Hyperprogression and impact of tumor growth kinetics after PD1/PDL1 inhibition in head and neck squamous cell carcinoma

Background: Hyperprogressive disease (HPD) rate in head and neck squamous cell carcinoma (HNSCC) patients treated with immune checkpoint inhibitors (ICI) was determined using tumor growth kinetics (TGK) and compared with rapidly progressive screen-failure (SF) patients. The impact of TGK on outcomes with salvage chemotherapy (SCT) was also evaluated. Results: HPD was found in 22/120 (18%) patients. Median TGK before the onset of immunotherapy (TGKpre) was 2.7 for SF patients and 4.8 for HPD patients, with no significant difference (p = 0.17). Disease control rate after initial progressive disease on ICI was 86% with SCT in case of tumor growth deceleration vs 39% in case of tumor growth acceleration. Conclusions: HPD was frequent, but TGK of HPD patients treated with ICI did not differ from SF patients, suggesting that there is no relevant causal relationship between HPD and ICI. After initial PD with ICI, tumor growth deceleration was associated with better outcomes, indicating that TGKR might be useful to detect late responders, meriting prospective investigations. Materials and Methods: TGK ratio (TGKR) was defined as the ratio of TGK on ICI (TGKpost) to TGKpre. HPD was defined as TGKR ≥ 2. TGKR >1 indicated tumor growth acceleration, while 0 < TGKR < 1 indicated tumor deceleration.


INTRODUCTION
Head and neck squamous cell carcinoma (HNSCC) is an aggressive epithelial cancer that derives from mucosa linings of the oral cavity, oropharynx, hypopharynx or larynx. It is also a major public health problem with it being the sixth most incident cancer worldwide, responsible of more than 700 000 cases every year and around 350 000 deaths [1]. Different methods are employed by this aggressive disease to avoid immune recognition, including direct T-cell suppression with soluble or surface inhibitory factors like Programmeddeath ligand 1 (PDL1), and the recruitment of immunosuppressive cell populations [2].
Allison first hypothesized that overcoming the anergic state of T-lymphocytes is possible via the blockade of co-inhibitory signals [3] and research has initially been focused mainly on the PDL1-PDL2/PD1 axis. Immune Research Paper www.oncotarget.com checkpoint inhibitors (ICI) have since completely revolutionized the treatment of Recurrent/Metastatic (R/M) HNSCC [4][5][6].
Primary resistance to ICI is frequent across all tumor types, including HNSCC, and concerns almost 60% of patients overall [7]. In a worrying manner, it was suggested that some patients even experience an acceleration of tumor growth kinetics (TGK) on immunotherapy [hyperprogressive disease (HPD)] [8]. Medical charts of 34 patients treated with PD1/PDL1 inhibitors from four different institutions were retrospectively reviewed in 2017 and HPD was found to be frequent (29%) and associated with a worse outcome [8]. Other studies found HPD in different tumor types with varying rates [9,10] but no consistent predictive genomic or clinical characteristic was associated. All were of a retrospective nature, without a control arm. In consequence, a causal relationship of HPD to ICI has not been proven and a natural evolution of the disease cannot be excluded in the observed cases. Many preclinical studies have hypothesized mechanisms, but no clear biological explanation has seen the day.
Furthermore, after failure of ICI, salvage chemotherapy (SCT) is usually the treatment of choice, but not a lot of data is available on the outcomes in this context. When facing an initial RECIST (Response Evaluation Criteria In Solid Tumors) progressive disease (PD) on ICI, physicians are often confronted with the choice between starting subsequent SCT or continuing ICI in hope for a late response, especially if patients do not experience a worsening general condition. The utility of TGK in this situation is unknown.
Our aim in this study was to know if HPD patients had slower tumor growth before treatment with ICI compared to patients with a naturally exponential growing disease, in which case a strong argument for a causal relationship between ICI and HPD can be made. For this, we used TGK to determine the rate of hyperprogression in clinical trial patients and then compared tumor growth before the onset of immunotherapy with rapidly deteriorating screen-failure (SF) patients. The impact of TGK on outcomes with SCT after initial PD with ICI was also evaluated.

