Colorectal cancer is one of the most common malignancies. In England and Wales it is the second most common in terms both of incidence and mortality [1] with approximately 36,000 new cases diagnosed in 2002 and 17,000 people dying from colorectal cancer in the same year [2]. In the USA it is the third most common cancer with an estimated 149,000 new cases in 2006 and approximately 55,000 deaths [3].

About 80% of all patients diagnosed with colorectal cancer (including some with advanced disease) undergo surgery [4]. Open resection is the standard method for surgical removal of primary colorectal tumours in the UK [5]; it results in significant morbidity. Over the past 15 years, laparoscopic resection has been considered as an alternative to open surgery although there are concerns about both its safety and effectiveness compared with open resections. There are three types of laparoscopic surgery: totally laparoscopic, laparoscopic-assisted and hand-assisted laparoscopic surgery (HALS).

In response to these concerns, the National Institute for Health and Clinical Excellence (NICE) issued guidance in 2000 on the use of laparoscopic surgery for colorectal cancer. This guidance stated that open rather that laparoscopic surgery was the preferred procedure and that laparoscopic surgery should only be undertaken as part of a randomised controlled trial (RCT) [5]. New data have since become available, particularly from three large RCTs [6–8] (each with around 800 participants) and an individual patient data (IPD) meta-analysis of these three trials [9] plus a further moderately sized trial [10]. The aim of this systematic review was to assess the clinical effectiveness of laparoscopic, laparoscopically assisted (hereafter together described as laparoscopic surgery) and HALS in comparison with open resection in the context of a reassessment by NICE.

Methods

Searching for the evidence

Published and unpublished reports of RCTs and systematic reviews evaluating the effectiveness of laparoscopic and HAL surgery for colorectal cancer were identified by the electronic searches. Searches were restricted to the years 2000 onwards (as earlier trials had been identified by the previous systematic review) [11] without language restriction and included abstracts from recent conference proceedings. Full details of the search strategy are reported elsewhere [12]. Additional data and relevant studies were identified from the reference lists of included studies and systematic reviews as well as by contacting lead authors of all included RCTs.

Inclusion and exclusion criteria

Individual RCTs and individual patient data meta-analyses of RCTs of laparoscopic surgery compared to open surgery for colorectal cancer were included. Studies including patients undergoing palliative treatment were excluded. The prespecified subgroups considered were defined by: location of cancer; stage of cancer; and mean age at diagnosis. The prespecified outcomes are listed in Table 1.

Table 1 Summary of outcomes reported in the included studies
Full size table
Table 1 (continued) Summary of outcomes reported in the included studies
Full size table

Quality assessment strategy

Two reviewers, working independently, assessed the methodological quality of included studies. Disagreements were resolved by consensus or arbitration. Primary RCTs were assessed using the Delphi criteria list [13] and the meta-analyses were assessed using the Oxman and colleagues checklist [14, 15].

Data extraction strategy

The titles and abstracts of all papers identified by the search strategy were screened. Two reviewers independently assessed full-text copies of all potentially relevant studies and extracted data from the included studies. Reviewers were not blinded to the names of the studies’ authors, institutions or sources of the reports. Any differences that could not be resolved through discussion were referred to an arbiter.

Data synthesis

For trials with multiple publications, only the most up-to-date data for each outcome were included. Dichotomous outcome data were combined using the Mantel–Haenszel relative risk (RR) method and continuous outcomes were combined using the inverse variance weighted mean difference (WMD) method. Ninety five percent confidence intervals (CI) and p values were calculated for the estimates of RR and WMD. The results are all reported using a fixed effects model. Chi-squared tests and I-squared statistics were used to explore statistical heterogeneity across studies and, when present, random effects methods were applied. Other possible reasons for heterogeneity were explored using sensitivity analyses. The meta-analyses were conducted using the Cochrane software RevMan 4.2.

Due to the lack of uniformity of the data presented by many studies, a qualitative review looking for consistency between studies was also performed. This was supplemented, where appropriate by considering the consistency in the direction of effects using the sign test [16].

Results

Forty-four reports describing 20 studies (19 RCTs and one individual patient data (IPD) meta-analysis [9] met the inclusion criteria for the review (Fig. 1).

