptor subtype *
Early stage (I + II):				0.360				0.827
Luminal A	36 (15)	40 (20)	24 (18)		43 (35)	13 (31)	6 (32)
Luminal B	153 (62)	106 (53)	87 (64)		56 (64)	22 (52)	11 (69)
HER2-Enriched	10 (4)	13 (16)	4 (3)		5 (6)	1 (2)	0 (0)
Triple negative	41 (17)	29 (15)	22 (16)		17 (20)	5 (12)	1 (5)
Missing: 33	10	17	1		2	2	1
Late stage (III + IV):				0.184				0.903
Luminal A	13 (9)	28 (15)	9 (9)		15 (14)	6 (19)	1 (10)
Luminal B	90 (65)	103 (56)	60 (63)		56 (53)	15 (47)	7 (70)
HER2-Enriched	12 (9)	8 (4)	6 (6)		9 (8)	2 (6)	1 (10)
Triple negative	18 (13)	35 (19)	20 (21)		25 (24)	7 (22)	1 (10)
Missing: 5	1	3				1

*Denominators are the proportion of women who had surgery of the sub-set of women of interest in the corresponding category by study period and site

The median length of the diagnostic interval at Site 1 remained constant at 1.8 months over the three enrolment periods. At Site 2, although not statistically significant, during COVID-19, there was an increase in T2 to 2 (1.5–4.3) months. At Site 1 the proportion of women achieving the GBCI pillar 2 target of 60 days ranged from 60%-80% (Fig. 3 C) and was significantly higher during the COVID-19 period (P < 0.001) when surgeons triaged patients for breast and COVID-19 symptoms at the tertiary surgical breast unit; this function is usually delegated to triage nurses. For Site 2 (Fig. 3 D) the proportions diagnosed by 2 months remained constant for the three enrolment periods at around 58%-62% (P 0.640).

Late state BC diagnosis and length of the diagnostic journey

As expected, the longer the diagnostic journey (T₁ and T₂ combined), the later the stage at diagnosis (Supplementary Tables 2–4) confirming that analysis of these time intervals can provide insight into patient and health system related determinants of stage at diagnosis.

Figures A and B present spline graphs of participant self-reported delays during the T1 and T2 periods, displayed across 3-month intervals and stratified by late-stage diagnosis and enrolment period. Women enrolled at Site 2 experienced a longer median T1 interval before the pandemic. In contrast, at Site 1, patients experienced longer T1 durations during and post the pandemic period – with longer during the summer 2023/2024 recess period (December 2023-February 2024) (Figs. 4 A). For the T2 interval, at Site 1, a pattern was observed of increasing median diagnostic delays ≥ 4 months among participant with late-stage at diagnoses during- and post-COVID-19. At Site 2, median delays ≥ 4 months occurred during all periods for women with late-stage disease (Figs. 4 B).

Fig. 4.

median spline illustrating length (in months) of the pre contact (T₁) and diagnostic (T₂) journeys of women diagnosed with last stage BC at Site 1 and Site 2. A Median duration of women diagnosed with late-stage BC from symptom recognition to first HS contact (T1), Site 1. B Median duration of women diagnosed with late-stage BC from symptom recognition to fist HS contact (T1), Site 2. C Median duration of women diagnosed with late-stage BC from first HS contact to pathology confirmed BC diagnosis (T2), Site 1. D Median duration of women diagnosed with late-stage BC from first HS contact to pathology confirmed BC diagnosis (T2), Site 2

Time to treatment initiation (T₃) and proportions of women meeting EUSOMA targets

The treatment interval at Site 1 increased from a pre COVID-19 median of 1.5 (IQR: 1.3-3.0) to 1.8 (IQR: 1.0–3.0) during COVID-19 and remained at 1.8 (IQR: 1.0-2.9) post COVID-19 (P = 0.005) (Fig. 2). At Site 2 although an increased median treatment interval was noted during COVID-19 it was not statistically significant. As illustrated in Figs. 3 E and F, at both sites and across the COVID-19 time periods, the proportions of women who had initiated first treatment by 2 months of a confirmed diagnosis fell short of the EUSOMA 80% minimum threshold. Sub-analyses (Supplementary Figs. 2A-D) revealed that at Site 1, significantly more women had surgery as first treatment within 2 months of diagnosis in the pre COVID-19 period (81%, P < 0.001) compared with during and post the pandemic enrolment periods.

Simulation-based sensitivity analyses indicated high sensitivity (> 93%) to the observed period-specific differences in the pre-contact (T₁) and treatment (T₃) interval distributions at both study sites. In contrast, sensitivity to the comparatively small differences in the diagnostic interval (T₂) was lower, estimated at 5.0% at Site 1 and 38.7% at Site 2. These findings support the robustness of the T₁ and T₃ comparisons, while non-significant findings for T₂ must be interpreted cautiously, as the study had limited sensitivity to detect differences in this regard.

