A retrospective observational study was performed. The study population consisted of all women with “high-risk” locally advanced cervical cancer who commenced NACT within WoSCAN between January 2012 and December 2016. Exclusion criteria was intent to proceed to definitive surgery. Patients were identified from the chemotherapy electronic prescribing and administration system (CEPAS). Follow up occurred until death, or the end of the study period (February 28, 2018), whichever occurred first.

Patients were staged according to International Federation of Gynaecology & Obstetrics (FIGO) 2009. Tumour and/or lymph node size on magnetic resonance imaging (MRI) was defined as maximum width on axial T2-weighted sequences. If MRI was unavailable, maximum cervical tumour width on pelvic examination was recorded. PET-CT was preferred to standard contrast enhanced CT for assessment of distant metastases. Surgical lymph node staging was not performed.

Data were collected from information gathered within CEPAS and ARIA, a radiotherapy management system; Clinical Portal, an electronic application providing socio-demographic information and details of treatment outcomes; and the Acute, Cancer, Deaths and Mental Health (ACaDMe) datamart [14], to obtain death records. Data were entered on a Microsoft Access database and anonymised. Statistical analysis was performed using R software, version 3.3.3 (R Foundation, Vienna, Austria) [15].

Median overall survival (OS) along with 95% Confidence Intervals (CI) were estimated using the Kaplan–Meier (KM) method. For OS and relapse, the date of commencement of NACT served as the start date and February 28, 2018, served as the censor date for those still alive at study end.

Cox proportional-hazard models were used to estimate unadjusted hazard ratios for survival, for the following clinical variables: definitive treatment (CCRT or other); age; baseline performance status (0 versus 1–3); FIGO Stage; para-aortic nodes (yes/no); type (squamous cell carcinoma or other) and grade of pathology (1 versus 2 versus 3); baseline albumin and haemoglobin (lower than normal range versus within normal range); baseline platelets (higher than normal range versus within normal range); Neutrophil lymphocyte ratio (NLR) (5 versus $\ge$5) [16]; Charlson comorbidity index (CCI) score [17]; Scottish Index of Multiple Deprivation (SIMD) (2012) [18]; and NACT regimen (cisplatin/paclitaxel versus carboplatin/paclitaxel versus cisplatin alone). Adjusted models were then created including age and significant variables from univariable analyses (p 0.1).

A total of 126 patients were included in the analysis, and the observed median follow up time was 30 months (Inter quartile range (IQR) 16.4–43.8). The median age was 44 years (IQR 34–54) and the majority (85.7%) had squamous cell carcinoma. 99.2% were staged with either MRI or PET-CT scan (MRI 96.0%, PET-CT 92.1%, both 89.7%). Primary tumour $\ge$5 cm was recorded in 78.6% of patients. Positive pelvic nodes and para-aortic lymph nodes on imaging were identified in 77% and 30% of women, respectively. Over 90% had $\ge$stage II disease and 27% had stage IV disease. Of the stage I patients, 8/9 had FIGO 1B2 disease with primary tumour $\ge$5 cm and 6/9 had nodal involvement. Table 1 shows the baseline characteristics.

The most common NACT regimen was 3-weekly cisplatin/paclitaxel (62.7%) followed by 3-weekly carboplatin/paclitaxel (34.9%); median number of cycles received was three (IQR 3–4) (Table 2). There were no chemotherapy related deaths.

CCRT was scheduled to commence on day 21 following the final NACT cycle. External beam radiotherapy (EBRT) was CT-planned using 3-D conformal techniques and 45–50 Gy delivered with a 4-field brick, superseded by volumetric arc therapy (VMAT) at the end of 2012. Weekly cisplatin 40 mg/m${}^2$ was administered concurrently for 5–6 weeks. High dose rate (HDR) intracavitary brachytherapy (BT), 24 Gy in 4 fractions prescribed to point A, was incorporated in weeks 5 and 6. After treatment, patients were followed up 3–4 monthly and a pelvic examination was performed. Imaging was requested according to clinician preference and/or at the time of suspected relapse.

The proportion of patients who proceeded to radical treatment was 96.8% (122/126) with 108 patients (85.7%) receiving CCRT and 14 (11.1%) receiving RT alone. The median time from last cycle of NACT to start of radiotherapy was 27 days (IQR 24–34). The remaining patients defaulted from treatment, had progressive disease and were unable to receive radical treatment, or were lost to follow up. Of the 122 radically treated patients, 116/122 had EBRT plus BT and the remaining 6/122 had EBRT +/– photon boost. BT was not administered to these 6 patients as a result of either patient refusal, compliance or suitability for the procedure. Treatment was completed as planned in 105/122 patients. There was one death during radiotherapy which was unrelated to treatment. Mean EQD2 to point A based on $\alpha$/$\beta$ ratio of 10 was 77.2 Gy (range 10–95.9). Median dose to point A of 82 Gy was obtained in 54/122 patients (Table 3).