Fig. 1
figure 1

Study selection process

Full size image

Quality and characteristics of available evidence

All RCTs were generally of a similar good quality (Table 2). The IPD meta-analysis [9] was not fully comprehensive in terms of the search methods employed and failed to report the selection criteria for including studies. No details were given about how the quality of included studies was assessed. However, the findings of the included studies were combined appropriately relative to the primary question the review addressed and the conclusions were supported by the data and the analysis reported.

Table 2 Summary of the quality assessment of the included RCTs
Full size table

In the 19 eligible RCTs, there were 19 relevant comparisons, none of which involved a comparison with HALS. Studies included varied in relation to settings, age and gender of participants, types of outcomes measured, and site and stage of cancer (Table 3). In general, studies reported the participants’ stage of cancer using either Dukes’ or tumour–node–metastases (TNM) classification. One study failed to report the stage of cancer at which participants were enrolled [17] and in another the stage was not clearly reported [7]. Where specified, the majority of participants receiving either laparoscopic or conventional open interventions had either Dukes’ B (TNM stage II) or Dukes’ C (TNM stage III) cancer.

Table 3 Summary of the baseline characteristics
Full size table

The IPD meta-analysis [9] included patients from four of the included trials: conventional versus laparoscopic-assisted surgery in colorectal cancer (CLASICC) [7], colon cancer laparoscopic or open resection (COLOR) [8], clinical outcomes of surgical therapy (COST) [6] and Lacy and colleagues [10]. A total of 1765 patients who were randomised before 1 April 2000 and had three years follow-up were included in this IPD meta-analysis.

Description of surgery received

Opposite method initiated

Opposite method initiated was defined as a laparoscopic operation initiated when an open resection was allocated, or vice versa. The opposite method to the one that the patient was randomised to was initiated in 46/1173 (3.9%) of those randomised to laparoscopic resections [7, 8, 10] and 4/268 (1.5%) of patients randomized to open surgery. Rates varied between the trials. In the IPD meta-analysis [9], the rates were similar in both groups (<1%).

Number of ports

The number of port sites used for laparoscopic resection varied between three and five across the studies reporting this outcome [18–24].

Conversion

A conversion was defined as a procedure initiated as laparoscopic but converted to an open procedure. Overall, 421 of 2027 (21%; range 0–46%) laparoscopic procedures were converted to open surgery [6–8, 18–20, 23–28]. A similar result was reported in the IPD meta-analysis [9].

Surgeons’ prior experience

Ten RCTs reported that surgeons performing the procedures were experienced in laparoscopic colorectal surgery [6–8, 10, 19, 21, 23, 25, 26, 29]. However, only three [6–8] reported a minimum level of experience required, which in each was that surgeons had undertaken at least 20 laparoscopic colorectal operations before participating in the trial.

Assessment of effectiveness

Duration of operation

Sixteen studies (n = 4125) provided information on the duration of operation (Table 4). In all but one study [18] the duration of operation was longer in the laparoscopic group (sign test, p < 0.001) and this difference was statistically significant (p < 0.05) in 12 studies. Only three studies [10, 22, 26] presented data in a form sufficiently similar to allow meta-analysis, which showed that laparoscopic surgery took 40 min longer than open surgery (95% CI 32–48, p < 0.001). This finding is consistent with the data not amenable to meta-analysis (Table 4). There was evidence of statistical heterogeneity, but the direction of effect was consistent across the studies. Using a random effects model did not change this pattern.

Table 4 Continuous outcomes for laparoscopic versus open resection
Full size table

Blood loss

Nine studies [8, 10, 19, 20, 23, 25, 26, 28, 30] provided information on blood loss but the data were not reported in a form sufficiently similar to allow for a quantitative synthesis (Table 4). Eight studies reported less blood loss following laparoscopic surgery [8, 10, 19, 20, 23, 26, 28, 30], and this was statistically significant in six [8, 10, 19, 20, 28, 30] (sign test, p = 0.039).

