Adjuvant intraperitoneal chemotherapy for the treatment of colorectal cancer at risk for peritoneal carcinomatosis: a systematic review

Abstract

Background: The peritoneal surface is the second most common site of disease recurrence, after the liver, following definitive surgery for colorectal cancer. Adjuvant intraperitoneal (IP) chemotherapy delivered at time of surgical resection has the potential to delay or prevent future spread to the peritoneal surface and improve clinical outcome. The exact role of adjuvant IP chemotherapy in colorectal cancer, including its associated morbidity and mortality, is not well defined.

Study design: Systematic review and pooled random effect analysis of comparative trials examining the addition of adjuvant IP chemotherapy compared to surgery alone in colorectal cancer. The primary outcome was overall survival, and the secondary outcomes were of post-operative morbidity and mortality.

Results: In nine colorectal cancer studies identified, seven were two-arm trials comparing adjuvant IP chemotherapy to surgery alone. Of these, four trials had outcome reporting and met criteria that allowed inclusion into a random effects model. Heterogeneity was measured by Cochran’s Q-test (Q = 13.9; p = 0.01) and random effect models were utilised. Pooling eligible trials together revealed a 0.55 odds ratio of death associated with the administration of IP chemotherapy compared to surgery alone (CI = 0.31, 0.98; p = 0.04). Trials selecting patients at elevated risk for the development of peritoneal carcinomatosis by clinicopathological biomarkers for administration of adjuvant IP chemotherapy reported more favourable overall outcomes. There was no increase in mortalities or IP chemotherapy-related abdominal complication rates among patients undergoing IP chemotherapy (OR = 1.4; CI = 0.52, 3.8; p = 0.5).

Conclusions: This systematic review supports the use of adjuvant IP chemotherapy in resectable colorectal cancer at risk for peritoneal spread. Future trials should seek to standardise inclusion criteria and IP chemotherapy modalities to better define the role of this treatment in patients with resectable colorectal cancer.

Keywords:

• Intraperitoneal chemotherapy
• colorectal cancer
• adjuvant regional chemotherapy
• gastrointestinal malignancies
• peritoneal carcinomatosis
• systematic review

Introduction

Colorectal malignancies are the most common gastrointestinal cancers in the United States. In 2016, approximately 135 000 new cases of colorectal cancers were diagnosed and nearly 50 000 of those were fatal [1]. Colorectal cancer is expected to remain the third most common cause of cancer-related death in the United States since 2016 [1]. Due to improved primary prevention and screening, in up to 40% of cases, colorectal cancer presents while still amenable to surgical removal [2]. While stages I and II colon cancer are associated with 5-year survival rates of 90% and 78%, mortality increases sharply thereafter with stages III and IV having 50%, and less than 5% of 5-year survivals [3,4]. Approximately 25% of colon cancer patients presents with distant disease, and the third of those not presenting with metastases eventually develops them [5,6]. This rate of recurrence persists in spite of adjuvant therapy chemotherapy, which is aimed at reducing recurrence and improving clinical outcome in stages II and III patients [4,7,8].

Therapies targeted to organs at high risk of recurrence are a current area of research focus, as part of an effort to improve patient selection for adjuvant treatment. The second most common site of recurrence in colorectal cancer is the peritoneum, which is involved in 25–35% relapses [9–13]. Patients who do develop peritoneal spread have a particularly dismal prognosis, with a shorter expected median survival compared to patients with non-peritoneal metastatic colon cancer [14]. Additionally, there is a disproportionally higher impact on quality of life associated with peritoneal recurrence due to the difficulties in managing GI failure with frequent intestinal obstructions. Thus, there appears to be rationale for the development of locoregional adjuvant therapy for the prevention of spread to the peritoneum, in particular in patients who are at high risk.

