Colorectal cancer is the second most common form of cancer in men and third most common cancer in women [1]. Surgery, sometimes combined with radio(chemo)therapy, is primary curative treatment. The three surgical techniques used are: open-, laparoscopic and robot-assisted laparoscopic surgery. Laparoscopic surgery compared to open surgery for colon cancer has been reported with benefits such as less postoperative pain [2‐4], earlier return of bowel function [2, 3], less intraoperative blood loss [3], and higher quality of life 30 days after surgery [5]. Clinical trials have also established similar long-term overall survival and recurrence [2‐4]. Regarding rectal cancer, laparoscopic surgery has been deemed as safe and effective as open surgery in the long-term [6] with similar positive short-term outcomes as for colon cancer [7]. However, two recent trials were unable to demonstrate non-inferiority for laparoscopic versus open surgery regarding short-term surrogate oncologic end-points [8, 9]. Robotic rectal cancer surgery has been associated with lower conversion rate, 1 day shorter hospital stay and longer operating time compared to laparoscopic surgery according to a large database study [10]. However, the only large randomized trial found no significant differences in conversion to open surgery, short-term surrogate oncological end-points, complication rates or health-related quality of life within 6 months [11].
Laparoscopic and robot-assisted laparoscopic surgery face higher cost for basic surgical equipment and instruments, and longer operating time than open surgery [12‐15]. However, it has been reported that higher up-front costs can be offset by shorter length of hospital stay and less time in the intensive care unit [16, 17]. Several cost studies have been conducted in the setting of randomized controlled trials early in the learning curve of laparoscopic surgery for colorectal cancer [12, 13, 15, 18]. It can be expected that with increased experience, resource utilization will decrease. In addition, costs vary by the degree of uptake of laparoscopic technique, in regions/hospitals with high uptake costs were lower, as was demonstrated in a retrospective database study of some 1400 Australian patients treated at public hospitals [17]. In Sweden, the uptake of laparoscopic surgery for colorectal cancer was 5–10% in 2007 and 20–25% in 2014, according to the Swedish Colorectal Cancer Registry SCRCR [19, 20].
Anzeige
The aim of this study was to evaluate the cost-effectiveness of laparoscopic versus open surgery in colorectal cancer using national registry data.
Methods
Study population
All patients with a diagnosis of colorectal cancer with subsequent surgery during 2013 and 2014 registered in the SCRCR were included. The only exclusion criterion was preoperatively staged T4 cancer based on the fact that there is a higher risk for selection bias to open surgery in more advanced tumors. During 2013, the SCRCR did not indicate if the index procedure was performed using robot-assisted laparoscopic technique. To be consistent, we chose to include robot-assisted operations during 2014 in the laparoscopic group. Patients were followed for 1 year after index surgery using data from the Swedish National Board of Health and Welfare retrieving data on readmissions and resource consuming events in inpatient care, such as reoperations, and data on deaths from the Cause of Death Registry and sick-leave from the Swedish Social Insurance Agency. The study protocol was approved by the local ethics committee (Dnr 661-16).
Health economic methodology
This study evaluates costs from two perspectives: society (including cost of production loss due to sick-leave) and the healthcare sector alone. Health economic evaluation methods compare the incremental cost to the incremental health benefit of a treatment [21]. The benefit is expressed as the clinical effectiveness of a treatment. In this study, a composite clinical end-point was chosen as the measure of effectiveness.
Accounting for censoring because of death or loss to follow-up provides unbiased estimates of cost [22]. This study applied the method proposed by Bang and Tsiatis [23] in which data is partitioned into smaller time intervals and then each observation of cost is weighted by the inverse probability of the observation being censored. As a result, observations of cost with high probability of being censored are weighted greater than otherwise.
Anzeige
Costs are expressed in 2016 USD, SEK was converted from using purchasing power parity from OECD [24].
Effectiveness measure
The outcome measure used was defined as a composite end-point comprised of three types of clinical events: resource consuming events in inpatient care, readmissions and death.
Events were scored with one point each to calculate the outcome variable. Resource consuming events in inpatient care included events such as surgical procedures for example stoma closure or reoperation for wound dehiscence, transfusions, as well as rectoscopy. A readmission was defined in this study as a resource consuming event in inpatient care without a specified event recorded in the National Patient Registry.