Lymph node retrieval

Seven [7, 18, 20, 23, 25–27] of the 12 studies providing data (Table 4) reported more lymph nodes retrieved in the open compared with the laparoscopic group, two [19, 29] reported more in the laparoscopic group and three studies reported no differences [6, 8, 10] (sign test, p = 0.289). Meta-analysis of the three trials [10, 26, 29] reporting data suitable for synthesis showed no statistically significant difference between groups (WMD −0.41; 95%CI −1.42 to 0.59, p = 0.42). The mean number of lymph nodes retrieved reported in the IPD meta-analysis [9] was 11.8 and 12.2 in the laparoscopic and open groups, respectively.

Length of hospital stay

All 14 studies [6–8, 10, 18–23, 25, 26, 28, 30] that provided information on length of hospital stay reported lower mean or median stay in the laparoscopic group, which was statistically significant in 11 studies [6, 8, 10, 19, 20, 22, 23, 25, 26, 28, 30] (Table 4) (sign test p < 0.001). Four studies reported data suitable for synthesis [8, 10, 22, 30] and the average length of stay was significantly shorter following laparoscopic surgery (WMD −2.58 days, 95% CI −3.12 to −2.03, p < 0.001). This result was consistent with the data from those trials that reported data not amenable to meta-analysis (Table 4). There was marked heterogeneity observed in this meta-analysis, but there was consistency in the direction of effect. Using a random effects model did not change this pattern. The main source of heterogeneity appears to be from the study by Zhou and colleagues [30], where the average age of participants was lower than in the other studies reviewed. Additionally, all participants in the Zhou study had rectal cancer.

Adverse events

Eight [7, 8, 10, 20, 24, 26, 28, 30], three [8, 18, 28], seven [8, 10, 19, 22, 24–26] and nine studies [7, 8, 10, 19, 20, 24, 26, 28, 31] reported data on anastomotic leakages, abdominal wound breakdown, wound infection and urinary tract infections, respectively. There was no statistical significant differences between the two groups, but clinically important differences could not be ruled out as the size and direction of effect varied across studies and the confidence intervals were wide (Fig. 2).

Fig. 2
figure 2

Adverse events for laparoscopic versus open surgery

Full size image

Seven RCTs [6–8, 10, 19, 26, 28] provided information on operative and 30-day mortality. In terms of operative mortality, the difference was not statistically significant and the confidence interval was wide (Fig. 2). Thirty-day mortality was less in the laparoscopic group than in the open group but again this was not statistically significant and no difference was detected (Fig. 2; RR 0.92, 95% CI 0.74 to 1.14).

Seven RCTs [6, 10, 18, 19, 23, 25, 26] provided information on recurrence (n = 1528). Recurrences appeared less frequently in the laparoscopic group than in the open resection group (Fig. 2), but the difference was not statistically significant. The results of this meta-analysis should be treated with caution as the follow-ups of the RCTs ranged from three to 108 months. The recurrence rate reported in the IPD meta-analysis was 14% in the laparoscopic group and 16% in the open group at three years (p = 0.43) [9]. There were only three reported cases of wound recurrences across the four RCTs [6, 21, 22, 28] that reported this outcome (laparoscopic = 2; open = 1) [6]. Eight studies [10, 20, 23, 25–27, 30, 32] provided information on port-site recurrence (3/483, 0.6%).

Only two studies reported incidence of incisional and port site hernia [26, 31]. The average follow-up in one was 2.5 years [31] and in the other 4.2 years [26]. Hernias were reported in 17 (one of which was a port-site hernia) out of 249 (7%) participants in the laparoscopic group and 13 out of 243 (5%) in the open group, but this difference was not statistically significant (Fig. 2).

Postoperative pain

Five studies included a measure of postoperative pain [7, 23, 26, 33, 34]. Between the first day and two weeks post-operation, four studies favoured the laparoscopic group [7, 23, 26, 33] and one did not show any difference [34] (sign test, p = 0.125). Three studies measured pain at one to three months postoperatively but this did not differ significantly between the two interventions [7, 23, 34]. Four studies reported that patients in the laparoscopic group required fewer days of postoperative analgesia than in the open group [6, 20, 25, 30] (sign test, p = 0.031). Other data on analgesic use was consistent with this [21, 28].

Time to return to usual activities

Only one study reported data on time to return to usual activities [26]. The average time to resume household activities in the laparoscopic group (mean 32 days, range 4 to 365 days) was lower than that in the open group for patients with rectosigmoid cancer (mean 44 days, range 7 to 198 days, p = 0.002).