Cytoreductive surgery (CRS) in combination with intraperitoneal (IP) chemotherapy has become a clinically proven treatment in the management of overt peritoneal carcinomatosis from colorectal cancer. This trend follows the established management of peritoneal involvement of appendiceal cancer, ovarian cancer and other primary peritoneal surface malignancies [15–19]. There is also a growing body of evidence supporting (IP) chemotherapy as a preventative strategy in gastrointestinal (GI) malignancies, including gastric cancer [20,21]. Preclinically, this strategy is supported by finding that cells undergo epithelial-mesenchymal transition during the peritoneal metastatic process. Recent studies have shown that this cell population acquires an increased ability to migrate and invade, and becomes exquisitely resistant to systemic chemotherapy [22,23]. Additionally, following implantation at distant sites, these metastatic deposits frequently have already created a microenvironment favourable for further growth and spread. They have been shown to attract tumour-promoting cell populations such as M2-tumour associated macrophages, T-regulatory cells, stellate cells and cancer-associated fibroblasts [24]. These factors are thought to mediate additional resistance to chemotherapy which may in part explain the failure of systemic intravenous chemotherapy to treat peritoneal carcinomatosis [25].

Furthermore, early administration of locoregional cytotoxic therapy with known pharmacokinetic advantages over IV chemotherapy may reach tumour cells prior to metastatic implantation. Reaching tumours prior to the development of cellular resistance-mediating pathways is known to yield improved efficacy [26–28]. Clinically, this strategy has been demonstrated to yield improved survival with the addition of IP chemotherapy in optimally debulked stage-III ovarian cancers [29]. Additionally, there is evidence from multiple randomised phase-II studies in gastric cancer supporting IP chemotherapy at the time of surgery to prevent peritoneal dissemination [20]. Currently, it is not known whether adjuvant IP chemotherapy at the time of definitive surgery might play a useful role in colorectal cancer [10,20,21,30,31]. A previous review on the topic identified seven comparative studies, while attesting to the heterogeneity of evidence, suggested an improved outcome with IP treatment. On the other hand, the largest study on the subject, which comprised 753 patients, was excluded from this review. Further, this review did not include a formal analysis quantifying the impact of IP chemotherapy.

To better assess the existing data supporting this approach, we systematically searched the literature and quantified the impact of adjuvant IP chemotherapy on clinical outcome with random effects models. We additionally sought to assess and quantify any potential morbidity associated with adjuvant IP chemotherapy.

Materials and methods

Search strategy

A systematic search of published literature was conducted using PubMed and Cochrane databases in August 2015. Combinations of the following terms were used to identify relevant studies: peritoneal or IP chemotherapy, carcinomatosis, colorectal cancer, cholangiocarcinoma, bile duct cancer, pancreas cancer, hepatocellular cancer and gastric cancer. Studies originally not published in English or not categorised as clinical trials were excluded. Additional review articles were screened for clinical studies not identified by the original search.

Selection criteria

Titles and abstracts were screened for eligibility by two independent researchers (PLF and MMK). All studies examining adjuvant IP chemotherapy for GI malignancies were included, studies in patients with established carcinomatosis were excluded and studies focussed on other malignancies including appendiceal and ovarian tumours were excluded as well. Other exclusion criteria included duplicates, studies published as conference abstracts only, phase I or preclinical studies, studies that did not report patient outcome data, studies that included more than 50% of patients with established peritoneal carcinomatosis, studies in which ≥50% of subjects received another form of regional therapy like intra-portal chemotherapy infusions, or studies that did not separate outcomes for those without macroscopic peritoneal carcinomatosis. Studies that used a non-chemotherapeutic IP therapy such as an immune or radiation therapy, either alone or in combination with IP chemotherapy, were also excluded. Studies that gave neoadjuvant systemic chemotherapy were excluded as well. In the case of disagreement about inclusion, the full text was examined to render a final decision.

Data analysis

Trials were sorted by histology and design. For each article, the following data were collected: author, year of publication, total number of patients including number of patients enrolled into IP chemotherapy vs. standard treatment arm in case of a randomised study, stage of disease, details on type of IP chemotherapy (including dose, administered volume and heat and timing), outcome data, toxicity, median follow-up, disease-free survival, peritoneal recurrence-free survival and overall survival. Reported toxicities were summarised and subjectively described.

Statistical analysis

Random effect models were used to correlate outcomes reported in treatment arms. First, as part of a comparative analysis, a “bubble chart” visualisation was generated. Each trial arm for both randomised and case control studies was plotted according to sample size and main reported clinical outcome. Studies that were randomised or well-matched case control designs were selected and compared using pooled odds ratio in random-effect models. Statistical heterogeneity was quantified by the I² inconsistency test and the tau² test, all summary statistics and analysis were performed within the R statistical and programming environment [32]. Graphical displays were generated through the “ggplot2” software package [33].

