Epidemiology and survival outcomes for liposarcoma patients in England: An observational cohort study using real world data

Abstract

Purpose: This real-world data study evaluates demographic and clinical characteristics and survival in patients with liposarcoma to inform our understanding of treatment patterns and associated outcomes in this rare cancer.

Materials and methods: A retrospective cohort study was conducted using existing data available through the Clinical Practice Research Datalink. Male and female patients aged 18 years or older who had their first ever record of liposarcoma between 1998 and 2018 were included in the study. The demographic and tumour characteristics were presented (overall, by first line treatment and for dedifferentiated liposarcoma only) as proportions (percentages) while continuous variables were presented as means with standard deviation and interquartile ranges. Survival rates from diagnosis date and first line treatment (with 95% confidence intervals) were also calculated.

Results: 1,315 patients were included, of which 46% (611) had a treatment recorded. Most patients were male and over the age of 60 years. Surgery was the most frequent treatment received following diagnosis (34% of all patients), followed by radiotherapy (8%) and chemotherapy (2.4%) with the remaining patients having no record of treatment available. Overall, there was a 77% probability of survival after diagnosis at 5 years.

Conclusions: Findings from this study help advance our understanding of real world patient characteristics, treatment patterns and survival outcomes in a rare and heterogeneous cancer, which may be useful for guiding clinical management. This study also identified challenges with using real world data, which can be minimised through improving data collection and standardisation.

Introduction

Soft tissue sarcomas (STS) are a rare, heterogeneous group of mesenchymal cancers representing approximately 1% of adult solid tumour cancers. ¹ They arise from connective tissue and include over 100 different histological subtypes which can vary in their presentation, biological behaviour, and outcome. ²

One of the most common sarcoma subtypes, representing ∼20% of cases, is liposarcoma (LPS).^3,4 It develops in fat tissues and can occur in any part of the body, most commonly in the retroperitoneum and deep soft tissue of proximal extremities. ⁵ LPS is made up of several histological subtypes (dedifferentiated (DD), well-differentiated (WD), myxoid/round cell and pleomorphic) with less common mixed and fibroblastic variants also reported. ⁶ LPS subtypes are associated with different prognoses and responses to treatment ⁷ with DDLPS being aggressive and treatment resistant.^5,7

Population-based incidence and survival data derived from current standard clinical practice are critical for improving patient care and outcomes. ⁸ Availability of population-based data for sarcomas, specifically LPS, remains scarce due to the rarity and heterogeneity of the disease and how the different subtypes are classified and recorded. There is need for further research into the epidemiology of LPS to inform the management of patients as well as future research, including new treatments.

In the context of assessing treatment outcomes, and further enabling continuous assessment of the risk/benefit ratio of newly approved medicines, population based real world data (RWD) may aid benchmarking in the evaluation of new therapies, where large scale randomised trials may be challenging.^9,10 This is particularly true for STS (and LPS specifically) where histology specific treatment decisions have been driven by RWD studies in the absence of randomised trial data, particularly in the frontline setting.^7,11,12

Regulators across the globe are recognising the value of real-world evidence (RWE) in supporting new drug development and approvals, especially in rare disease areas.^10,13 In the United Kingdom (UK), the Medicines and Healthcare products Regulatory Agency (MHRA) delivers the Clinical Practice Research Datalink (CPRD) research data service, which provides access to one of the largest, longitudinal research databases in the UK. It contains anonymised primary care patient data for approximately 60 million patients from a network of general practices (GPs) across the country and has been linked to several other health related datasets.^14–16 CPRD data has not only been used to support patient recruitment to trials, but also to provide RWD virtual control arms as disease course or standard of care treatment comparators, and for pharmacovigilance post-approval.^15,17

Study objectives

The primary objectives of this study were to describe a) the demographic and clinical characteristics of LPS patients in England, b) the characteristics of LPS tumours based on patient clinical and demographic characteristics, and c) the overall and stratified survival rates for LPS patients. Where possible, we interrogated these parameters for DDLPS as this subtype is aggressive and treatment resistant.^5,7

For the secondary objectives, the study described the proportion of LPS patients by treatment groups, treatment duration and survival time from first treatment initiation.

Methods

Study design and setting

We conducted a retrospective cohort study (using data available through the CPRD ¹⁸ ) to describe the demographics, tumour characteristics (at diagnosis), treatments and survival of patients with LPS. In doing so, we aimed to present a detailed picture of the treatment and survival of patients with this rare form of cancer in England.

Data sources

CPRD Aurum contains anonymised primary care electronic healthcare records (EHR) in the form of clinical codes (e.g. SNOMED) from GP practices using EMIS^® GP software. ¹⁹ It is the largest of CPRD’s primary care databases, including over 60 million patients (over 40 million of which are research acceptable).^19,20 The January 2022 release of the CPRD Aurum database ²⁰ was used for cohort definition and determining patient demographics. The coverage of the data extracted was from 1^st January 1987–31^st December 2021.