No significant differences were observed across the periods for neoadjuvant chemotherapy (NACT) initiation. At Site 2, 65%-75% women had initiated NACT by 3 months and was highest pre COVID-19 (75%, P = 0.009). There were no significant differences in surgery as first treatment by enrolment period at Site 2. The impact of COVID-19 on diagnostic and treatment initiation outcomes for each site is summarised in Table 3.

Table 3.

Summary impact of COVID-19 on breast cancer diagnosis and treatment initiation outcomes

Changes compared with pre-pandemic period	SITE 1		SITE 2
	During- COVID-19	Post- COVID-19	During- COVID-19	Post- COVID-19
Breast cancer patient volumes	Increased	Increased	Decreased	Increased
Stage III/IV proportions	Increased	Partial recovery	Increased	Recovery
Median T1 interval	Increased	Increased	No significant change	No significant change
Median T2 interval	No change	No change	Increased	Recovery
Median T3 interval	Increased	Increased	Increased	Recovery

Discussion

Our study reported the impact of the COVID-19 pandemic on stage at breast cancer diagnosis and on diagnostic and treatment timelines for urban women with BC in the public health system in Johannesburg, South Africa. During the COVID-19 enrolment period, both hospital sites recorded the highest proportions of late-stage diagnoses, echoing findings from the USA and Europe [33] examining pandemic effects on BC management. At Site 1, this finding appeared to be mainly explained by significantly longer pre-contact intervals during COVID-19, which did not recover subsequently. The pandemic also caused community members’ hesitancy to visit healthcare facilities due to fear of contracting the virus. The fear of contracting COVID-19 while using public transport and at the facility may have worsened the delays observed during the pandemic [13, 14]. The persistently longer post COVID-19 T₁ and T₃ intervals highlights how chronic system strain persisted beyond lockdowns. At Site 2 the already very high pre-COVID-19 late-stage proportion was further increased during COVID, seemingly linked to longer median diagnostic intervals and there was a drastic reduction in patients managed at the site because of the fire outbreak. Both diagnostic intervals and patient volumes recovered post-pandemic.

It is noteworthy that across all three enrolment periods, the high proportion of late-stage diagnoses fell well short of the GBCI pillar 1 target, which aims for at least 60% of patients to be diagnosed with early stage (I or II) disease and diagnostic intervals within 60 days. Our findings reflect that more than 35% of the women experienced pre-contact intervals longer than three months and 27% − 41% of women experienced diagnostic intervals > 2 months from first health system contact to a pathologically confirmed diagnosis. These results are in stark contrast to findings in high-income country settings where population-based mammography screening is well-established and delays to diagnosis and treatment initiation are few. In the US National BC early detection program between 1996 and 2005, median diagnostic intervals were between 23 and 25 days. South Africa’s public health system has no formal breast cancer screening programme. Although guidelines are in place for population level screening, this is not formalised particularly in the public health sector where mammography services are under-resourced. Consequently, mammography services are prioritised for women accessing public tertiary services for breast complaints based on urgency of clinic symptoms and based on clinical examination, patient age for those with no obvious cancer symptoms. Our findings are however consistent with those from other SSA countries, where population-based BC screening is also lacking, community awareness of symptoms remains low and health system resources are insufficient to cope with increasing BC incidence [4, 5, 34–36]. In such settings, reports of difficulties in accessing health care and diagnostic errors are common (37, 38, 39].

Treatment initiation intervals were longest during COVID-19 at both sites where the proportion of patients who initiated treatments within a timely period decreased, confirming disruptions in tertiary cancer management services during the pandemic [13–15]. All patients presenting with breast lumps—particularly older women—were prioritised for clinical assessment and biopsy. However, the national suspension of elective surgical procedures and restricted theatre availability resulted in significant treatment delays and decreases in proportions of patients who received surgery during the pandemic, which worsened in the post pandemic period studied, irrespective of stage at diagnosis. There was also a shift from predominantly conventional to hypofractionated radiotherapy and shorter course chemotherapy treatments [19, 20]. Our findings echo similar challenges experienced in other high- and low-to-middle HDI country settings. The global response was a shift in hormone receptor-positive (HR+) early BC treatment, where guideline-concordant neoadjuvant endocrine therapy (NET) was employed as a bridging strategy to defer surgery by three to six months, particularly for patients with small, node-negative, ER+/HER2– tumours [18, 40, 41]. At our sites, increases in neoadjuvant systemic treatments were influenced by both later-stage at diagnosis and constrained surgical services.