Results

Study population

The database search resulted in a total of 444 studies, of which 386 studies were excluded based on title and abstract. The most common reason for exclusion was that the study examined IP chemotherapy as a treatment for established peritoneal spread. A flow chart describing the literature search according to the preferred reporting items for systematic reviews and meta-analysis (PRISMA) statement is shown in Figure 1 [34]. Full texts of the remaining 57 studies were screened for eligibility and 30 studies were excluded for reasons outlined in Figure 1. There were nine studies in colorectal cancer identified for detailed analysis. The remainder consisted of 17 studies in gastric cancer and one in pancreatic cancer. There were no studies identified in biliary tract malignancies.

Figure 1. Article identification and screening according to PRISMA reporting guidelines [34].

Among the nine colorectal cancer trials, 853 patients were treated with 595 receiving adjuvant IP therapy and surgery and 359 undergoing surgical resection alone. A total of 681 patients were enrolled in phase II randomised studies with 393 patients receiving IP chemotherapy and 288 undergoing surgery alone. Prospective single arm and comparative case control studies grouped together included 494 patients in intervention arms and 359 patients in standard surgery arms. The earliest article on adjuvant IP therapy in colorectal cancer was published in 1985 and the most recent in 2013 (Table 1). All RCTs had pre-specified inclusion criteria.

Table 1. Characteristics and outcomes of studies examining surgery plus adjuvant IP chemotherapy versus surgical resection alone in colorectal cancer.

		Stage
Author (study type)	Total/(H)IPEC/other	I–III	IV	IP method	IP agent and dose	Systemic Tx	Median follow-up	Clinical outcome	p Value
Tentes et al. [35] (RCT)^a	107				HIPEC: MMC 15 mg/m^2 + oxaliplatin 130 mg/m^2	III and IV patients (pts) received IV 5FU and LV		3 years OS HIPEC 100% vs. 3 years OS EPIC 69%	0.011
	40	38	2	HIPEC	EPIC: 5FU 600 mg/m^2, POD 0–5		17 months
	67	67	0	EPIC			28 months
Nordlinger et al. [36] (RCT)^b	753	753			EPIC: 600 mg/m^2 FU, q 24 h POD 4–9	Randomized to adjuvant FU/LV or FU/folinic acid	6.8 years	5-year OS: 72.3% (regional) vs. 72.0% control	OS: p = 0.69
	235	235		EPIC	Intraportal: 500 mg/m^2 FU, q 24 h POD 1–7			5-year DFS: 0.94 HR for regional vs. control
	415^a	415		Regional: (EPIC or intraportal)
Vaillant et al. [37] (RCT)	267	267	0		5FU 0.6 g/m^2/day ×3 h for 6 d, POD 4–14	Intra-op IV 5FU EPIC group only	58 months	DFS: 5 year – 74% ± 4% (HIPEC) vs. 69% ± 4%	DFS – p = 0.30
	133	133	0	EPIC				OS: 5 year – 68% ± 4% vs. 62% ± 4%	OS – p = 0.26
Scheithauer et al. [38] (RCT)	241	241	0		LV 200 mg/m^2 + 5FU 350 mg/m^2 on day 1 and day 3 × q 4 weeks ×6	IV 5FU/LV both groups	4 years	DFS: 48 months – 77.3% (HIPEC) vs. 57.5%	DFS – p = .0015
								OS: 4 years – 83% (HIPEC) vs. 65%	OS – p=.0004
Sugarbaker et al. [39] (RCT)	66	ND^c	ND^c		1040 mg increased by 130 mg each cycle until tox, 5 d course 2 months post-op	IV 5FU 12 mg/kg non-HIPEC only	ND	MS 47.5 months HIPEC vs. 46.3 months IV group	p = 0.54
	36			EPIC
Sammartino et al. [40] (CC)	75	75	0		Ox 460 mg/m^2	HIPEC: intraportal IV 5FU/LV; post op for pT4, LN + and G3	37.5 months	OS: NS difference	NS
	25	25	0	HIPEC				DFS: 36.8 months HIPEC vs. 21.9 months control	<0.01
Noura et al. [41] (CC)^d	52	36	16		MMC 20 mg	42 pts adjuvant chemo	83.1 months	5-year cancer specific OS: 54.3% NIPEC v. 9.5% Sx alone	0.0001
	31	24	7	NIPEC
Virzi et al. [42] (PSA)	12	9	3	HIPEC	CDDP 25 mg/m^2/l + MMC 3.3 mg/m^2/l	None	49 months	5-year PFS and Peritoneal PFS 74.1 and 90.9%, 5-year OS 83.3%	NA
Palermo et al. [43] (PSA)	45	45	0	EPIC	5FU 20 mg/kg qD ×5d at week 1, 5, 9, 13, 17, 21	EBRT: Total dose 45.0 Gy	130 months	DFS: 5 and 10 years – 51% (±8%) and 49% (±8%)	NA