RESULTS
In total, there were 192 patients in 9 clinical trials testing anti-PD1/PDL1 agents alone or in combination with anti-CTLA4 or anti-KIR antibodies. Among those, 158 patients were treated with ICI (the other 34 patients received exclusive chemotherapy) and 120/158 patients were eligible for HPD analysis. The remaining 38 patients were not included in the final analysis because of the absence of available pre-baseline imaging (2 patients) and the absence of post treatment imaging (36 patients) ( Figure 1). Median follow-up time since the start of immunotherapy was 34.3 months (95% CI, 32.2 to 35.5).  (Figure 3).

SF tumor growth kinetics comparison
In total, 65 patients were screen-failed in the 9 clinical trials. Of these, 50 SF cases were attributed to rapid clinical deterioration and were included in the final analysis ( Figure 1). The following reasons were the cause of SF in the included patients: death, symptomatic cerebral metastases, elevated liver enzymes attributed to metastatic disease, corticosteroid use for disease control and worsening general condition. 46/50 patients were eligible for TGK pre assessment as 1 patient was deceased, 1 patient was lost to follow up and 2 patients didn't have an available CT-scan.

TGKR after initial progression on checkpoint inhibitors
Out of 39 patients who presented initial RECIST 1.1 PD with ICI and were subsequently treated with salvage chemotherapy, 32 patients were eligible for TGK R assessment. 7 patients were ineligible because of the absence of pre-baseline scan. Seven (7) out of 14 (50%) patients with disease deceleration (TGK R < 1) on ICI had PR or CR and 5 (36%) had SD. DCR was 86%. 3 out of 18 (17%) patients with disease acceleration (TGK R > 1) on ICI had PR and 4 (22%) had SD. DCR was 39% ( Figure 5). Median time from last ICI administration to first imaging on salvage chemotherapy was 2.3 months (95% CI, 2 to 2.7).

Hyperprogression: a real phenomenon?
TGK did not differ between rapidly deteriorating SF and ICI-treated HPD patients, strongly suggesting that there is no relevant causal relationship between HPD and ICI, at least in some cases. In other words, patients that are considered to have ICI-induced hyperprogression have a tumor growth before ICI treatment that is similar to patients that have a rapidly evolving disease, not treated with ICI. Exponentially growing tumors are expected to continue to grow exponentially when untreated.
In our study, we used TGK to evaluate hyperprogression as this was the methodology used in the first published work on HPD in HNSCC by Saâda-Bouzid et al [8]. If another definition was used [e.g., 50%  Tumor growth ratio (TGR) in 4 weeks], results might have been different, highlighting the need of an international consensus to define hyperprogression. More recently, HPD evaluation in non-small-cell lung cancer (NSCLC) patients was highly concordant using TGK and TGR which suggests the interchangeability of these two methods [11].
Multiple mechanisms have been raised: oncogenic signaling activation [11], upregulation of alternative immune checkpoints [12], major immune reactions caused by PD1/PDL1 inhibitors, previous irradiation [8], tumor proliferation via a direct (DNA damage with free radicals) or indirect (angiogenesis and tissue  remodeling promotion) effect [13][14][15], expansion of PD1expressing T-regulatory cells and modulation of tumourpromoting cells [16]. Interaction with tumor-associated macrophages via the Fc domain of the anti-PD1 antibody with reprogramming into immunosuppressive M2like macrophages has also been incriminated [17]. No consistent predictors have yet been validated, further supporting our results.
On the other hand, 36 patients in our cohort were excluded because of the absence of post-baseline imaging that might be due to in some cases to hyperprogressionrelated death. This may lead to underestimating the phenomenon, although some of these deaths might also be due to disease natural behavior.
The absence of randomization and the difference in screening windows could of course be a source of bias, but Overlapping confidence intervals of this dot plot show that distribution is similar. Figure 5: Impact of Tumor Growth Kinetics ratio (TGKR) on outcomes with salvage chemotherapy (SCT) after initial RECIST 1.1 progression with immune checkpoint inhibitors (ICI). Patients were more likely to respond to SCT in case of tumor growth deceleration (0 < TGK R < 1) and had higher disease control rate (DCR) than those with tumor growth acceleration (TGK R > 1). we believe that our comparison with rapidly deteriorating SF patients is pertinent as those are the patients that might best represent a naturally exponential growing disease.
Furthermore, the heterogeneity of the immunotherapy regimens in this study could also be a confounding factor, but a previous work evaluating HPD in earlyphase immunotherapy trials found hyperprogression with combination ICI as well as monotherapy, with no significant difference [18], mirroring our results.
We chose to compare growth kinetics before starting immunotherapy (TGK pre ) between the two groups (and not the ratio of TGK post to TGK pre ), as we thought that this was the most appropriate time interval to show similarity in the speed of disease progression. SF patients were often lost to follow-up, received ICI in another setting or had variable timing in CT-scan evaluation after screen-failure, so comparison of the ratio of TGK post to TGK pre would have been less appropriate.
In addition, we have found that HPD evaluation with total tumor burden and with RECIST 1.1 was concordant for about seventy percent of patients only, further emphasizing the need to find a consensus in HPD diagnostic criteria. This is the largest cohort evaluating HPD in HNSCC to date. Our findings underscore the difficulties in interpreting the evolution of the disease under ICI and present an important message to clinicians that are treating patients with R/M HNSCC.
We found that hyperprogression was frequent (18%) and correlated only with a high NLR. Previous retrospective studies (albeit with less included patients) have already shown a correlation of high NLR with poor outcomes [19,20], and so our results confirm this literature data, even if the relative small number of HPD patients might have weakened the statistical power of the comparison.