For the initial 90 days after index surgery, every clinical event was counted (one point each), regardless of underlying cause. From day 91 and up to 1 year after index surgery, only procedure-related events were counted. To define procedure related causes for readmissions and deaths, two colorectal cancer surgeons (EA and EH) independently selected diagnoses using ICD10-codes. Procedure-related causes for resource-consuming events in inpatient care were determined before data collection using NOMESCO (Nordic Medico-Statistical Committee, version 1.16) chapters. If disagreement occurred, it was resolved by including the diagnosis or procedure as related. In sensitivity analyses, all events from index surgery up to 1 year were included. The list of predefined procedure-related events can be found in online supplementary material.
Resource use and unit costs
All resource use in this study was measured and valued using gross-costing [25]. It implies that the resource use associated with surgical technique is broken down into relatively large factors, e.g. resources consumed per average episode of inpatient stay associated with a certain NOMESCO-code. Thus, it was assumed that only the type and frequency of resource consuming event and readmission, differed between surgical techniques, not the exact amount of resources consumed during each event.
Resource-consuming events in inpatient stay, including the index operation, was costed by collecting the type of event (NOMESCO-code) and frequency from the National Patient Registry. Then, the corresponding unit cost was collected from a national database (Swedish Association of Local Authorities and Regions) covering approximately 75% of Swedish inpatient care, regardless of private or public care provider. Hospitals reporting to the database collected per patient resource use in their administrative systems during inpatient stay. The minimum requirement is to collect resources consumed per individual in the operating theater including anesthesia, intensive care unit, postoperative care including time in recovery room and medical service use on the ward, X-ray, laboratory tests, expensive materials and pharmaceuticals.
The resource use per readmission was costed using the DRG (diagnosis related group) weight as registered in the national patient registry. That DRG-weight was multiplied by the cost of one DRG weight ($5723), as determined for Sweden according to 2016 data.
The classification of sick-leave related to the procedure was based on the same definition as for clinical effectiveness. The productivity cost per sick-leave day was calculated using the human capital approach [21], the average monthly wage (Statistics Sweden) plus the social security and supplementary pension fees was multiplied by the per patient sick-leave days.
Anzeige
Statistical analysis
Baseline characteristics were compared at group level using median, inter-quartile range, mean values, standard deviation and 95% confidence intervals (CI) for continuous variables while for categorical variable frequencies and percentages were reported. The former was tested using Mann–Whitney U and the latter χ2 test, respectively. All tests were two-sided and 5% significance level was applied.
To account for differences in baseline characteristics that could potentially affect the estimation of mean effectiveness and cost between laparoscopic and open surgery, multivariate regression methods were used. The included potential confounders were surgical technique, preoperative tumor-stage, ASA-classification (American Society of Anesthesiologists), tumor location, sex and age. Because it is the sample arithmetic mean that is interesting to decision-makers, generalized linear models (GLM) were utilized for the primary analysis. Both distributional family and link function were tested using modified Park’s test [26], Pregibon link test [27], modified Hosmer–Lemeshow test [28] and the Copas-test [29]. To transform mean cost back to its original scale and avoid introducing covariate imbalances, recycled predictions were used for the two models estimated using GLM [22].
As both clinical resource use and costs typically exhibit a heavily right skewed distribution, non-parametric bootstrap was used to estimate percentile-based 95% CI [30].
The robustness of the results was explored in several sensitivity analyses. The clinical effectiveness measure was tested by including all events from 3 months and for the reminder of the follow-up instead of using predefined causes. The robustness of the cost difference was tested by varying costs per resource use category and surgical technique by 30% at a time. For both the clinical effectiveness and costs, ordinary least squares (OLS) regression was used as robustness check of the GLM.
Anzeige
Statistical analysis was performed using STATA® version 14 (StataCorp, College Station, Texas, USA).
Results
After excluding preoperatively staged T4-tumors a total of 7707 patients were eligible for analysis, 6060 patients in the open surgery group and 1647 in the laparoscopic surgery group.