RCT: randomised control trial; CC: case control; PSA: prospective single arm; HIPEC: heated intraperitoneal chemotherapy; NIPEC: normothermic intraperitoneal chemotherapy; EPIC: early post-operative intraperitoneal chemotherapy; Sx: surgery; Ox: oxaliplatin; MMC: mitomycin C; 5FU: 5-fluorouracil; CDDP: cisplatin; OS: overall survival; PFS: progression free survival; ND: not declared; NS: not significant; CC: case control; PSA: prospective single arm.

^a EPIC vs. HIPEC, no surgery alone control.

^b IP treatment group included intraportal.

^c Lymph node positivity was a requirement.

^d All pts with positive peritoneal cytology, only cancer-specific survival reported.

The studies in Table 1 show heterogeneity in terms of inclusion criteria (tumor node metastasis (TNM) stage), number of individuals enrolled, IP administration modalities and IP chemotherapy selection. Stage-IV patients comprised the minority of enrolled patients, and only Noura et al. reported more than 10% of patients with stage-IV disease (31%). Distant disease cannot be completely excluded with preoperative workup alone, and a small number of stage-IV patients are to be expected in patients undergoing operations with curative intent. The smallest trial was that of Virzi et al. with only 12 patients in a prospective single arm study, while the largest was that of Nordlinger et al. with 753 patients in total enrolled. Administered IP chemotherapy regimens included 5-FU, platinum and mitomycin-C-based regimens. Among the 5 RCTs identified, all utilised some form of early post-operative IV chemotherapy, with only Tentes et al., also having an intra-operative IV chemotherapy group. Taken together, these differences in trial characteristics significantly limit the impact of any conclusions from pooled statistical analysis.

Overall mortality in patients treated with surgery vs. surgery plus IP chemotherapy

Overall survival was compared for patients undergoing surgery and IP chemotherapy vs. surgery alone (Figure 2). Displaying all trial arms in a bubble chart, there was a subjective trend toward improved survival among arms administering IP chemotherapy (red circles positioned higher on the y-axis) vs. surgery only (blue circles). Attesting to the known heterogeneity in outcome reporting, overall survival rates ranged from 50% to 75% at reported intervals between 36 and 60 months [10]. We next sought to identify whether study outcomes could be separated based on the chemotherapy agent utilised. As displayed in Figure 3, there was a trend toward improved overall survival with 5-FU-based IP regimens, compared to oxaliplatin regimens, or just surgery alone. The trend is skewed by the strikingly positive results of Tentes et al., who compared intraoperative administration of IP MMC and oxaliplatin vs. post-operative IV 5FU and found a dramatic survival difference in favour of IP chemotherapy [35].

Figure 2. Colorectal cancer trial treatment arms by overall survival and reported clinical outcome. Each bubble represents one arm of a study. Size represents number of patients. First initial of first author and study year is displayed as a label above each corresponding bubble. Red represents surgery plus adjuvant IP treatment arm, blue surgery arm alone.

Figure 3. Colorectal cancer trial treatment arms by IP chemotherapy agent. Each bubble represents one arm of a study. Size represents number of patients. First initial of first author and study year is displayed as a label above each corresponding bubble. Colors denote IP chemotherapy agent utilised.

Next, we selected randomised controlled trials that compared IP chemotherapy plus surgery to surgery alone and also reported overall survival, for initial random effects modelling. This group of studies included Vaillant, Scheithauer, Nordlinger and Noura et al. The choice whether to specify fixed or random effects in the meta-analysis was guided by the degree of effect heterogeneity in the included trials. Under Cochran's Q-test (13.9, p = 0.01), the null hypothesis of a homogeneous effect is rejected at the 5% level. The I² statistic of 78% (CI 42%, 92%) also reveals considerable “unexplained” heterogeneity in the studies [44]. Under such circumstances, random effects explicitly accounting for intra-trial variance should provide more conservative estimates of the pooled effects than modelled fixed effects will. This study, therefore, reports estimates from meta-analysis with random effects. When the studies were pooled in this manner, IP chemotherapy was associated with a statistically significant decreased risk of death of 0.55 (Figure 4, 95% CI = 0.3053, 0.9849; p = 0.0443).