In the urban Johannesburg context, despite COVID-19 restrictions and redeployment of theatre staff, Site 1 maintained essential surgical services for BC patients, largely through the surgical breast unit’s triage-based prioritisation workflow. This could account for the large differences in patient volumes managed at the two sites during the pandemic. The treatment response at both sites was like those in other global settings. Luminal patients across all stages were commenced on endocrine therapy as either primary or bridging treatment, while non-luminal early-stage cases required individual motivation at hospital theatre vetting meetings to secure access to limited operative slots. Furthermore, access to systemic therapy was further constrained by COVID-19 surges and the Site 2 fire, endocrine therapy was consequently used as an interim measure for locally advanced and metastatic luminal cancers, whereas non-luminal advanced and metastatic patients experienced prolonged delays in initiating systemic therapy. Post-pandemic recovery saw a gradual re-expansion of surgical activity and the establishment of an on-site chemotherapy service at Site 1 in mid-2021, which markedly improved treatment access and reduced reliance on inter-hospital transfers. However, persistent constraints in theatre time, workforce shortages, and increasing patient volumes continue to limit compliance with international standards for treatment timeliness. These resource constraints are reflected in the fact that at no periods analysed did either site meet the minimum EUSOMA benchmark standard of 80% initiation of first treatments within 8 weeks of patient diagnostic workup. This suggests that the increasing BC burden at both sites is negatively impacting already constrained tertiary BC treatment resources. Without substantial investment in human and infrastructural resources, neither site is likely to achieve EUSOMA standards nor manage the rising burden of BC in South Africa.

The relationship between delayed diagnosis and treatment initiation, and worse survival outcomes, is well established [42–45]. A prior analysis of 3-year overall survival (OS) among the SABCHO cohort of women who had NACT or surgery as first treatment [46] revealed that for women who had surgery within 3 months of diagnosis, the 3-year OS was 85% compared with 79% for those who experienced surgical delays of more than 3 months, and the mortality risk increased by 11% for every additional month delayed. For the group of women who had NACT, 3-year OS was 66% vs. 72% OS for women who initiated treatment later than 3 months following diagnosis. We can therefore postulate survival for women in our cohort would be negatively impacted by the delays experienced pre-, during and post-COVID-19.

Finally, our study adapted the Model of Pathways to treatment (31) (Fig. 1) to conceptualise the journey from symptom recognition to BC diagnosis and first treatment initiated. The model was relevant to this study because it distinguishes delays occurring before healthcare contact from those occurring within the health system during diagnosis and treatment initiation, while recognising the contribution of patient, healthcare provider/system, and disease related factors across the pathway. This framework enabled us to examine whether, and at which points along the pathway, time intervals differed across the three periods and between the two study sites.

Implications for policy and practice

Our findings show that interventions are required at a community level to reduce pre-contact delays which were long and have not recovered after the pandemic. Aggressive community-based campaigns focused on breast cancer risk and early symptoms are needed to encourage self-examinations and dispel perceptions that small or painless lumps can be ignored. Diagnostic intervals did increase during COVID-19 at Site 2 albeit not significantly. Nevertheless, they were long at both sites during all three enrolment periods and are a time interval opportunity to accelerate women to diagnosis. One opportunity would be to omit secondary hospitals from suspected breast cancer referral journeys as we have previously shown that referral routes involving these hospitals are associated with late-stage diagnosis [10].

At the tertiary setting, additional resources are required to cope with the increasing patients’ volumes and to achieve international treatment initiation timeliness targets. The pandemic revealed fragility at all levels of the health system, there was fortunately a return to normalcy post COVID-19 at Site 2, but due to the persistent constraint on surgical operating time, this was only partially evident for Site 1. The pandemic-driven expansion of neoadjuvant chemotherapy and neo-endocrine treatments significantly influenced both surgical strategies and health system operations. In several centres, mastectomy rates increased as surgeons sought to minimise dependence on postoperative radiotherapy, which requires frequent hospital attendance. Extended neo-endocrine use temporarily alleviated surgical backlogs and, in some cases, enabled more patients to become eligible for breast-conserving surgery. However, the long-term oncological outcomes of these adaptations—particularly the expanded use of neo-endocrine therapies in lower-burden hormone responsive disease—remain uncertain and warrant continued surveillance, especially in LMIC settings where long-term follow-up and outcome reporting are often challenging. Overall, the COVID-19 pandemic catalysed enduring changes in breast cancer management, highlighting both the adaptability and the vulnerabilities of oncology services operating under constrained conditions. Our investigation illustrated the value of decentralised chemotherapy services and the need for parallel expansion of surgical capacity where post-COVID-19, staffing and theatre backlogs remain major determinants of treatment delay.