CPRD Aurum is also routinely linked to several other datasets to provide a more comprehensive picture of patient care, thus helping to enhance public health and clinical research. ¹⁶ Information regarding how CPRD undertake the linkage (including the joining criteria) is provided in a paper by Padmanabhan et al. ¹⁶

The ONS Death Registration dataset (containing information on date and causes of death ²¹ ) was used to more accurately define the date of death for the survival analysis. Copyright © (2022), re-used with the permission of The Health and Social Care Information Centre, all rights reserved.

The National Cancer Registration and Analysis (NCRAS) Cancer Registration (CR) data include diagnostic and treatment information for registered tumours diagnosed or treated within England ²² and was used to determine tumour characteristics, treatment patterns and apply certain inclusion/exclusion criteria for cohort definition. The NCRAS SACT (which includes information on systemic treatments ²³ ) and the NCRAS RTDS (which contains records of radiotherapy treatments) ²⁴ were used to ascertain receipt of relevant treatment and treatment duration.

Composite measures of the Index of multiple deprivation (IMD) at the Lower Super Output Areas (LSOA) level in England for 2015 have been linked with CPRD data at the patient (using the patient’s postcode ²⁵ ) and practice level (using the GP postcode ²⁶ ). Patient level IMD data were used to describe the cohort by IMD quintile (as a proxy for socio-economic status). Practice level IMD data were used if patient level information was unavailable.

Study population

Male and female patients newly diagnosed with LPS between 1^st January 1998–31^st December 2018 (to coincide with the end of the coverage period for the NCRAS datasets) were included in the study (Figure 1).

Figure 1.

Flow diagram of the patient counts following the application of the inclusion and exclusion criteria. *Research acceptable male and female patients in the January 2022 version of the CPRD Aurum database who had at least 1 day of follow up during the study period. **All patients in the study population. ***Patients in sub-cohort 1 whose index date was recorded in the NCRAS cancer registration data. ****Patients in sub-cohort 2 whose index date fell within the coverage periods for NCRAS SACT and RTDS data (1^st January 2014–31^st December 2018).

The index date was the first ever (incident) diagnosis event recorded within the study period (Appendix A), defined as a medical code recorded in the CPRD Aurum observation table (Appendix B.1) OR an ICD-10 code (Appendix B.2) AND specified morphology code (see Appendix B.3) in the NCRAS CR data. Using primary and secondary care data for defining the study cohort ensured capture of as many relevant cases as possible and reduced the risk of bias. ²⁷

Where a patient had a first LPS record in CPRD Aurum up to 30 days before a LPS record in the NCRAS CR data, the NCRAS date and diagnosis record was used as the index record as these were deemed different records for the same cancer. Additionally, patients had data deemed “acceptable” for research (based on the acceptability flag generated by CPRD ¹⁹ ) and were registered at their GP practice for at least 6 months before their index date (which had to fall within the period they were registered at their GP practice). Prior registration was required to reduce the risk of bias due to misclassification of incident LPS diagnoses. Patients also had to be aged 18 years or older on their index date.

From the above study population, patients were assigned to Sub-cohorts (Figure 1). Sub-cohort 1 (for the first primary objective) included all patients in the study population while Sub-cohort 2 (second and third primary objectives) was limited to patients whose index diagnosis was recorded in the NCRAS CR data. Lastly, Sub-cohort 3 (for all secondary objectives) comprised of patients within Sub-cohort 2 who also had their index date within the coverage periods for NCRAS SACT and RTDS data (1^st January 2014–31^st December 2018). All patients in the cohorts were registered at practices in England.

Patients were also flagged if they had DDLPS, which was defined as a tumour record in the NCRAS CR data (see Appendix B.2) with the morphology code “8858” (if the index date was recorded in the NCRAS CR data) OR a CPRD Aurum medical code specifying DDLPS (if the index date was recorded in CPRD Aurum- see Appendix B.1).

Patients were followed up from their index date until the study end (defined as the earliest of the patient’s end of registration, the CPRD derived death date, practice last collection date and the end of the study period (31/12/2018)).

The definitions of key variables included in the study are provided in Table 1.

Table 1.

Definitions of the patient demographic, tumour and treatment characteristics.