Figure 4. Random effect comparison overall mortality odds ratios in colorectal cancer Mantel Haenszel random effects pooled analysis of comparative colorectal cancer trials reporting overall survival. Size of boxes is relative to weight of each study. 95% of confidence intervals represented by whiskers, arrow indicating interval clipping. Asterisk (*) denotes IP group included all “regional” chemotherapy, which included intraportal and intraperitoneal administration.

The effect would be stronger without the inclusion of Nordlinger et al., which, in the largest study on the subject, randomised 753 stage-II and -III colorectal cancer patients to IP regional chemotherapy followed by systemic chemotherapy, or to resection plus systemic chemotherapy alone. The IP regional chemotherapy group in the Nordlinger study included patients who received chemotherapy washing of the peritoneal cavity and others that received intraportal regional liver chemotherapy perfusion. Unfortunately, outcomes for the groups receiving intraportal therapy vs. IP therapy were not separated, and together they were reported to achieve a 5-year survival rate exceeding 70% with no significant difference from surgery alone [36]. It is possible that the inferior outcomes among patients who received liver perfusion as their adjuvant treatment might have negated the greater impact seen in the IP chemotherapy group. Of note, despite the study of Nordlinger et al., having significantly more patients than the other studies, included its weight in the random effect model was calculated as 31% of reflecting the significant differences among study size, an advantage compared to fixed-effect analysis which weights studies relatively more equally [45].

Heterogeneity in patient selection and study design

A stronger single in favour of IP chemotherapy may have been further obscured by the heterogeneity in tumour stage and risk for peritoneal recurrence across the studies analysed. The majority of studies attempted to select patients at high risk for peritoneal carcinomatosis via the TNM system (patients with T stage(s) ≥3 in Tentes et al., Vaillant et al., Scheithauer et al.) assuming a higher risk of peritoneal dissemination with increased T and lymph node stages [35,37,38]. Only the studies of Noura et al. and Sammartino et al. included additional criteria such as positive peritoneal lavage (Noura et al.) or presence of mucinous or signet ring tumour histology (Sammartino et al.) [40,41]. Interestingly, these studies showed the highest differences in measured clinical outcome with the addition of IP chemotherapy to definitive surgery. Noura reported a 5-year cancer specific OS of 54.3% vs. 9.5% for those receiving IP chemotherapy vs. surgery alone. Furthermore, Sammartino et al., which reported only PFS and was therefore excluded from the random effects model, did show an increase in disease free survival of 36.8 months for the IP chemotherapy group for 21.9 months for surgery alone.

An additional concern from this pooled analysis is the potential effect of the under-reporting of negative studies. Statistical tests of symmetry and visual inspection of a funnel plot (i.e. a scatterplot of the odds ratios against standard errors for studies in the meta-analysis) can help to detect publication bias. Inspection of Figure 5 reveals an acceptable distribution of the four analysed studies, each lying within the meta-analytic 99% of confidence interval. Four tests of symmetry each fail to reject the null hypothesis of symmetry at the 5% level: Begg’s test (p = 0.09), Egger’s test (p = 0.07), Peters’ test (p = 0.07) and Harbord’s test (p = 0.08) [46–49]. The conclusions that can be drawn from these tests are limited by the small number of trials in the analysis, but evidence of publication bias does not emerge.

Figure 5. Funnel plot with fixed and random effects shown.

IP chemotherapy and associated operative morbidity

IP chemotherapy-related complications due to the cytotoxic agent were infrequent and relatively minor. In-hospital mortality was rare, and the most commonly reported complications were those who are known to be associated with systemic chemotherapy. The most common abdominal complications reported were anastomotic failure, peritonitis and small bowel obstruction (Table 2). A random effect analysis of the four eligible trials was used to quantify the association between morbidity and IP chemotherapy.

Table 2. Peri-operative morbidity and mortality of studies examining surgery plus adjuvant IP chemotherapy vs. surgical resection alone in colorectal cancer.