The consistent failure of LMICs to meet GBCI and EUSOMA targets is concerning. Policy makers should be encouraged to use evidence such as that presented by this study to streamline referrals from primary health care to tertiary centres with the capacity for timely BC diagnosis and treatment services. There is also an urgent need for political role players to explore private-public partnerships to finance the scaling of surgical and chemotherapy provision at breast clinics. Moreover, meaningful community-level engagements focused on breast education encouraging self-examinations and dispelling perceptions that small or painless lumps can be ignored is essential to reduce pre-contact delays.

Limitations

The key limitation of this study and all others assessing diagnostic intervals is the reliance on patient self-reported information on the pre-contact duration and delay periods within the referral facilities. This introduces inevitable recall bias, and some vulnerable patients may have felt unable to report the full extent of their own or health system delays due to fear of negative impacts on the quality of their healthcare in the public health system. Nevertheless, this approach did enable us to reveal the impacts of COVID-19 on stage at diagnosis, durations of pre-contact, diagnostic and treatment initiation intervals at both Johannesburg tertiary hospital sites, and the extent of recovery of these time intervals post-COVID-19. Secondly, our investigation was focused on the impact of COVID-19 on BC stage at diagnosis and timely treatment initiation, and treatment completion rates and survival outcomes were not assessed for this investigation.

Conclusions

The already high prevalence of late-stage at BC diagnosis pre COVID-19 was exacerbated during COVID-19 and recovered post-pandemic. Initiation of cancer treatment was delayed during the pandemic; there was a lower proportion of patients receiving timely cancer treatments and this did not recover to pre-COVID-19 proportions. During all enrolment periods GBCI and EUSOMA targets were never achieved, and existing human and infrastructure resources remain inadequate to cope with the steadily increasing BC burden in Johannesburg.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BC

Breast Cancer

COVID-19

coronavirus disease

DCIS

ductal carcinoma in situ

DSTI-NRF

Department of Science, Technology and Innovation and the National Research Foundation

EUSOMA

European Society of Breast Cancer Specialists

GBCI

Global Breast Cancer Initiative

GP

general practitioner

HDI

Human Development Index

HREC

Human Research Ethics Committee Medical

HS

health system

IQRs

inter-quartile ranges

NACT

neoadjuvant chemotherapy

NET

neoadjuvant endocrine therapy

OS

overall survival

SABCHO

South African BC and HIV Outcomes

SD

standard deviation

SSA

sub-Saharan Africa

WHO

World Health Organisation

Author contributions

R.P. conceptualisation and design, methodology, analysis and interpretation of data, drafted, writing, revisions and finalised work; R.P.C. substantial contribution to data analysis, writing, editing, extensive review and revisions; O.O.O. data cleaning, writing, editing and revisions; M.E.A. data cleaning, editing and revisions; W.C.C. conceptualisation and design, data curation and extraction, extensive review and contribution to analysis; N.M. review and editing; P.N. review and editing; L.K. review and editing; F.J. review and editing; S.N. review and editing; R.M.M. review and editing; K.S. review and editing ; P.O. writing and revisions; V.M. conceptualisation and design, methodology, review and editing; S.A.N. conceptualisation and design, methodology, substantial supervision and revisions; M.J. conceptualisation and design, methodology, data curation, analysis, supervision, substantial writing and revisions. All authors reviewed and approved the final version of the manuscript to submit for publication.

Funding

This study was funded by the DSTI-NRF Centre of Excellence in Human Development at the University of the Witwatersrand, Johannesburg, the NIH/NCI grants; Grant/award numbers: NIH/R01-CA19262701 and R01-CA250012; South African Medical Research Council/University of the Witwatersrand Common Epithelial Cancer Research Center (MRC/ WITS CECRC); German Federal Ministry of German Federal Ministry of Research, Technology and Space (BMFTR) 01KA2220B to the RHISSA Programme for the NORA Consortium; Science for Africa Foundation to the Programme Del-22-008 with support from Wellcome Trust and the UK Foreign, Commonwealth & Development Office and is part of the EDCPT2 programme supported by the European Union. The opinions expressed and the conclusions drawn are those of the authors and are not to be attributed to the Centre of Excellence in Human Development.

Data availability

All data and materials are made available within the manuscript and supplementary files.

Declarations

Ethics approval and consent to participate

The University of the Witwatersrand Human Research Ethics Committee (HREC) Medical approved the study (M141102) on the January 7th, 2015. Written consent was provided by patients enrolled into the study at Chris Hani Baragwanath Academic Hospital (Site 1) and Charlotte Maxeke Johannesburg Academic Hospital (Site 2).

Consent for publication

All authors consent to the publication of this paper.

Competing interests

The authors declare no competing interests.