Variable	Definition
Patient demographics
Age	Age at index date (based on the year of birth recorded in CPRD Aurum)
Age group	Age at index date (based on the year of birth recorded in CPRD Aurum) grouped into seven age bands: “18–29 years”, “30–39 years”, “40–49 years”, “50–59 years”, “60–69 years”, “70–79 years” and “80+ years”
Sex	Patient’s sex (as recorded in CPRD Aurum)
Body mass index (BMI)	A valid BMI record in CPRD Aurum that was prior to and most proximal to the index date and was within 2 years (730.5 days) prior to the index date. If a patient did not have a BMI recorded, the measure was calculated using the most recent valid adult height (valid range between 1.2 and 2.2 m) and weight (valid weight range between 25 and 450 kg) recorded. Extreme values of BMI (less than 10 kg/m2 and greater than 85 kg/m2) were excluded as likely errors
	BMI records were grouped as “<18.5”, “18.5–24.9”, “25–29.9”, “30–39.9”, “40+” or “missing” (if the patient did not have a relevant record)
Area based deprivation	The most recent national quintile deprivation score available at the time for England (patient or practice level). Namely the 2015 English IMD score
Region	The region of the practice the patient was registered at (as recorded in CPRD Aurum). For practices in England, the regions are based on the ONS regions
Tumour characteristics
Morphology	Morphology information recorded in the NCRAS CR data based on specified morphology codes (see Appendix B.3)
Grade	Grade information as recorded in the in the NCRAS CR data using the scale 1–3 or 1–4. Where there are several tumours with different grades, the grade of the predominant tumour is recorded. Tumours for which grade was inappropriate or could not be assessed (“GX”) were grouped as “unknown/not appropriate”
Stage	Stage information as recorded in the NCRAS CR data using the scale of 1–4. The data is derived using a combination of best T, N and M staging (Ann arbor staging for lymphomas is also included). Tumours for which stage information was outdated or invalid were grouped as “unknown/not appropriate”
Basis of diagnosis	Basis of diagnosis information as recorded in the NCRAS CR data. The data were categorised as “histology” (including histology of a metastases and histology of a primary tumour), “other” (combining death certificate, clinical, clinical investigation, specific tumour markers and cytology’) and “missing”
Prior cancer	A record of any cancer recorded in the NCRAS CR data prior to the index date
Charlson comorbidity index (CCI)	The total charlson co-morbidity score for a 2-year period as recorded in the NCRAS CR data. Patients with a score greater than 2 were grouped together as “3+”
Dedifferentiated	A flag indicating the morphology of the tumour was dedifferentiated (defined as a tumour record in the NCRAS CR’s data with the morphology code “8858”). When derived for primary objective 1, this also included relevant CPRD Aurum medical codes which specified dedifferentiated tumours (see Appendix B.1)
Treatment characteristics
Date of first line treatment	The first treatment after index date and within the patient’s follow up (excluding the end of the study period). The type of treatment recorded against this date (surgery, chemotherapy, radiotherapy or none) as recorded in the NCRAS datasets was reported
Date for second line treatment	The second treatment after index date and within the patient’s follow up (excluding the end of the study period). The type of treatment recorded against this date (surgery, chemotherapy, radiotherapy or none) as recorded in the NCRAS datasets was reported
Date of third line treatment	The third treatment after index date and within the patient’s follow up (excluding the end of the study period). The type of treatment recorded against this date (surgery, chemotherapy, radiotherapy or none) as recorded in the NCRAS datasets was reported
Duration of first line treatment	The number of days between the start of the first treatment episode and the date that episode ended. Chemotherapy treatment records occurring within 30 days of a previous record of the same drug type were considered part of the same regimen, and therefore treatment episode. Radiotherapy treatment records occurring within 30 days of a previous record of the same intent were considered part of the same treatment episode. Records with missing end of treatment dates for chemotherapy or radiotherapy were imputed as 6 weeks (42 days) after treatment start date based on clinical advice

As this was a descriptive study (which in part aimed to described patients with this rare form of cancer), the number of cases who met the inclusion and exclusion criteria determined the sample size.

Statistical analysis

All data (except the NCRAS datasets) were extracted using tools developed by CPRD or Stata 17 as these are held in house. Extracts from the NCRAS datasets were requested from National Health Service (NHS) England who provided data specific to this study. All data management and analyses were conducted using Stata 17.

Demographic statistics were presented for all patients, by first treatment after index date (for those whose index date was recorded in the NCRAS CR data) and for patients with DDLPS. For body mass index (BMI), we required the data to be recorded within 2 years prior to the index date as this measurement could be impacted by the disease in the form of weight loss. ²⁸ Only treatment records after the index date were included (to prevent misclassification of treatment) and patients in the cohort who did not have a first treatment recorded were grouped as “No treatment”.

For Sub-cohort 2, summary statistics of tumour characteristics were calculated overall, for patients with DDLPS and by first treatment. Additionally, the median (with interquartile range [IQR]), one-, three- and 5-year survival rates (with 95% confidence intervals [CIs]) were calculated overall and by first treatment. Patients with less than 1 day of survival time (e.g., their index date was the same as or after their ONS death date) were excluded from the analysis. Kaplan-Meier curves were generated to describe survival overall, for patients with DDLPS, by sex and by 10-years age groups.

For Sub-cohort 3, duration of first treatment (for patients receiving chemotherapy or radiotherapy) and survival rates based on treatment initiation (with 95% CIs) were also calculated. Patients with less than 1 day of survival time (e.g., their treatment initiation date was the same as or after their ONS death date) were excluded from this analysis. The maximum of 4 years was calculated as no patient in Sub-cohort 3 had follow-up time of 5 years (due to the coverage period of NCRAS SACT and RTDS data). Kaplan-Meier curves were generated to describe survival overall, for patients with DDLPS, by sex and by 10-years age groups.

Where present, missing data were reported as counts and percentages. Imputation was only applied when generating the end of treatment date for patients who had this information missing. This was set to 42 days after the start of treatment date.