Author (study type)	IP method	Overall complications	Complication notes
Tentes et al., [35] (RCT)^a	HIPEC vs. EPIC	In hospital mortality: 1 (2.5%) vs. 9 (13.4%) (p = 0.009) Morbidity: 16 (40%) vs. 22 (33%) (p > 0.05)	Top 3: Anastamotic failure (10% vs. 10%) Respiratory (7.5% vs. 7.5%) Wound infections (10% vs. 1.5%)
Nordlinger et al. (RCT)^b	Surgery vs. regional: (EPIC or intraportal)	Morbidity: 77 (21%) vs. 110 (29%)	63 of the complications in the regional group were abdominal complications 41 were WHO grade 3–4
Vaillant et al. [37] (RCT)	Surgery alone vs. EPIC	In hospital mortality: 0 (0%) vs. 2 (1.5%) Morbidity: 17 (12.7%) vs. 28 (21%)	18 total abdominal complications versus 27 extra-abdominal complications. Most frequent complications were wound sepsis (6 vs. 3). The cause of death in both patients was PE
Scheithauer et al. [38] (RCT)	Surgery alone vs. EPIC	In hospital mortality: none Morbidity: 53% vs. 56% (WHO grade 1–2) 13% vs. 3% (p = 0.01 ≥WHO grade 3)	Top 3: Leukopenia (19 vs. 18) Granulocytopenia (18 vs. 15) Diarrhea (12 vs. 7)
Sugarbaker et al. [39] (RCT)^c	Surgery + IV 5FU vs. EPIC	Morbidity 53% vs. 45%	Top 3 non-haematologic: SBO (8 vs. 3) 5FU peritonitis (5 vs. 0) Diabetes (1 vs. 2)
Sammartino et al. [40] (CC)	Surgery alone vs. HIPEC	Morbidity 9 (18%) vs. 5 (20%)	One patient in HIPEC group got post-operative pancreatitis, and one received a laparotomy on POD 1 for bleeding
Noura et al. [41] (CC)^d	NIPEC	In hospital mortality: none grade 3–4 morbidity: 0 vs. 1 (1.9%)	Only IP chemotherapy related event was a grade 3 skin ulceration at the catheter insertion site
Virzi et al. [42] (PSA)	HIPEC	In hospital mortality: none grade 3–4 morbidity: 3	Transitory acute renal failure, mucocutaneous candidiasis, urinary retention
Palermo et al. [43] (PSA)	EPIC	In hospital mortality: none grade 3–4 morbidity: 33.5%	Most common morbidity was leukopoenia and thrombocytopenia. 4 patients experienced grade 3 peritonitis, and 4 patients developed SBO requiring surgery (disease recurrence)

HIPEC: heated intraperitoneal chemotherapy; NIPEC: normothermic intraperitoneal chemotherapy; EPIC: early post-operative intraperitoneal chemotherapy.

^a EPIC vs. HIPEC, no surgery alone control.

^b IP treatment group included intraportal.

^c Lymph node positivity was a requirement.

^d All pts with positive peritoneal cytology, only cancer-specific survival reported.

IP chemotherapy was not associated with a statistically significant increase in morbidity of general post-operative complications (Figure 6(A), OR = 1.7, p = 0.17). Reported complications were broadly grouped into two categories, chemotherapy-related issues such as agranulocytosis, febrile neutropenia and other known haematologic and general issues, or IP chemotherapy specific abdominal complications such as anastomotic leak, diarrhoea, abdominal pain, obstruction/ileus or ostomy/wound breakdown (Table 2). When these IP chemotherapy-specific complications were entered into random effect models, there was no associated increase in morbidity associated with IP chemotherapy (Figure 6(B), OR = 1.4, p = 0.5). While not statistically significant, the trend toward increased complications with IP chemotherapy was in large part driven by Scheithauer et al., which performed the largest series examining delayed post-operative IP chemotherapy (as late as 35 d after definitive surgery) in using peritoneal dialysis catheters vs. surgery alone, and reported a significantly higher severe treatment-associated side effect rate (13% vs. 3%; p = 0.01) in the IP arm [38]. Among studies included in the random effects model, only four mortalities were reported, two in the IP arm of Vaillant et al., and one in the regional and one in the systemic arm of Nordlinger et al., resulting in no significant associations.