Key baseline characteristics are outlined in Table 1, stratified for open and laparoscopic surgery as well as colon and rectal cancer. Median age was significantly higher in the open group than in the laparoscopic colorectal cancer surgery group, 74 vs. 72 (p value < 0.001) but not between open and laparoscopic rectal cancer surgery, 69 vs. 70 (p value = 0.067). A larger share of patients in the open colon cancer surgery group was categorized as ASA-grade three or four than in the laparoscopic colon cancer surgery group, indicating patients with more co-morbidities in the open group. There were significantly more missing data regarding preoperative T and N stages for open than laparoscopic surgery (both colon and rectal cancer).
Table 1
Demographic and clinical characteristics
Colon cancer | P value | Rectal cancer | P value | |||
|---|---|---|---|---|---|---|
Open (n = 4346) | Lap (n = 1139) | Open (n = 1714) | Lap (n = 508) | |||
Age | ||||||
Median (IQR) | 74 (67–81) | 72 (65–80) | < 0.001 | 69 (61–76) | 70 (63–77) | 0.067 |
Sex, n (%) | ||||||
Male | 2090 (48) | 605 (53) | 0.003 | 1103 (64) | 266 (52) | < 0.001 |
Female | 2256 (52) | 534 (47) | 611 (36) | 242 (48) | ||
ASA, n (%) | ||||||
1 | 477 (11) | 192 (17) | < 0.001 | 304 (18) | 121 (24) | 0.005 |
2 | 2226 (52) | 629 (55) | 988 (58) | 290 (57) | ||
3 | 1408 (33) | 303 (27) | 397 (23) | 90 (18) | ||
4 | 175 (4) | 11 (1) | 17 (1) | 5 (1) | ||
5 | 2 (0) | 0 (0) | 0 (0) | 0 (0) | ||
Missing | 58 | 4 | 8 | 2 | ||
T-stage, n (%) | ||||||
1–2 | 675 (30) | 300 (39) | < 0.001 | 487 (30) | 179 (37) | 0.008 |
3 | 1553 (70) | 468 (61) | 1128 (70 | 311 (63) | ||
Missing | 2118 | 371 | 99 | 18 | ||
N-stage, n (%) | ||||||
0 | 2176 (67) | 669 (69) | 0.417 | 705 (44) | 220 (45) | 0.502 |
1–2 | 1054 (33) | 304 (31) | 914 (56) | 266 (55) | ||
Missing | 1116 | 166 | 95 | 22 | ||
M-stage, n (%) | ||||||
0 | 3833 (89) | 1079 (95) | < 0.001 | 1574 (92) | 482 (95) | 0.018 |
1 | 473 (11) | 58 (5) | 138 (8) | 25 (5) | ||
Missing | 40 | 2 | 2 | 1 | ||
Neoadjuvant chemotherapy, n (%) | ||||||
No | 4250 (98) | 1130 (99) | 0.002 | 1389 (81) | 456 (90) | < 0.001 |
Yes | 94 (2) | 9 (1) | 323 (19) | 52 (10) | ||
Missing | 2 | 0 | 2 | 0 | ||
Neoadjuvant radiotherapy, n (%) | ||||||
No | 4330 (100) | 1137 (100) | 0.725 | 645 (38) | 208 (41) | 0.180 |
Yes | 10 (0) | 2 (0) | 1068 (62) | 300 (59) | ||
Missing | 6 | 0 | 1 | 0 | ||
Adjuvant therapy, n (%) | ||||||
No | 2860 (68) | 797 (71) | 0.072 | 1123 (67) | 364 (73) | 0.010 |
Yes | 1355 (32) | 331 (29) | 554 (33) | 134 (27) | ||
Missing | 131 | 11 | 37 | 10 | ||
Hospital type, n (%) | ||||||
District hospital | 1841 (42) | 575 (50) | < 0.001 | 563 (33) | 153 (30) | 0.436 |
County hospital | 1476 (34) | 259 (23) | 587 (34) | 175 (34) | ||
University hospital | 1025 (24) | 305 (27) | 564 (33) | 180 (35) | ||
Missing | 4 | 0 | 0 | 0 | ||
BMI | ||||||
Median (IQR) | 25 (23–28) | 26 (23–28) | 0.062 | 25 (23–28) | 26 (23–28) | 0.952 |
Missing | 373 (9%) | 28 (2%) | 45 (3%) | 9 (2%) | ||
Death within 12 months, n (%) | ||||||
No | 3804 (88) | 1081 (95) | < 0.001 | 1620 (95) | 491 (97) | 0.052 |
Yes | 542 (12) | 58 (5) | 94 (5) | 17 (3) | ||