Due to CPRD’s policy, the results were presented in a way to prevent deductive disclosure by small cells.

This study was conducted by a multidisciplinary team comprising of experts in CPRD data, epidemiology, oncology, and data science. Collaborators were therefore able to draw on the expertise of fellow team members in the planning and execution of the study as well as the interpretation of the results.

Results

The source population contained 33.8 million male and female patients with research acceptable data, who had at least 1 day of follow-up during the study period. Of those patients, 2.7 million had a record in the NCRAS CR dataset.

During the study period (1st January 1998–31st December 2018) 1,315 patients qualified for inclusion in the study (Figure 1) with an average follow-up time (from index date) of 5.6 years (5.1 SD). The cohort contributed 7,424.5 patient years.

Of those patients, 611 (46%) were in Sub-cohort 2 (patients whose index date was recorded in the NCRAS CR data) and 169 (13%) were in Sub-cohort 3 (patients within Sub-cohort 2 who also had their index date within the coverage periods for NCRAS SACT and RTDS data).

Overall, most patients diagnosed with LPS were male (58%) and over the age of 60 years (60%) (Table 2). There were only small differences in the proportions based on IMD quintile. DDLPS was present in 18% (N = 240) of cases, of which 62% were male and over the age of 60 years. For the 611 patients whose index date was in the NCRAS CR data, 339 (55%) had no record in the available data of receiving any form of treatment (Appendix C.1). For all patients with a record of treatment, the most frequent first treatment (34% of all patients) was surgery (Appendix C.1), with a small proportion of patients receiving radiotherapy (8%) and chemotherapy (2.4%) as first treatment (Appendix C.2). Patients receiving no treatment made up the largest numbers in every age group, except the 18–29 years group, suggesting that age differences do not appear to be a key factor in the recorded treatment pattern differences (Appendix C.1). Subsequent treatment for patients following first treatment was investigated but not reported due to small cell counts.

Table 2.

Descriptive demographics for patients with liposarcoma record in NCRAS cancer registration data or CPRD Aurum primary care data- overall and by dedifferentiated tumour subgroup ^a .

		All patients	Dedifferentiated only
		N (%, unless otherwise stated)	N (%, unless otherwise stated)
Total number of cases		1315	240
Gender	Male	758 (57.6)	149 (62.1)
	Female	557 (42.4)	91 (37.9)
Mean age b (SD)		63 (14.9)	64.77 (14.0)
Age (IQR)		53–75	56–76
Age group	18–29 years	22 (1.7)	<5 (0.4–1.7)
	30–39 years	68 (5.2)	9–12 (3.8–5.0)
	40–49 years	176 (13.4)	23 (9.6)
	50–59 years	255 (19.4)	39 (16.3)
	60–69 years	315 (24.0)	69 (28.8)
	70–79 years	295 (22.4)	65 (27.1)
	80+ years	184 (14.0)	31 (12.9)
Practice region	North east	73 (5.6)	6 (2.5)
	North west	226 (17.2)	51 (21.3)
	Yorkshire & the Humber	56 (4.3)	10 (4.2)
	East midlands	26 (2.0)	5 (2.1)
	West midlands	186 (14.1)	32 (13.3)
	East of England	59 (4.5)	8 (3.3)
	London	179 (13.6)	39 (16.3)
	South east	323 (24.6)	59 (24.6)
	South west	187 (14.2)	30 (12.5)
Area based deprivation c quintile	1 – least deprived	331 (25.2)	66 (27.5)
	2	278 (21.1)	48 (20.0)
	3	269 (20.5)	50 (20.8)
	4	215 (16.3)	37 (15.4)
	5 – most deprived	222 (16.9)	39 (16.3)
Body mass index (BMI, Kg/m2)	<18.5	10 (0.8)	<5 (0.4–1.7)
	18.5–24.9	176 (13.4)	37 (15.4)
	25–29.9	266 (20.2)	52 (21.7)
	30–39.9	161 (12.2)	29 (12.1)
	40+	33 (2.5)	7–10 (2.9–4.2)
	Missing	669 (50.9)	111 (46.3)

SD: standard deviation; IQR: interquartile range.

^aDue to CPRD’s policy, the results were presented in a way to prevent deductive disclosure by small cells.

^bAge is calculated as year of diagnosis minus year of birth as recorded in CPRD Aurum.

^cArea based deprivation is measured using patient-level IMD scores for England. Where patient level measures were missing, practice level IMD scores were used.

Across all cohorts, BMI (recorded within 2 years prior and most proximal to the index date) was missing for approximately half of the patients.

There was a high proportion of missing grade and stage information (including “Unknown/ Not appropriate”) for all patients in Sub-cohort 2 (42% and 83% respectively). Where grade information was recorded, most LPS tumours were at the lowest tumour grade (Grade 1). The majority of LPS tumours were diagnosed via histology (94%) (Table 3). In terms of clinical history, the Charlson Comorbidity Index (CCI- as recorded in the NCRAS CR data) was missing for 27% patients. For patients who had this information recorded, only 11% had any form of comorbidity. For most patients (87%), LPS was their first record of cancer.