Figure 6. Random effect comparison of post-operative morbidity associated with intraperitoneal chemotherapy in colorectal cancer Mantel Haenszel random effects pooled analysis of comparative colorectal cancer trials reporting post-operative morbidity. Size of boxes is relative to weight of each study. 95% of confidence intervals represented by whiskers, arrow indicating interval clipping. (A) All reported complications included. (B) Abdominal only complications, including, GI obstruction, ileus, anastomotic leak, diarrhoea, abdominal pain, ostomy and wound complications. Asterisk (*) denotes IP group included all “regional” chemotherapy, which included intraportal and intraperitoneal administration. (**) denotes Scheithauer et al. reported grade III and above complications only.

Several studies were excluded from the pooled analysis, but still provided valuable insights. Sammartino et al. enrolled patients at high risk for peritoneal involvement, including those with T3/T4 tumours, signet cell or mucinous histology and reported improved median disease-free survival with no difference on OS [40]. This study, in addition to the small initial report of Sugarbaker et al., which did not find a difference in favour of IP chemotherapy, was both excluded from the random effects analysis as they did not explicitly report survival odds ratios. Of note, the study of Tentes et al. was also excluded from the random effects model as it did not include a surgery alone control group. As mentioned previously, this study is notable for its dramatic reported survival advantage for the immediate IP chemotherapy group [50].

Discussion

Adjuvant systemic chemotherapy for unselected stages II and III colorectal cancer patients is known to improve outcome, but only in a minority of patients. Recent efforts, therefore, have shifted focus on improving patient selection for adjuvant therapy. Studies have demonstrated promise in areas such as using molecular markers like high microsatellite instability, BRAF mutant tumours or new genetic predictors (Oncotype DX) [51,52]. Clinical predictors such as lymphovascular invasion, T4 tumours with signs of perforation or presence of intestinal obstruction in stage-II cancers are currently under investigation.

The use of adjuvant IP chemotherapy, in this regard, remains a relatively unexploited area, despite its potential to focus prevention at the site of likely recurrence, afford a more personalised use of treatment and reduce toxicity. This concept, to a certain degree, is an extension of hepatic artery chemotherapy infusion pump results reported by Kemeny et al. [53]. Using a similar site-directed approach, the rationale of adjuvant hepatic artery perfusion is to target the organ at highest risk of recurrence with 5-FUDR, seeking to increase the dose delivered, target the expected pattern of recurrence and reduce toxicity compared to systemic treatment. For similar reasons, we focussed on the peritoneal cavity, the second-most likely site of distant recurrence in colorectal cancer.

Our main finding is an association between adjuvant IP chemotherapy and improved overall survival following definitive surgery for early stage colorectal cancer. This finding shares similarities with previous analyses on adjuvant IP chemotherapy in gastric cancer, where several reviews have identified improved outcomes in patients undergoing surgery plus IP chemotherapy compared to surgery alone [20,21,54,55]. Specifically, our group has previously identified associations between survival and IP chemotherapy that was delivered intra-operatively vs. post-operatively, or between the use of mitomycin C and cisplatin [20]. These reported conclusions, however, are limited by the paucity of randomised studies, as well as heterogeneity in eligibility criteria, IP chemotherapy agents used and outcome reporting. With the more limited dataset of adjuvant IP chemotherapy in colorectal cancer, similar comparisons were not possible. Furthermore, previous random effects analysis on the use of adjuvant IP chemotherapy in gastric cancer derived its signals from trials which delivered immediate, intra-operative IP chemotherapy, while all trials included in presented random effects model utilised early post-operative IP chemotherapy. The later postoperative rather than immediate intraoperative administration of IP chemotherapy loses the advantage of hyperthermia, which has been shown to have an anti-neoplastic advantage in several clinical and preclinical studies [20,56,57].

