The specialized search strategy led to 12,689 results. After removing duplicates, all of the 12,675 titles and abstracts were screened for inclusion. The remaining 158 possibly relevant studies were retrieved as full text articles. Based on the full text assessment, 148 studies were excluded. The majority of the excluded studies did not meet our study design criteria (79 studies). In addition, a number of studies compared different medical treatments and medication or the intervention did not meet our definition of cancer care pathways (23 studies). Finally, 10 studies matched our methodological requirements. In an additional file the PRISMA flow diagram is presented [see Additional file 3]. In Table 1 the key characteristics of the included primary studies included are presented. The excluded full text studies and the reason for exclusion are listed in Additional file 4. For the references of all excluded full text studies, see Additional File 5.
Below we present the studies which were conducted in the secondary healthcare setting separate from the study conducted within both primary and secondary healthcare, because these settings differ, for instance in location, patient popuation and organization.
Results of studies conducted in the secondary healthcare (hospital care) setting
The majority of the included studies (nine studies) were conducted in the setting of secondary healthcare, within hospitals or in oncology centres [19–27]. These studies represented 1,494 patients. The evidence base of pathway interventions in two studies remained “not sure”, these studies are pending review because the authors did not reply to our emails and phone calls [19,25].
The specific interventions regarding the pathways described in the included studies, showed considerable variation. In three studies a description of the pathway as well as a figure of the pathway was provided [20,23,26]. In six studies the care pathway focused on the perioperative phase in order to guide surgical management [20–24,26]. One study presented a pathway for the hospital staff as well as a pathway for patients .
In most studies focussing on pathways for the surgical care, key components which were addressed in these pathways were described. The following components were mentioned most frequently: nutrition and diet [20,22–24,26], diagnostic modalities and laboratory tests [20,22,23,26], medication [20,23,24,26], patient education [22,24,26], preoperative consultation and visits [20,23,26], drains [21,23], activity [20,26], clinical procedures and treatment [20,26], discharge planning or -instruction [20,26], assessment and preadmission testing and evaluation [20,23], and psychosocial support and education [20,26]. Other components which were mentioned in one study only: performance status, outcome criteria, follow-up criteria, and follow up care , pain management and pain control, and deep vein thrombosis prophylaxis , preoperative bowel preparation and fasting, and removal of nasogastric tube , removal of a catheter and mobility .
One study included in this review was not focused on surgical care and investigated pain management, including an initial consultation with a control pain doctor and weekly follow up sessions .
The specification of the study designs of the included studies were based on the description of the Cochrane Effective Practice and Organisation of Care (EPOC) Group .
Randomised controlled study (NRS) designs
We included two studies which applied randomized study designs [24,27]. In these studies patients were randomized to either a pathway group or a non-pathway group.
Interrupted time series studies
Two studies used an interrupted time series study design (ITS), in these studies a pre-pathway group was compared to two or more pathway groups [21,22]. In one study pathway groups at 12 and 36 months after implementation were compared to a pre-pathway group , and in the other study a pre-pathway group and pathway groups at six, 12, and 18 months after implementation were used .
Non-randomized controlled trials
In three studies a non-randomized controlled trial study design was applied [20,23,26]. In these studies patients in the non-pathway group received general care, and simultaneously patient in the pathway group were managed based on the pathway. In one study a historical control group was compared with two other groups; a pathway group and a non-pathway group .
In the articles patients with different tumors were studied: three studies described the effects of pathways for head and neck cancer [20,21,27]. Other studies presented the results of pathways for gastric cancer [23,24], gynaecological cancer , and breast cancer .
The number of included participants varied, and ranged from less than 70 patients to more than 600 patients [20–24,26,27].
Three studies were conducted in in Asia [23,24,26]. In addition, two studies were performed in the United States of America (USA) [21,22], and one in Europe .
Three studies were conducted in general and non-academic hospitals or oncology centres [24,26,27]. Other studies were performed in an academic hospital [20,21,22]. In one study the setting was not clearly reported .