Table 3.

Characteristics of index liposarcoma tumours based on patient clinical data in the NCRAS Cancer Registration (CR) record ^a .

		All NCRAS patients	Surgery	Chemotherapy	Radiotherapy	No treatment	Dedifferentiated only
		N (%)	N (%)	N (%)	N (%)	N (%)	N (%)
Total number of cases b		611	208	15	50	339	72
Morphology	Liposarcoma, NOS, mixoid liposarcoma & fibroblastic liposarcoma	327 (53.5)	103 (49.5)	3-6 (20.0–40.0)	28 (56.0)	190 (56.0)	0 (0.0)
	Liposarcoma, well differentiated	212 (34.7)	70 (33.7)	<5 (6.7–26.7)	12 (24.0)	129 (38.1)	0 (0.0)
	Dedifferentiated liposarcoma	72 (11.8)	35 (16.8)	8 (53.3)	10 (20.0)	20 (5.9)	72 (100.0)
Grade	1	294 (48.1)	96 (46.2)	<5 (6.7–26.7)	17 (34.0)	178 (52.5)	5-8 (6.9–11.1)
	2	27 (4.4)	16 (7.7)	0 (0.0)	3-9 (6.0–18)	<5 (0.3–1.2)	10 (13.9)
	3	29 (4.7)	13-16 (6.3–7.7)	<5 (6.7–26.7)	6 (12.0)	<5 (0.3–1.2)	16 (22.2)
	4	5 (0.8)	<5 (0.5–1.9)	0 (0.0)	<5 (2.0–8.0)	<5 (0.3–1.2)	<5 (1.4–5.6)
	Unknown/not appropriate	224 (36.7)	65 (31.3)	7 (46.7)	16 (32.0)	137 (40.4)	29 (40.3)
	Missing	32 (5.2)	9–14 (4.3–6.7)	<5 (6.7–26.7)	<5 (2.0–8.0)	14 (4.1)	8 (11.1)
Stage	1	66 (10.8)	30 (14.4)	0 (0.0)	6 (12.0)	26-32 (7.7–9.4)	<5 (1.4–5.6)
	2	18 (2.9)	12 (5.8)	<5 (6.7–26.7)	5 (10.0)	0 (0.0)	6-12 (8.3–16.7)
	3	14-17 (2.3–2.8)	6 (2.9)	0 (0.0)	5 (10.0)	<5 (0.3–1.2)	9 (12.5)
	4	<5 (0.2–0.7)	<5 (0.5–1.9)	<5 (6.7–26.7)	0 (0.0)	<5 (0.3–1.2)	<5 (1.4–5.6)
	Unknown/not appropriate	205 (33.6)	64 (30.8)	<5 (6.7–26.7)	17 (34.0)	122 (36.0)	25 (34.7)
	Missing	304 (49.8)	95 (45.7)	10 (66.7)	17 (34.0)	183 (54.0)	24 (33.3)
Basis of diagnosis	Histology	573 (93.8)	203 (97.6)	11–14 (73.3–93.3)	45–49 (90.0–98.0)	309 (91.2)	68–71 (94.4–98.6)
	Other	34–37 (5.6–6.1)	<5 (0.5–1.9)	<5 (6.7–26.7)	<5 (2.0–8.0)	26–29 (7.7–8.6)	<5 (1.4–5.6)
	Unknown	<5 (0.2–0.7)	<5 (0.5–1.9)	0 (0.0)	0 (0.0)	<5 (0.3–1.2)	0 (0.0)
Prior cancer	No	532 (87.1)					58 (80.6)
	Yes	79 (12.9)					14 (19.4)
Charlson comorbidity index	0	379 (62.0)					55 (76.4)
	1	34 (5.6)					<5 (1.4–5.6)
	2+	33 (5.4)					<5 (1.4–5.6)
	Missing	165 (27.0)					9 (12.5)

^aDue to CPRD’s policy, the results were presented in a way to prevent deductive disclosure by small cells. This included not presenting values for certain stratifications.

^bOne patient received two treatments on the same day and appears in two treatment subgroups.

In patients with DDLPS, information on grade and stage of the disease was missing (including unknown/not appropriate) in a large proportion of cases and 19% of patients had prior cancer diagnosis. The majority (75%) of DDLPS patients had CCI of 0 whilst this information was missing for 12.5% of patients (Table 3).

Following diagnosis, the median waiting time to first treatment was 52 days (IQR 84, 26–110 days) in the overall cohort, and 67 days (IQR 78, 38–116 days) for patients with DDLPS (Appendix C.3). Surgery had the lowest median waiting time (44 days: IQR 75, 23–98 days). The mean duration of treatment (for patients receiving chemotherapy or radiotherapy) was 36.6 days overall (Appendix C.4).