The unadjusted measure of clinical effectiveness is presented in Table 2. The difference in mean number of events was 0.33 (95% CI 0.23 to 0.43) in favor of laparoscopic surgery. Increased resource consuming events in inpatient care was the main contributor to this difference with 0.22 events (95% CI 0.13 to 0.30), followed by readmissions 0.07 (95% CI 0.03 to 0.11) and deaths 0.05 (95% CI 0.04 to 0.06). One reason for increased resource consuming events in inpatient care (n.b. data not shown) was more prevalent in the open surgery group were reoperation of wound dehiscence, 2.3% and 1.7% for open colon and rectal cancer surgery, respectively. After open rectal cancer surgery 4.8% of the patients were given blood transfusions compared to 1.7% in the laparoscopic surgery group.
Table 2
Mean unadjusted differences in clinical effectiveness and resource use open versus laparoscopic colorectal cancer surgery
Mean events open surgery | Mean events laparoscopic surgery | Difference in mean events | 95% CI | |
|---|---|---|---|---|
Primary analysis, events | ||||
Clinical effectivenessa | 1.35 (1.87) | 1.02 (1.67) | 0.33 (4.47) | 0.23–0.43 |
Resource consuming events in inpatient carea | 0.87 (1.51) | 0.66 (1.33) | 0.22 (3.60) | 0.13–0.30 |
Readmissionsa | 0.39 (0.80) | 0.33 (0.71) | 0.07 (1.90) | 0.03–0.11 |
Deathsa | 0.08 (0.28) | 0.03 (0.18) | 0.05 (0.63) | 0.04–0.06 |
Secondary analysis | ||||
Clinical effectiveness | 1.86 (2.55) | 1.40 (2.26) | 0.46 (6.07) | 0.32–0.59 |
Resource consuming events in inpatient care | 1.24 (2.08) | 0.93 (1.78) | 0.31 (4.93) | 0.20–0.42 |
Readmissions | 0.51 (0.96) | 0.42 (0.87) | 0.08 (2.29) | 0.03–0.11 |
Deaths | 0.10 (0.31) | 0.05 (0.21) | 0.06 (0.70) | 0.03–0.08 |
Sick leave | ||||
Sick leave, daysa,,b | 29.9 (81.7) | 29.0 (77.0) | 0.9 (196.9) | − 3.5–5.3 |
Sick leave, daysb | 33.1 (85.8) | 31.8 (80.4) | 1.3 (206.5) | − 3.4–5.9 |
Anzeige
As can be seen in Table 3, the adjusted mean difference between open and laparoscopic surgery was 0.23 events (95% CI 0.12 to 0.33), meaning that laparoscopic surgery patients on average had 0.23 events less than open surgery patients during 1-year follow-up.
Table 3
Mean adjusted differences in clinical effectiveness open versus laparoscopic colorectal cancer surgery
Difference mean | Standard error | 95% CI | |
|---|---|---|---|
Clinical effectivenessa, b | 0.23 | 0.05 | 0.12–0.33 |
Clinical effectivenessc | 0.22 | 0.06 | 0.11–0.34 |
Results regarding societal and healthcare costs associated with the two surgical techniques are presented in Table 4. Since number of deaths differ between laparoscopic and open surgery, costs in the primary analysis were weighted according to the method proposed by Bang and Tsiatis [23]. The difference in mean costs for the societal perspective were $6698 (95% CI 4474 to 8922) and $6513 (95% CI 4644 to 8381) for the healthcare sector, respectively, in the unadjusted analyses. Not weighting costs made little difference, $6494 (95% CI 4360 to 8628) and $6363 (95% CI 4590 to 8135) for the societal and healthcare perspectives, respectively. The difference in unweighted total cost for the index operation was $3136 (95% CI 2492 to 3781). As seen in Table 4, resource-consuming events in inpatient care had the biggest impact on the difference in costs, $6065 (95% CI 4242 to 7887) followed by readmissions $448 (95% CI 232 to 664) and sick-leave $185 (95% CI − 914 to 1285).