Length of stay was the most common used indicator and was reported in five studies [20–24]. The most frequently used quality indicators were complications [20,21,24] and readmission [20–22]. Other outcomes which were reported in the included studies, were: patient satisfaction [26,27], patient anxiety [26,27], morbidity , and quality of life . However, these quality outcomes measures were not comparable between the studies.
Costs and hospital charges were described in six studies [20–24,27], but these studies showed considerable differences in definitions and results. Nevertheless, in most studies the actual costs instead of charges were reported [20,22–24,27], because costs are set constant over time. In one study the median total charges per patient was used as the primary outcome . In addition, in all studies in which costs were reported, fixed as well as variable costs were included in the total costs. Besides, the studies showed differences in aspects which were included in the total costs. In four studies patient visits, consultation, assessment and diagnostic- and laboratory tests, as well as treatment were included [20,22,24,27]. Medication was included in five studies [20,21,22,24,27]. Facilities, like inpatient ward costs, operation room, medical and surgical supplies were reported in two studies [21,22]. One study included professionals fees . And in another study the costs for patient monitoring and patient education were included .
Additional, two studies reported extra information about the methods for conducting the cost analysis: in one study the hospital and professional costs were combined into a model that has been developed to set costs constant over time  and in another study quality adjusted life days (QALD’s) were generated and the results were presented in a Cost Effectiveness Acceptability Curve (CEAC) related to the willingness to pay .
For the outcome measure “length of stay (LOS)" we were able to carry out a meta-analysis.
Based on the validated criteria suggested by the Effective Organisation of Care Group (EPOC) , all studies were assessed as “high risk of bias”, except for one study. In an additional file the results of the risk of bias assessment are show [see Additional file 6]. To appraise the methodological quality of the included cost evaluations the Evers checklist was used, which is recommended for Cochrane Reviews . See Additional file 7.
Effects on LOS
The effects of cancer care pathways on LOS were reported in five studies [20–24]. All included studies that measured LOS, reported results in favour of cancer pathways. However, in only two studies both mean and standard deviation were reported [22,24]. One study reported the mean LOS, and the SD was calculated by using the Revman Calculator . In another study the median and range was reported and the mean as well as the standard deviation were estimated  (see ‘dealing with missing data’ in the methods section). In one study the median was reported only , therefore we were not able to calculate the mean and standard deviation. Two studies consisted of two subgroups which were separately studied [22,23].
After conducting a meta-analysis with data of four studies [21–24], which represented a study population of 1079 patients, substantial heterogeneity between the studies was observed (I2 = 72%). Therefore a forest plot with the pooled effects of all included studies reporting on LOS was not presented. The results of the meta-analysis of subgroups is presented in the section ‘subgroup analysis’.
Effects on patient outcomes
Quality outcome measures were reported in four studies [20–22,24]. However, only two studies reported the measured effects in terms of complications [21,24], and one study reported outcomes measures in terms of readmissions . Therefore, statistical pooling of quality outcomes could not be performed. Both studies reporting effects of complications described less observed complications among the pathway groups [21,24]. The study reporting effects of readmissions described less readmissions for the pathway group within 30 days after surgery .
Effects on costs
Out of nine studies, six studies reported on costs effects [20–24,27]. In four studies, including two studies with each two subgroups, lower costs were reported for pathway groups [20–22,24], and three of these studies reported a significant reduction of costs related to cancer care pathways [20,23,24]. However, in one subgroup of a study the total hospitals costs and the preoperative costs were lower in the pathway group, but the postoperative costs were higher in the pathway group . Another study reported lower total and medication costs, but higher total daily costs in the pathway group .
Nevertheless, we observed a considerable methodological variation regarding the different methods used for cost calculation. In some studies a full cost approach was used, whereas other studies included only direct hospital costs. For more information about the included and excluded costs or charges, see ‘interventions characteristics’ in the results section. In Table 2 the costs differences are presented. In addition, we have provided the un-discounted cost data in a separate table shown in Additional file 8, to allow readers recalculate the results using any discount rate.