Patients who received RT alone or no definitive therapy were older and more likely to have stage IV disease and/or positive para-aortic lymph nodes. See supplementary file (Supplementary Table 1) for more details.

The Kaplan-Meier plot for OS is shown in Fig. 1. At the study end (median follow up of 30 months), 49 patients had died and the primary cause of death was cervical neoplasm in 90% of cases. There were no treatment related deaths. Median OS was not reached (NR) (95% CI 41.4 months – NR) but the 3-year OS was 61.8% (95% CI 53.4–71.6%).

Fig. 1.

Kaplan-Meier plots for Survival. (a) Overall survival. (b) Overall Survival by definitive treatment — Concurrent chemoradiotherapy vs. Other. Abbreviations: CCRT, concurrent chemoradiotherapy.

Using univariable Cox proportional hazard models, the following factors had a significant negative influence on survival: definitive therapy other than CCRT, age, FIGO stage, positive para-aortic nodes, performance status $\ge$1, baseline albumin and haemoglobin normal reference range, baseline platelets $>$normal reference range, NLR $\ge$5, and CCI score $\ge$1.

Using multivariable analysis, the factors that remained independently associated with poorer survival were: definitive therapy other than CCRT (Hazard Ratio (HR) 2.23 (95% CI 1.01–4.91) and NLR $\ge$5 (HR 2.8 (95% CI 1.32–5.90)) (Table 4).

Two patients were lost to follow up. Of the remaining 124 patients, 50 (40.3%) had relapsed by the end of the study period, of whom 31 women (25%) went on to receive subsequent treatment. The crude local control rate was 75.8% (30/124 cases with persistent or recurrent cervical disease). The pelvic control rate was 68.5% (39 locoregional failures). Relapse restricted to the pelvis occurred in 21 patients. Isolated distant relapse was documented in 7 patients and in combination with pelvic and/or para-aortic failure in a further 13 patients, resulting in total distant failure rate of 16% (20/124). Of all relapses, 40% (20/50) had distant disease. Five women were salvaged with surgery or radiotherapy, for central pelvis or para-aortic nodal recurrence, respectively. A further 26 patients had treatment with palliative intent.

We report 3-year OS of 61.8% in a cohort of women with very locally advanced cervical cancer who received 3-weekly platinum based NACT prior to CCRT/RT. Both the definitive therapy received following NACT and baseline NLR were found to significantly impact upon survival when adjusted for other potential cofounders. The vast majority of patients proceeded to radical treatment (CCRT 85.7%/RT 11.1%). Distant failure was lower than predicted at 16% but pelvic control was 68.5%.

Evidence supporting the use of NACT prior to RT/CCRT is limited, with most data comprising retrospective case studies from India [19-21], or small phase II trials [5,22-24] that often had short follow up periods and reported response rate rather than survival. Until very recently, there was no randomised evidence available. CIRCE was a two-arm, randomised, open-label phase II trial undertaken within a single centre in Brazil which reported findings in 2019. This study randomised 107 patients to either NACT with 3-weekly cisplatin and gemcitabine followed by CCRT or CCRT alone [25]. Similar to our study, it reported a 3-year OS of 60.7% in the 55 patients who received NACT, although OS was found to be markedly lower compared with those who received CCRT alone (3-year OS 86.8%). CIRCE included predominantly stage IIB and IIIB patients; very few women had stage IV disease. Only 9% in the neoadjuvant arm were recorded as FIGO stage IVA in comparison to 27% stage IV (including 4% stage IVB) in our cohort. Unfortunately, the number of patients who had para-aortic lymph node involvement was not reported as this would have allowed a better comparison with our study. Furthermore, NACT regimen was different with cisplatin/gemcitabine forming the chemotherapy backbone. Extrapolating from OS data comparing various platinum doublets in patients with relapsed and/or metastatic disease [26], the choice of gemcitabine or paclitaxel is unlikely to be a determining factor. Similarly, carboplatin is not inferior to cisplatin when combined with paclitaxel in the advanced setting [27]. In our population, there was no survival difference detected in those receiving carboplatin/paclitaxel compared to cisplatin/paclitaxel.