Table 4
Mean unadjusted differences in cost (USD, $) open versus laparoscopic colorectal cancer surgery
Mean open surgery (standard deviation) | Laparoscopic surgery (standard deviation) | Difference (standard deviation) | 95% CI | P value | |
|---|---|---|---|---|---|
Societal perspective | |||||
Weighted | |||||
Total costa | 40 217 (41 370) | 33 519 (38 768) | 6698 (99 600) | 4474–8922 | < 0.0001 |
Resource consuming events in inpatient carea | 30 876 (34 196) | 24 811 (30 585) | 6065 (81 619) | 4242–7887 | < 0.0001 |
Readmissiona | 1885 (4120) | 1437 (3346) | 448 (9679) | 232–664 | < 0.0001 |
Sick-leavea | 7457 (20 412) | 7271 (19 329) | 185 (49 243) | -914–1285 | 0.7412 |
Total cost | 45 116 (46 020) | 37 484 (44 010) | 7632 (111 238) | 5148–10 116 | < 0.0001 |
Resource consuming events in inpatient care | 34 389 (38 471) | 27 579 (35 651) | 6811 (92 424) | 4747–8874 | < 0.0001 |
Readmission | 2415 (4956) | 1871 (4255) | 544 (11 747) | 282–806 | < 0.0001 |
Sick leave | 8312 (21 625) | 8035 (20 360) | 277 (52 111) | -886–1441 | 0.6403 |
Unweighted | |||||
Total costa | 38 800 (39 724) | 32 305 (37 087) | 6494 (95 570) | 4360–8628 | < 0.0001 |
Resource consuming events in inpatient carea | 29 884 (32 672) | 23 908 (27 304) | 5976 (77 092) | 4254–7697 | < 0.0001 |
Readmissionsa | 1804 (3926) | 1385 (3210) | 418 (9255) | 212–625 | 0.0001 |
Sick-leavea | 7049 (19 235) | 6836 (18 125) | 214 (46 358) | -821–1249 | 0.6855 |
Total cost | 43 223 (43 813) | 35 894 (41 845) | 7330 (105 875) | 4965–9694 | < 0.0001 |
Resource consuming events in inpatient care | 33 152 (36 919) | 26 624 (34 329) | 6527 (88 752) | 4546–8509 | < 0.0001 |
Readmissions | 2285 (4662) | 1783 (4024) | 502 (11 087) | 254–749 | 0.0001 |
Sick leave | 7793 (20 190) | 7496 (18 927) | 297 (48 612) | -789–1382 | 0.5919 |
Health-care perspective | |||||
Weighted | |||||
Total costa | 32 761 (35 091) | 26 248 (31 194) | 6513 (83 665) | 4644–8381 | < 0.0001 |
Total cost | 36 804 (39 679) | 29 449 (36 831) | 7355 (95 356) | 5226–9484 | < 0.0001 |
Unweighted | |||||
Total costa | 31 750 (33 943) | 25 470 (30 254) | 6281 (80 965) | 4473–8088 | < 0.0001 |
Total cost | 35 430 (38 036) | 28 398 (35 427) | 7033 (91 466) | 4990–9075 | < 0.0001 |
Index episode of care | |||||
Costa | 18 655 (12 738) | 15 519 (7574) | 3136 (28 849) | 2492–3781 | < 0.0001 |
Adjusting for differences between the two techniques (Table 5) with regards to the potential confounders outlined in the method section, decreased the difference to $4505 (95% percentile-based CI 2257 to 6799) for the societal perspective and to $4480 (95% percentile-based CI 2739 to 6203) for the health care sector. Neither weighting the costs nor accounting for different relationships between mean and variance (GLM versus OLS) had significant impact on the results.