Health-related quality of life (QoL)

Four studies, using a variety of instruments, reported the QoL of people undergoing laparoscopic or open resections [7, 28, 34, 35]. Three studies reported higher QoL following laparoscopic surgery [7, 34, 35] and one reported similar scores [28], but this was a randomised study embedded within an enhanced recovery program.

Overall survival

Six RCTs [6, 10, 19, 25, 26, 30] provided information on overall survival. Length of follow-up of the RCTs ranged from one to 108 months. In the time-to-event IPD meta-analysis [9] of four trials, no evidence of a statistically significant difference in overall survival was found (hazard ratio 1.07; 95% CI 0.83 to 1.37, p = 0.61). As the IPD meta-analysis did not include all relevant studies, the data from all six RCTs reporting survival data were included in a meta-analysis (Fig. 3; RR 1.03, 95% CI 0.98 to 1.09). The results of this meta-analysis should be treated with caution as the length of follow-up of the RCTs varied and the analysis only considered the proportion of deaths and not time to death.

Fig. 3
figure 3

Overall survival and disease-free survival for laparoscopic versus open surgery

Full size image

Disease-free survival

Four RCTs [6, 10, 25, 26] provided information on disease-free survival (Fig. 3: RR 1.01 95% CI 0.95 to 1.07, p = 0.83). This result is consistent with the IPD meta-analysis [9] where disease-free survival up to three years was found to be greater (by 0.5%) in the laparoscopic group although this was not statistically significant (hazard ratio 0.99; 95% CI 0.80 to 1.22; p = 0.92) [9].

Important subgroup differences for laparoscopic versus open techniques

Patients undergoing conversions

Three studies reported separate outcome data for patients undergoing conversions [7, 19, 25]. The pattern observed in converted patients, for duration of operation, urinary tract and wound infection, and overall survival was similar to that reported above. Converted patients however, displayed higher blood loss and longer length of hospital stay. In addition, tumour recurrence appeared to be greater than that observed for patients who were successfully managed according to their treatment allocation although lymph node retrieval was higher. Converted patients showed poorer QoL at baseline and at every follow-up assessment than patients who underwent laparoscopic resection [34].

Effect of surgeon experience

Three trials reported the effect of surgeon experience on outcomes [6–8]. The COST trial found no experience-based trends for conversion, length of stay or QoL measures [6, 34]. However, the CLASICC trial reported a decline in number of conversions by year of recruitment from 38% in the first year to 16% in the sixth year [7]. The COLOR trial also found that the duration of surgery for laparoscopic procedures became shorter with increasing numbers of patients per centre (p = 0.03), although the number of lymph nodes harvested and length of hospital stay did not differ significantly [8].

Location of cancer

Subgroup analysis showed no evidence that the treatment effect size for anastomotic leakages was different for colon compared with rectal cancer (Fig. 4). However, the evidence is limited as only two RCTs reported anastomotic leakages in rectal patients [7, 30] and hence confidence intervals are wide. A similar result was observed for wound infections and urinary tract infections (Fig. 4).

Fig. 4
figure 4

Subgroup analyses by location of cancer for laparoscopic versus open surgery

Full size image

Stage of cancer

Two RCTs provided subgroup analysis by stage of cancer for overall survival [6, 26]. In both of these trials there was no significant difference in overall survival of patients undergoing laparoscopic resection compared to open resection for cancer stages I, II or III (p > 0.05). The IPD meta-analysis compared overall and disease-free survival for patients undergoing laparoscopic with open resection by stage of cancer [9]. These analyses were based upon data from 426 (stage I), 612 (stage II) and 480 (stage III) patients, although data were not available from all of these for the whole three year follow-up period. Using the log-rank test, the authors found no evidence of a statistically significant difference at three years in overall and disease-free survival between the randomised groups by stage of disease. They reported p values of 0.92, 0.44 and 0.53 for stages I, II and III, respectively, for disease-free survival [9].