Other studies investigating NACT have utilised alternative regimens. The CXII trial [5], for example, used carboplatin and paclitaxel but it was given as six weekly doses in comparison to our study (3-weekly scheduling). The selection of weekly treatment in CXII is based on a historical meta-analysis that suggested dose density and short intervals (14 days) between chemotherapy may be the key to improve outcomes if NACT is delivered prior to radical treatment for locally advanced cervical cancer, although radical treatment predominantly consisted of RT alone as the data was collated before CCRT was widely accepted [28]. The survival in our study is similar to that reported from CXII, with 3-year OS of 61.8% (95% CI 53.4–71.5) versus 67% (95% CI 51–79) with CXII [5]. Although OS was slightly poorer, it is unlikely that 3-weekly treatment is significantly detrimental compared to weekly scheduling as our cohort had poorer prognostic features than the CXII population. The majority of cases in CXII consisted of Stage IIB or IIIB disease with Stage IVA representing only 7% of the study group (Stage IVB was excluded) [5]. Interestingly, the 3-year OS reported within the CCRT arm of CIRCE is higher than that described by the larger multicentre cohort study, RetroEMBRACE, which reported a 3-year OS of 74% using state of the art RT techniques, although only 77% received concomitant chemotherapy [4]. Half of patients in RetroEMBRACE were noted to have Stage IIB disease. Less than 5% had Stage IV disease, including a small number of stage IVB patients (0.7%) [4]. Our local control rate of 75.8% contrasts with 91% at 3 years from RetroEMBRACE [4], suggesting that overall RT dose was suboptimal. However, there were crucial differences in patient characteristics, particularly tumour stage and size. The majority of our patients (almost 80%) had documented primary tumour at diagnosis in excess of 5cm. Local control at 3 years drops from 95% in tumours 5 cm to 85% in tumours $\ge$5 cm with an associated reduction in OS from 81% to 66% at 3 years, respectively [4]. Further, most of our patients had involved nodes (77%), contrasting with 40% in RetroEMBRACE. This is pertinent for two reasons — firstly, node positivity influences OS (78% at 3-years in node negative population compared with 67% in node positive population) [4]; and secondly, positive nodes can be a predictor for distant metastases [29,30]. Allowing for the more advanced nature of our study cohort, the 3-year OS of 61.8%, compares more favourably with the results from RetroEMBRACE. NACT should not be considered as a substitute for effective RT techniques, especially IGBT, and does not appear to augment local control, but NACT may reduce micro-metastases in a group with very high risk disease. The rate of distant relapse in our study was 40% in patients with confirmed recurrence as opposed to 80% in RetroEMBRACE [30], despite the fact that there was a significant variance in node positivity rates at diagnosis.

The main strength of this analysis is the inclusiveness in terms of study participants who received NACT for locally advanced cervical cancer in the era of modern conformal RT planning. However, as a retrospective series, the added benefits of NACT prior to CCRT cannot be evaluated since there was no control arm. The principal limitation of our study is the evolution of RT practice. For instance, the pelvic EBRT dose was increased from 45 Gy to 50 Gy due to concerns over pelvic control rates, and VMAT was introduced in 2012. CT was employed to delineate organs at risk at the time of BT, but dose was prescribed to Point A as per the historical Manchester system. Thus, the recommended ESTRO guideline [31] that stipulates volume based planning was not adhered to. Also, we did not have the facility for combined intracavitary/interstitial BT which was performed in almost 25% patients in RetroEMBRACE [4]. Furthermore, toxicity information relating to both chemotherapy and radiotherapy are not routinely captured electronically in our systems and therefore were unavailable for this study. However, in a prior audit, NACT was found to be reasonably well tolerated [32].

As highly developed radiotherapy techniques become increasingly available, and the role of IGBT firmly established in optimising local control, there is a drive to enhance pelvic control rates by characterising elective nodal volumes and implementing simultaneous integrated boost to target involved nodes [33]. It has also been postulated that escalating dose to eradicate nodal burden will reduce distant relapse and/or improve OS. Conversely, some emerging data refutes this [29,34], suggesting that additional SACT is required. The potential role of adjuvant immunotherapy is currently under investigation in the CALLA trial [35]. As yet, it is unclear whether this will prove to be a successful strategy and, if so, which subgroups will benefit most, as recruitment is likely to continue until 2024. NACT has the advantage of being low cost and readily available. Ultimately, it would be helpful to identify women at highest risk of poor outcome based on clinico-pathological factors and/or molecular markers and direct them to the appropriate therapeutic pathway. The prognostic significance of NLR also warrants further consideration as per previous findings [36], higher baseline NLR was associated with shorter survival.