Table 5
Mean adjusted differences in cost (USD) open versus laparoscopic colorectal cancer surgery
Adjusted difference | Standard error | 95% CI | P value | |
|---|---|---|---|---|
Societal perspective | ||||
Total costa,b,c | 4504 | 1178 | 2257–6799 | < 0.001 |
Total costb,d | 4944 | 1346 | 2305–7584 | < 0.001 |
Total costa,d | 4319 | 1131 | 2101–6537 | < 0.001 |
Total costd | 4803 | 1279 | 2295–7310 | < 0.001 |
Health-care perspective | ||||
Total costa,b,c | 4480 | 889 | 2739–6203 | < 0.001 |
Total costb,d | 4781 | 1186 | 2456–7106 | < 0.001 |
Total costa,d | 4236 | 979 | 2317–6155 | < 0.001 |
Total costd | 4576 | 1134 | 2353–6800 | < 0.001 |
The results from the deterministic sensitivity analysis can be found in Table 6. Since different regression methods did not change the results ordinary linear regression with the primary cost outcome as dependent variable was used for all sensitivity analyses. A 30% cost increase for resource consuming events in inpatient care for laparoscopic surgery or its mirror image of 30% decrease in resource consuming events in inpatient care for open surgery changed the results the most; in these analyses, open surgery was statistically significantly less costly than laparoscopic surgery. Assuming a 30% rise in sick-leave cost for laparoscopic surgery, or the equivalent decrease for open surgery, laparoscopic surgery was no longer statistically significantly associated with lower cost than open surgery.
Table 6
Mean adjusted deterministic sensitivity analysis of cost differences
Adjusted difference (open-laparoscopy) ($) | 95% CI | |
|---|---|---|
Total costa,b | 3242 | 1577–4906 |
Resource consuming events in inpatient care | ||
Laparoscopy − 30% | 9119 | 7456–9119 |
Laparoscopy + 30% | − 2511 | − 4339–(− )2511 |
Open minus − 30% | − 3473 | − 4935–(− )3473 |
Open plus +30% | 10081 | 8033–10081 |
Readmission | ||
Laparoscopy − 30% | 3640 | 1907–3640 |
Laparoscopy + 30% | 2968 | 1229–2968 |
Open − 30% | 2941 | 1215–2941 |
Open + 30% | 3668 | 1920–3668 |
Sick-leave | ||
Laparoscopy − 30% | 5014 | 3310–5014 |
Laparoscopy + 30% | 1594 | − 184–1594 |
Open − 30% | 1592 | − 66–1592 |
Open + 30% | 5016 | 3177–5016 |
Because both clinical effectiveness and costs were in favor of laparoscopic surgery, no joint analysis of cost-effectiveness was warranted.
Discussion
To our knowledge, this is the first large study of routine care demonstrating that laparoscopic surgery for colorectal cancer was both more effective and less costly than open surgery in a 12-month perspective.
A recent study using Medicare data from the United States concluded that laparoscopic colon cancer surgery was associated with lower expenditures for Medicare beneficiaries [31]. However, the authors did not include rectal cancer surgery in the analysis, neither did they present societal costs for their health economic evaluation. Another large study [32] from the United States used a population-based administrative database to compare economic and clinical outcomes between laparoscopy and open surgery for different colorectal procedures. It concluded that laparoscopy was associated with lower costs and led to better clinical outcomes than open surgery. The study did not focus on colorectal cancer alone, did not study sick-leave costs, and compared charges, which is not directly comparable to cost. A European study [33] from the Netherlands studied accumulated hospital costs during 90 days after index surgery. Data was collected in a population-based database containing 29 hospitals and patients were stratified according to tumor location, ASA-grade and age. Overall, laparoscopy was significantly less costly than open surgery after colon cancer surgery but not after rectal cancer surgery. Again, only health care costs were studied and the follow-up was shorter than in the current study.