Impact of tumor growth kinetics
Some chemotherapeutic agents have been shown to exert immune-reactive events such as upregulation of MHC (major histocompatibility complex) class molecules and increased antigen presentation [21,22]. Chemotherapy can augment tumor immunity by decreasing the number of Immunosuppressive cells like myeloid-derived suppressor cells (MDSC) or T-regulator cells in the microenvironment which can lead to an accumulation of helper T-cells on site [23,24] and by promoting anti-tumor CD4+ T-cell phenotype [25]. In fact, Improved responses to chemotherapy has been reported after vaccination immunotherapy in various tumor types [26].
Our results confirm that salvage chemotherapy seems to be more effective with better outcomes (ORR, PFS, and OS), suggesting that chemotherapy is enhanced by immunotherapy. Indeed, ORR was 28%; which is higher than the historical controls before the era of immunotherapy; varying between 6 and 24% [27][28][29].
By showing that patients who presented initial PD on ICI associated to a tumor deceleration (TGK R < 1) had higher ORR that those with tumor acceleration (TGK R > 1), we think that some of the responses in our study were due/enhanced by circulating anti-PD1/PDL1 antibodies causing delayed onset of response. The median time from the last immunotherapy administration and the first evaluation on salvage chemotherapy was 2.3 months, which is compatible with the half-life of these agents [30,31]. We by consequence also believe that TGK R might be a useful tool to help select potential late responders to ICI and thus avoid toxic and possibly inefficient chemotherapy.
Of course, this hypothesis should be evaluated in other data sets and if confirmed, it could potentially be tested prospectively by randomizing to continued immunotherapy versus switch to salvage chemotherapy in case of tumor growth decelaration.
In conclusion, HPD was found in 18% of cases and correlated with high NLR. Growth kinetics before ICI-treated HPD patients were similar to SF patients, suggesting that the published rates of HPD might be due to natural disease behavior, at least in some cases. After initial RECIST PD with ICI, tumor growth deceleration was associated with better outcomes compared to tumor growth acceleration, indicating that TGK R might be useful to detect late responders and avoid SCT, meriting prospective investigations.