Our study seeks to quantify the impact of IP chemotherapy delivered at the time of definitive surgery in patients with colorectal cancer. Despite the sentinel study of Verwaal et al. showing that CRS and IP chemotherapy has merit in the management of established peritoneal carcinomatosis due to colorectal cancer, our report, as well as others, shows that there is a lack of high-quality randomised trials focussed on this clinical question. This support for IP chemotherapy in metastatic disease is distinctly different from the case of gastric cancer, where adjuvant IP chemotherapy is supported by a number of randomised studies, but evidence for IP chemotherapy in within the setting of peritoneal carcinomatosis is limited [58,59]. Therefore, future development of IP chemotherapy in colorectal cancer might benefit from early identification of patients most likely to benefit. It is noteworthy that the two most recent studies identified here used histopathological and cytological criteria in addition to clinical criteria to identify patients at high risk for peritoneal recurrence. This approach yielded improved clinical outcomes [40,41]. Mucinous histology and the presence of signet ring cells, used by Sammartino et al., are known to be associated with increased peritoneal spread [40]. The same is true of positive peritoneal cytology, which was used by Noura et al. [41]. In line with these findings, the recently begun multi-institutional Adjuvant HIPEC in High Risk Colon Cancer (COLOPEC) trial, designed to randomise to either surgery plus IP chemotherapy or systemic chemotherapy, is limited to T4 tumour stage or intra-abdominally perforated colon cancer [31]. The selection criteria utilised by Sammartino et al. and Noura et al. were in contrast to that of Nordlinger et al. [31,40,41]. This phase-III multi-institutional trial failed to disclose which patients in the regional chemotherapy arm received intraportal therapy vs. IP therapy. Additionally, Nordlinger et al. enrolled a large number of stage-II or -III patients with N0 disease. Both of these factors might have contributed to the study’s negative result [36].

Findings from this study will lend support for choosing well-defined criteria to select patients at increased risk for peritoneal involvement following surgical resection. Such selection criteria will need to extend beyond the traditional clinicopathological criteria of TNM system, grade and histopathology (signet ring cell, mucinous histpathology), and should involve a direct measure like positive IP cytology or peritoneal micrometastasis detected intraoperatively via use of near-infrared imaging and indocyanine green (ClinicalTrials.gov identifiers NCT02032485 and NCT01982227, respectively). A study employing positive cytology as a selection criterion for IP chemotherapy administration at time of gastrectomy has been recently opened at the National Cancer Institute for Gastric Cancer (ClinicalTrials.gov NCT03092518).

Overall, the prognostic value of mutational derangements across tumours is likely to surpass stage and histology as the basis for surgical selection of patients. Indeed, genetic predictors specific for peritoneal surface involvement have already been shown to have value for differentiating between low- and high-risk appendiceal and colorectal peritoneal carcinomatosis [60]. Although metastatic deposits contain significant mutational heterogeneity, recent work has identified certain genomic signatures in primary tumours that are predictive of future metastatic sites [61,62]. Thus, correlative tissue profiling could help to identify patients at high risk for peritoneal spread and likely to benefit from IP chemotherapy.

Given recent changes in surgical care, the adjuvant IP chemotherapy approach will need to be adjusted to fit better overall trends in cancer care. For colon cancer in particular, recent evidence points to the non-inferiority of minimally invasive resections that also reduce length of stays. Taking this into account, a one-time administration of IP chemotherapy at the time of resection is likely to be more easily accepted compared to multi-day post-surgical EPIC treatment(s). This approach is also supported by the direct comparison of the two modalities in Tentes et al., reporting a clinical outcome advantage in patients who received adjuvant IP chemotherapy immediately intra-op vs. delayed post-operatively. Furthermore, the recently opened multi-institutional COLOPEC study in Europe also uses a one-time IP chemo infusion during surgery [31].

While the quantitative analysis presented here supports the need for additional study of adjuvant IP chemotherapy in high-risk colorectal cancer, our study has significant limitations. These include the small number of randomised trials, which was further narrowed in the random effects analysis due to heterogeneous outcome reporting. Additionally, significant heterogeneity remained in the clinicopathological characteristics of patients included in the model, in terms of risk for recurrence, administration of additional adjuvant treatment, as well as risk for peritoneal involvement.

Conclusion

This systematic review provides rationale in favour of adjuvant IP chemotherapy in resectable colorectal cancer patients at risk for peritoneal spread. Summary analyses do not support the traditional assumption that IP chemotherapy is associated with adverse post-operative outcomes such as intra-abdominal abscess formation and anastomotic failure. Future clinical trials should seek to enrich patients at high risk for peritoneal carcinomatosis as well as formalise standard inclusion criteria and IP chemotherapy treatment agent use to better define the role of this therapy in colorectal cancer.

Disclosure statement

No potential conflict of interest was reported by the authors.