Of the patients in Sub-cohort 2, 605 were included in the survival analysis as the ONS death registration death date for six patients was recorded on their index date (diagnosis date). Overall, patients with LPS had 77% probability of survival at 5 years (Table 4). The survival probability decreased with increasing age and for patients with a Stage 3/4 tumour. Patients with a Grade 4 tumour had a survival probability of 27% 1 year following their diagnosis record.

Table 4.

The survival rates of patients with liposarcoma from diagnosis record.

		Median follow up (in years)	IQR		Survival rate - 1 year (with 95% CIs)	Survival rate - 3 years (with 95% CIs)	Survival rate - 5 years (with 95% CIs)
Total number of patient alive on diagnosis date a	N = 605
All patients		4.42	2.03	8.87	0.92 (0.89–0.94)	0.81 (0.78–0.85)	0.77 (0.73–0.80)
Gender	Male	4.80	2.13	8.93	0.94 (0.91–0.96)	0.83 (0.79–0.87)	0.79 (0.74–0.84)
	Female	4.15	1.65	8.68	0.89 (0.84–0.92)	0.79 (0.73–0.84)	0.73 (0.67–0.79)
Age group	18–29 years	4.28	3.44	9.97	1.00 (−)	1.00 (−)	0.83 (0.27–0.97)
	30–39 years	7.30	3.98	13.58	0.97 (0.80–1.00)	0.97 (0.80–1.00)	0.97 (0.80–1.00)
	40–49 years	5.92	2.82	11.63	0.99 (0.90–1.00)	0.91 (0.80–0.96)	0.91 (0.80–0.96)
	50–59 years	5.80	2.74	10.64	0.97 (0.91–0.99)	0.93 (0.87–0.96)	0.90 (0.82–0.94)
	60–69 years	4.60	1.92	7.81	0.96 (0.91–0.98)	0.80 (0.72–0.86)	0.78 (0.70–0.84)
	70–79 years	4.24	1.80	7.84	0.91 (0.85–0.95)	0.76 (0.67–0.83)	0.68 (0.58–0.76)
	80+ years	1.89	0.51	4.07	0.67 (0.55–0.77)	0.55 (0.43–0.66)	0.44 (0.30–0.56)
Grade	1	5.16	2.69	9.54	0.97 (0.94–0.98)	0.91 (0.87–0.94)	0.87 (0.82–0.91)
	2	2.34	1.19	4.02	0.96 (0.75–0.99)	0.63 (0.39–0.80)	0.46 (0.21–0.68)
	3	1.40	0.66	4.39	0.73 (0.51–0.86)	0.47 (0.27–0.65)	0.40 (0.20–0.60)
	4	0.54	0.24	0.69	0.27 (0.01–0.69)	0.27 (0.01–0.69)	—
	Unknown/not appropriate	5.43	2.03	9.92	0.89 (0.84–0.93)	0.77 (0.71–0.82)	0.72 (0.66–0.78)
	Missing	1.83	0.58	2.80	0.86 (0.68–0.95)	0.77 (0.55–0.89)	0.77 (0.55–0.89)
Stage	1	2.65	1.86	3.97	0.98 (0.90–1.00)	0.92 (0.80–0.97)	0.84 (0.57–0.95)
	2	2.73	1.24	3.33	0.94 (0.63–0.99)	0.88 (0.59–0.97)	0.70 (0.26–0.91)
	3	0.93	0.24	2.05	0.62 (0.27–0.84)	0.62 (0.27–0.84)	0.62 (0.27–0.84)
	4	1.61	0.85	2.10	0.75 (0.13–0.96)	—	—
	Unknown/not appropriate	4.03	1.89	6.00	0.93 (0.89–0.96)	0.83 (0.77–0.88)	0.80 (0.73–0.85)
	Missing	7.08	2.74	10.54	0.91 (0.87–0.93)	0.79 (0.74–0.84)	0.74 (0.69–0.79)
Dedifferentiated	No	5.13	2.37	9.56	0.93 (0.91–0.95)	0.84 (0.81–0.87)	0.80 (0.76–0.84)
	Yes	1.44	0.45	3.32	0.81 (0.69–0.89)	0.53 (0.38–0.66)	0.43 (0.27–0.58)

IQR: interquartile range.

^aThis analysis includes patients who have at least 1 day of follow-up from index date. Six patients were excluded as they died on or before their recorded index date (diagnosis date) based on the ONS Death registration date.

Ninety-one patients were included in the survival analysis from treatment initiation (the most common of which was surgery). At 4 years, patients with LPS who were treated had 73% probability of survival (Table 5). The observed rates for those with DDLPS were 0.85 (95% CIs 0.59–0.95) at 1 year, 0.63 (95% CIs 0.28–0.84) at 3 years and 0.63 (95% CIs 0.28–0.84) at 4 years. The survival probability was also lower for patients with a Stage 3/4 tumour (compared to other tumour stages).

Table 5.

The survival rates of treated patients with liposarcoma from treatment start date.