Patient selection and data collection
All patients with a histologically confirmed R/M HNSCC treated in Léon Bérard cancer center in a clinical trial testing PD1/PDL1 antibodies alone or in combination with an anti-Cytotoxic T-Lymphocyte Antigen 4 (CTLA4) or an anti-Killer Immunoglobulinlike Receptor (KIR) antibody between March 2014 and November 2018; were included ( Figure 1). All imaging was retrospectively reviewed by medical oncologists and independent radiologists. SCT was defined as the first line of chemotherapy administered after failure of ICI. The following data was collected and recorded: age, gender, primary tumor location, tobacco use, human papilloma virus (HPV) status (p16 immunostaining and/or DNA in situ hybridization), previous multimodal therapy at the initial stage, neutrophil-to-lymphocyte ratio (NLR) (0 to 72 hours before start of treatment; with a cut-off ≥5 defining high NLR), antibiotics intake in the 3 months prior to immunotherapy, the number and dates of previous and following lines of systemic therapy. Also collected was the best overall response on and after failure of immunotherapy using RECIST 1.1, progression-free survival (PFS) and overall survival (OS).
Samples were provided by the local Biological Resources Center (BB-0033-00050, CRB Centre Léon Bérard, Lyon France). 4-μm thick tissue sections of FFPE tissue were prepared according to conventional procedures. For each sample, hematoxylin and eosin (HES) staining was performed to determine the number of tumor cells. Immunohistochemistry (IHC) was performed on an automated immunostainer (Ventana Benchmark ultra, Roche, Meylan, France) using Ultra View DAB Kit according to the manufacturer's instructions. Sections were incubated with a rabbit monoclonal human anti-PDL1 Ab (diluted at 1:50, Quartett, Berlin, Germany) clone QR1. The Ventana amplification kit was used and an anti-rabbit-HRP was applied on sections. Staining was visualized with DAB solution with 3,3′-diaminobenzidine as a chromogenic substrate. Finally, the sections were counterstained with Gill's hematoxylin. All samples were examined by a qualified anatomopathologist for combined positive score (CPS), defined as the number of PD-L1-positive cells (tumour cells, macrophages and lymphocytes) divided by the number of tumour cells × 100 (a minimum of 100 viable tumour cells must have been present for the specimen to be considered evaluable). CPS ≥ 1 was the cut-off for PDL1 positivity.
The data collection cutoff point was December 4, 2019.

Hyperprogressive disease definition
In order to be eligible for TGK R assessment, patients had to have a pre-baseline scan, a baseline scan and posttreatment scan. Minimal delay between 2 CT-scans was 14 days and patients had to have started ICI therapy in the 6 weeks following baseline scan (Figure 1). TGK before (TGK pre ) and after (TGK post ) anti-PD1/PDL1 therapy were evaluated. TGK pre was defined as the difference of the sum of the largest diameters of the target lesions per unit of time between pre-baseline and baseline imaging [(S 0-S pre )/(T 0 -T pre )]. TGK post was defined in the same manner between on immunotherapy and baseline imaging [(S POST -S 0 )/(T POST -T 0 )]. HPD was defined as TGK R (ratio of TGK post to TGK pre ) ≥ 2. TGK R > 1 indicated tumor growth acceleration, while 0 < TGK R < 1 indicated tumor deceleration. TGK R < 0 indicated tumor shrinkage.
TGK R was calculated with RECIST 1.1 for all patients. Since TGK only evaluates the variation of target lesions and does not include new lesions in the assessment of tumor growth, TGK R was also calculated using total tumor burden (TTB).

Screen-failure inclusion
All patients that were screened for the same clinical trials and were subsequently ineligible were analyzed. Patients that were considered screen-failed because of rapid clinical deterioration attributed to disease progression were included in the 'screen failure' (SF) group. Patients were not included if they were subsequently treated with ICI in another clinical trial or setting. TGK pre was calculated in this group in the same manner in order to compare tumor growth with HPD patients before the onset of immunotherapy (Figure 1).

Statistical analysis
Descriptive statistics were used to summarize the baseline characteristics of the patients. Chi-squared or Fisher's exact test was used for statistical comparisons of categorical data. Mann-Whitney test was used to compare distribution. Partial response (PR) and complete response (CR) defined objective response (OR). Disease control rate (DCR) was defined as the sum of CR, PR, and stable disease (SD). PFS time was defined as the period from the date of initial treatment administration to the date of clinical disease progression, mortality from any cause or the last follow-up. OS time was defined as the period from the date of initial treatment administration to the date of mortality from any cause or the last follow-up. The Kaplan-Meier method was used to assess PFS and OS. Comparison was done using the log-rank test. Data of patients who were lost to follow-up were censored at the time of last contact. Statistical analysis was done using MedCalc 18.11.6 statistical software (MedCalc Software, Mariakerke, Belgium).