Laparoscopic surgery has previously been regarded as costlier to the healthcare sector than open surgery. Earlier studies include factors such as a steeper learning curve, conversion to open surgery and more expensive equipment. The cost difference reported in earlier studies, often RCTs, was not found in this study where we included routine care on a national basis and using a time when higher uptake of laparoscopic surgery was established. Uptake has been slow in Sweden, as can be seen in the percentage of laparoscopic surgery, varying around 20% for colon cancer and 25% for rectal cancer, when for example Denmark [34], Australia [35], Great Britain and Ireland [36] reached these levels some 5 years earlier. This could mean that with even higher uptake the difference in favor of laparoscopy could be larger.
Resource-consuming events in inpatient care was the main contributor to the difference in mean cost. In previous studies, bowel obstruction and incisional hernia [37] and wound infection [37, 38] have been reported to be more prevalent after open than after laparoscopic surgery. Many observations on preoperative clinical T and, to some extent, N-stage were missing in open surgery group. There were also more patients with distant metastases in the open group. It is possible that it was considered less necessary with detailed preoperative assessment in patients with distant metastases.
The national registries used in this study all have high validity [19, 39, 40]. Further, there are no reasons to believe that there was a systematic bias between the surgical techniques regarding classification of ICD-10 or NOMESCO codes. In Sweden, it is mandatory by law to send information regarding inpatient care to government agencies. In addition, the national register on cost per patient has high coverage, about 75% of all episodes of care are registered.
One limitation of this study is that patients were not randomized and thus possible selection bias due to observable and unobservable differences in patient characteristics between the two surgical techniques. To correct for observable imbalance in characteristics, patients in the open colon surgery group were generally frailer than in the other groups, a multivariate regression model was utilized. The results were robust as demonstrated by different regression methods as well as in various sensitivity analyses. The unobserved characteristics might still influence the results, but it is unlikely that those would explain the entire difference across surgical techniques. Moreover, Sheetz et al. [31] found robust evidence for laparoscopic surgery being less costly than open surgery for colon cancer even after controlling for unobserved characteristics using an instrumental variables approach.
For an economic evaluation to be conclusive, the preferred outcome measure is quality-adjusted life-years (QALY), which accounts for both survival and quality of life for the remainder of patients’ life. SCRCR had not yet begun collecting data regarding quality of life for the study period, i.e. estimation of QALYs was not possible. However, no study has concluded that quality of life would be worse after laparoscopic surgery than after open surgery, rather the opposite [5]. Together with the outcomes in this study, lead us to believe that such analysis would strengthen reached conclusions. Furthermore, the SCRCR did not differentiate whether the index procedure was performed using robot-assisted laparoscopic or laparoscopic technique during 2013 and in order to be consistent we decided to include robot-assisted laparoscopic technique in the laparoscopic group for 2014 as well, thus in the entire analysis. According to the SCRCR less than 10% of all rectal cancer prodecures in 2014 were operated on using robot-assisted laparoscopic technique. To avoid bias in cost estimates, the unit cost for the laparoscopic index procedures were applied also to the robot-assisted laparoscopic procedures performed during 2014. The costs accumulated during the 1-year follow-up are assumed to be related to type of index procedure, but it is not expected that laparoscopic and robot-assisted laparoscopic procedure for colorectal cancer would differ in resource-consuming events or readmission during the follow-up.
In the future, it would be interesting to redo the current analysis with quality of life weights collected from the SCRCR and to increase the knowledge of which resource consuming events in inpatient care that contributed to the difference in clinical effectiveness and cost found in this study.
In conclusion, laparoscopic surgical technique can be regarded as preferable from the patient, the societal and the healthcare perspective as primary treatment for colorectal cancer, except those with advanced tumors, as clinical effectiveness and costs were favorable compared with open techniques and other trials have reported no significant differences in long-term oncological outcome.
Acknowledgements
Open access funding provided by University of Gothenburg.
Compliance with ethical standards
Disclosures
Eva Angenete reports grant from Sahlgrenska University Hospital, ALF Agreement concerning research and education of doctors (ALFGBG-716581). Ingela Björholt and Eva Lesén reports remuneration for work paid to employer (PharmaLex Sweden AB) from University of Gothenburg, during the conduct of the study. Eva Haglind reports Grants from Swedish Cancer Society (2016-362), grants from ALF-Gbg (4307771) during the conduct of the study. Jacob Gehrman have no conflicts of interest or financial ties to disclose.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.