		N (%)	Median follow up in years (IQR)	Survival rate - 1 year (with 95% CIs)	Survival rate - 3 years (with 95% CIs)	Survival rate - 4 years (with 95% CIs)
Total number of cases		91	1.58 (0.47–2.98)	0.90 (0.80–0.95)	0.79 (0.65–0.88)	0.73 (0.54–0.85)
Gender	Male	51 (56.0)	1.63 (0.74–2.53)	0.91 (0.78–0.97)	0.82 (0.61–0.93)	0.82 (0.61-0.93)
	Female	40 (44.0)	1.47 (0.29–3.47)	0.88 (0.68–0.96)	0.75 (0.52–0.88)	0.64 (0.35–0.83)
Age group	18–39 years	5 (5.5)	1.96 (0.97–3.58)	1.00 (–)	1.00 (–)	1.00 (–)
	40–49 years	7 (7.7)	2.12 (1.52–3.96)	1.00 (–)	0.67 (0.05–0.95)	0.67 (0.05–0.95)
	50–59 years	17 (18.7)	0.82 (0.32–2.54)	0.89 (0.43–0.98)	0.89 (0.43–0.98)	—
	60–69 years	23 (25.3)	1.57 (1.07–2.20)	0.95 (0.69–0.99)	0.72 (0.40–0.89)	0.72 (0.40–0.89)
	70–79 years	23 (25.3)	1.20 (0.25–2.53)	0.83 (0.55–0.94)	0.83 (0.55–0.94)	0.83 (0.55–0.94)
	80+ years	16 (17.6)	2.34 (0.60–3.82)	0.86 (0.54–0.96)	0.76 (0.42–0.92)	0.57 (0.17–0.84)
Grade	1	38 (41.8)	2.22 (1.20–3.41)	0.97 (0.80–1.00)	0.91 (0.66–0.98)	0.78 (0.39–0.94)
	2	15 (16.5)	1.17 (0.63–3.20)	1.00 (–)	0.75 (0.30–0.93)	0.75 (0.30–0.93)
	3/4	12 (13.2)	0.68 (0.25–1.56)	0.50 (0.15–0.77)	0.38 (0.09–0.67)	0.38 (0.09–0.67)
	Unknown/not appropriate	12 (13.2)	1.30 (0.47–3.36)	0.89 (0.43–0.98)	0.71 (0.23–0.92)	0.71 (0.23–0.92)
	Missing	14 (15.4)	1.09 (0.43–1.80)	0.92 (0.54–0.99)	—	—
Stage	1	27 (29.7)	1.63 (0.29–2.53)	1.00 (–)	0.90 (0.47–0.99)	0.90 (0.47–0.99)
	2	17 (18.7)	1.57 (1.06–2.98)	0.93 (0.59–0.99)	0.73 (0.35–0.91)	0.73 (0.35–0.91)
	3/4	13 (14.3)	0.82 (0.50–1.62)	0.68 (0.29–0.88)	0.54 (0.18–0.80)	0.54 (0.18–0.80)
	Unknown/not appropriate	34 (37.4)	2.31 (0.57–3.77)	0.90 (0.71–0.97)	0.84 (0.61–0.94)	0.73 (0.42–0.89)
Dedifferentiated	No	65 (71.4)	1.96 (0.79–3.20)	0.92 (0.81–0.97)	0.83 (0.67–0.92)	0.76 (0.54–0.89)
	Yes	26 (28.6)	0.98 (0.32–1.70)	0.85 (0.59–0.95)	0.63 (0.28–0.84)	0.63 (0.28–0.84)

Kaplan-Meier plots for the survival rates are provided in Appendix D.

Discussion

This study is the first report exploring the epidemiology, treatments and outcomes of LPS patients based on data from one of the largest linked primary care databases in the UK.

This study primarily aimed to describe the demographics, tumour characteristics and survival of patients newly diagnosed with LPS. The majority of LPS patients were male and over the age of 60 years. LPS was the first cancer a patient was diagnosed with and the tumours were of the lower grade and stage when reported (though the tumour stage and grade information was missing for a high proportion of patients). Patients with DDLPS had similar characteristics as patients with other LPS subtypes, including myxoid liposarcoma. The likelihood of survival following diagnosis was worse for those with Grade 4 tumours (where this information was recorded).

For the secondary aims, we described the proportion of LPS patients by treatment, quantified first treatment duration following diagnosis and described the survival time of LPS patients following first treatment. Most patients did not have a first treatment recorded, and whilst patient age did not appear to be a driver for the high proportion of patients receiving no treatment (as reported by Tirotta et al. ²⁹ ), our results may be impacted by missing data. For those who did receive treatment, the most common first treatment was surgery, in keeping with surgical resection being the mainstay of localized disease management. ¹² DDLPS patients had a longer waiting time to first treatment (compared to other subtypes) and a greater proportion of patients with this type of tumour were treated with chemotherapy (Table 3). Those with Grade 3/4 tumours (where this information was recorded) had the lowest likelihood of survival following first treatment.