Discussion

This paper reports an update of the review [14] that underpinned NICE’s guidance in 2000. Other reviews have been published since this guidance was issued, with the most recent being the one by Reza and colleagues [36]. We considered data from over 4,500 randomised participants across 19 RCTs of generally good quality. Our review includes nine more RCTs than included in the review by Reza and colleagues [36] plus an additional IPD meta-analysis [9] that included unpublished data. In summary, we found that convalescence is more rapid after laparoscopic surgery (reflected in less postoperative pain and blood loss, shorter hospital stay, and more rapid return to usual activities). The duration of operation for laparoscopic resection is longer. Lymph node retrieval, completeness of resection and QoL do not appear to differ between the two approaches, although clinically important differences could not be ruled out. The occurrence of complications such as anastomotic leakage, abdominal wound breakdown, incisional hernia, wound and urinary tract infections are similar, again with wide confidence intervals. Operative and 30-day mortality, were also similar in both groups.

The major development since the 2000 review [11] has been in the evidence on recurrence, disease-free survival and overall survival. We found no evidence of a difference in the number of recurrences (including wound recurrences), disease-free survival and overall survival. Furthermore, after laparoscopic resection, port-site recurrences were found in fewer than 1% of patients. This updated review also attempted to assess relative effectiveness in terms of differences in wound related morbidities such as incisional and port-site hernias, and persisting pain. Few data were identified for hernia.

Although there were marked differences in study populations and setting for duration of operation and length of hospital stay, resulting in significant heterogeneity, consistency on the direction of effect was observed.

There were relatively few data for any of the subgroups. The data that were available suggest that there may be important differences between colon and rectal cancer as well as between patients undergoing conversions. However, this is tentative, and it was impossible to judge whether or not there are potentially important differences between treatments within clinical subgroups of colorectal cancer patients. In addition, there is emerging experience in the literature in support of considering colon and rectal cancer as separate entities as rectal cancer has unique technical and pelvic dissection issues. Moreover, the systematic review was conducted on an intention-to-treat basis. Therefore, any reduction in the rate at which patients undergoing laparoscopic surgery are converted to open surgery might be expected to increase the difference observed between laparoscopic and open surgery.

Several limitations must be noted when interpreting the results of this review. An extensive literature search was conducted and both published and unpublished data were sought. Despite these efforts, it is possible that some unpublished studies may have been missed. Moreover, some trials excluded patients with advanced disease while others included only patients with colon cancer, thus limiting subgroup analyses and making results not generalisable to all groups of patients.

For many of the review outcomes the data were sparse. Nonetheless, the direction and magnitude of effect of these data appeared to be consistent.

The biggest limitation of this review is that the data available relate to at most a three-year time horizon. More long-term follow-up data are therefore required before it is certain that there is no difference in longer-term recurrence and survival.

In common with other laparoscopic procedures, laparoscopic surgery for colorectal cancer is technically more difficult than open surgery. The effect of learning may explain why some trials patients randomised to laparoscopic surgery actually received open surgery (opposite method initiated) and why so many trial patients allocated to laparoscopic surgery were converted during the procedure from laparoscopic to open surgery. Increased experience in selecting which patients are suitable for laparoscopic surgery as well as improving operator expertise might be expected to reduce both these rates.

In conclusion, with the supplement of new high-quality data that have become available and the IPD meta-analysis, this review supports the use of laparoscopic surgery for the treatment of colorectal cancer beyond an RCT setting provided that is carried out by surgeons with appropriate experience and competence. Based on this review and other considerations, NICE changed its guidance in 2006 and laparoscopic resection is now an accepted alternative to open resection in the UK. Nevertheless, there is insufficient evidence to judge whether the procedures differed in respect to long-term outcomes as the best data relates to a three-year follow-up. However, three of the largest trials [6–8] are still to be concluded which will provide more reliable data on long-term outcomes. In addition, a multicentre trial involving over 800 patients has started in Japan to evaluate whether laparoscopic surgery is the optimal treatment for colorectal cancer in which the primary outcome of interest in the study is overall survival [37]. As these data become available, they should be used to update systematic reviews. Also, the authors of the IPD meta-analysis should be encouraged to extend their data in terms of both follow-up and inclusion of other relevant studies by involving other groups. Lastly, there is very limited data available on HALS and, if this technique is to be adopted widely, methodological sound RCTs comparing HALS with both laparoscopic and open surgery are necessary.