To date, only one other study in England (for data collected between 2013 and 2017) has reported epidemiology and survival data for STS and its main types. ⁸ Over the 5-year period, the reported 1-year and 5-years net survival rates for LPS were 90% and 81.5%, respectively. ⁸ Another report from the same authors investigated treatment patterns for sarcomas that occurred in the retroperitoneum, which included several STS types (but more commonly LPS and leiomyosarcoma), ³⁰ and outcomes associated specifically with surgery in England. ²⁹ The study found that types of treatment received (including surgery, radiation, chemotherapy and other care) varied considerably by age group and that surgery survival outcomes were much improved if treatment was received in a high-volume sarcoma centre. ²⁹

Our results were similar to those reported by Bacon et al. (in terms of age and survival rates ⁸ ) and we were able to additionally present further information on LPS patients including BMI and prior cancer history.

Strengths and limitations

There were several strengths to this study. The large sample size, general representativeness of the English population and longitudinal coverage ¹⁹ makes it a robust source for gathering RWE. Additionally, the data are based on real world practice and contain a wealth of clinical information that can enhance research. We also incorporated data from several sources in order to build a more comprehensive picture of the characteristics and outcomes for LPS patients. Lastly, the study is authored by experts in fields of LPS and conducting studies using RWD.

However, limitations in the study should also be acknowledged. Despite the increased recognition of the utility of RWD studies and RWE in supporting drug development and approvals, there are specific requirements on data reliability and relevance. ⁹ Several examples (including our study) have highlighted the challenges associated with conducting large RWD studies in rare diseases and in sarcomas specifically. In the main, these were attributed to disease heterogeneity, variability in coding and data completeness and quality. This underlines the need for improved data collection and reporting.^8,9,31 For example, while there was much information on the characteristics of LPS patients in this study, there were areas of limited data (e.g., stage, grade) which are vital in improving knowledge of the disease and the development of new effective treatments.

Additionally, we may have missed eligible patients from the study due to the cohort definition. In this study we used the CPRD database and the ICD-10 C49 code for identifying patients with LPS, regardless of anatomical location. In contrast, the Bacon et al. study extracted disease classification data from the larger but unlinked national NCRAS database and ICDO-3 disease classification, which includes the original disease anatomical as well as morphology and behaviour codes. ⁸ It is also noted that identifying STS patients (specifically liposarcoma patients) using ICD coding is limited due to the inconsistency in the coding used for diagnosis and treatment for this form of cancer. ³² Future studies should aim to use a definition which permits a much broader capture of patients. Additional analyses stratified according to anatomical site and histological type should also be considered to better understand outcomes in patients whose disease occurs in complex anatomical regions. However, when using real world data, these analyses have to be performed in compliance with local regulations surrounding data protection to prevent re-identification of patients with rare characteristics or conditions.

There may also have been misclassification in terms of LPS subtypes or confirmation of a LPS diagnosis (e.g., erroneous coding, missing information). The incomplete data from stage and grade, though consistent with variations in sarcoma data capture as previously reported, ⁸ also limited our ability to report on these characteristics. There was also incomplete information on other characteristics (e.g., BMI). Furthermore, the study results will be specific to England and therefore may not be generalisable to populations based in other countries due to differences in demographics or healthcare systems.

Conclusion

LPS is a challenging type of cancer that, until recently, has had limited epidemiological data, particularly in England. Our research indicates that LPS is a rare cancer with a significant treatment burden. We observed a decrease in survival rates with increasing patient age and advancing disease grade and/or stage. Furthermore, our study underscores the complexities inherent in collecting and reporting data for rare diseases such as LPS. Consequently, the enhancement and standardisation of data reporting, especially for rare heterogeneous cancers such as STS and LPS, is of paramount importance. This will optimise the use of patient data registries and RWE, ultimately leading to improved patient outcomes.

ORCID iD

Jessie O. Oyinlola https://orcid.org/0009-0000-1099-4035

Ethical statement

Ethical approval

CPRD has overarching ethics approval from the Health Research Authority (HRA) to support research using anonymised patient data for public health benefit. Once CPRD receives anonymised data from a GP practice, the data is fully compliant with the Information Commissioner’s Office (ICO) anonymisation code of practice and patient privacy is protected. Requests by researchers to access the data are reviewed via the MHRA CPRD RDG process to ensure that the proposed research is of benefit to patients and public health. More information is available on the CPRD website: https://www.cprd.com/safeguarding-patient-data. The study received research governance approval from the MHRA CPRD’s RDG on 20^th January 2022 (RDG Protocol number 21_000628).

Consent to participate

Informed consent from patients is not required when working with CPRD primary care and linked data. This is because the data made available for access by approved researchers is effectively anonymised. CPRD has favourable ethical opinion from the East Midlands-Derby NHS Research Ethics Committee (REC) to provide coded and anonymised data for public health research. More information can be found here: Clinical Practice Research Datalink (CPRD) Research Database - Health Research Authority (hra.nhs.uk).

Data Availability Statement

It is not possible to share data from this study as access to CPRD data, including UK Primary Care Data, and linked data such as Hospital Episode Statistics, is subject to protocol approval via CPRD’s Research Data Governance (RDG) Process. Further information can be found on the CPRD website (https://www.cprd.com/data-access).