Implementation of cancer care pathways
Information about the implementation of pathways was reported in five studies [20–23,26]. To categorize the detailed information about the reported implementation process of the pathways, we used the refined taxonomy for guideline implementation of Mazza and colleagues . This taxonomy was based on the Cochrane Effective Practice and Organisation of Care (EPOC) data collection checklist and developed to classify the nature and content of implementation strategies. The taxonomy consisted of four domains: professional, financial (healthcare professionals, patients), organizational (healthcare professionals, patients, structural) and regulatory.
Present materials at meetings: in two studies healthcare professionals were given information regarding the pathway in order to implement the pathway adequately [20,26]. Also several conferences and seminars were organized for outpatient and inpatient healthcare teams working with a disease site working group. Further, physicians were briefed on the pathway . In another study nurses underwent a two-hour training session to refresh their information on cancer risk factors, symptoms, diagnostic methods, treatment, pre-operative and post-operative nursing care for patients, and discharge procedures. In addition, doctors and nurses were given information about the clinical pathway and their duties and responsibilities while implementing it .
Creation of an implementation team: in three studies a multidisciplinary group was involved in the development of the pathway [20,21,23]. In one study a core group determined which pathways were developed and a disease site working group was organized to draft the pathway . In another study the pathway was developed and continued to be modified by a multidisciplinary team which included surgeons, nurses, and allied healthcare representatives . In addition, in one other study the involvement of the multidisciplinary team was less clear . In this study the development of the pathways within a multidisciplinary team was not mentioned, but the pathways were based on the results of clinical trials and consensus of experts. Furthermore, these pathways were developed in cooperation with the department of anesthesia pain service and a pharmacist reviewed the recommendations. Moreover, the nursing team played an active role in developing these pathways [ 22].
Change in information & communication technology: in one study was described that almost one year after the pathway was implemented, an electronic medical record (EMR)-based care pathway was being used .
No implementation activities in the regulatory and financial domains were described in the primary studies.
Subgroup analyses were conducted in order to formulate more thorough conclusions relevant for clinical practice. These analyses were performed according to the protocol described previously.
Type of tumor
The included studies were subdivided by type of tumor. We created subgroups of the study with subgroups including patients with gynaecological cancer  and studies including patients with gastric cancer [23,24]. Based on the random effects model, pathways for patients with gastric cancer showed a statistical significant pooled reduction of more than two and a half days compared to usual care (WMD: –2.75; CI: –4.67—0.83). In the study with subgroups including patients with cervical and endometrial cancer, we observed a statistical significant pooled LOS reduction of more than one and a half day (WMD: –1.58; CI: –2.10—1.05). Furthermore, the total pooled LOS reduced almost 2 days (WMD: –1.87; CI: –2.42—1.31), which was a statistical significant result, associated with a moderate amount of heterogeneity (I2 = 50%). Nevertheless, it should be mentioned that the effects on LOS for pathways regarding patients with gynaecological cancer was based on one overall study, which contained two subgroups. See figure 1 with the subgroup analyses of the effects on LOS.
Insert here: Figure 1 Subgroup analyses of the effects on LOS
The primary studies were ordered by country to examine possible different market effects. Therefore studies carried out in North America (both in the USA) versus the studies performed in Asia (Korea and Japan) were analyzed. However, no association between the country and the impact of pathways on LOS was detected.
Year of publication
Subgroup analysis based on the year of publication was similar for the subgroups subdivided by country, which showed no association.
Subgroup analysis based on differences in the setting in which the studies were conducted could not be performed, because one study was performed an academic hospital setting, one study in a non-academic setting and the setting of the other study was unclear. In addition, subgroups based on the risk of bias of the included studies could not be performed also, because all studies were assessed as “high risk of bias” except for one. Furthermore, subgroups based on the age of the population were not possible, due to the fact that the mean age of the included studies was not distinctive enough to form different subgroups.
First, we tested the robustness of the pooled LOS effects using different statistical calculation models, i.e. fixed versus random effects model. The pooled effects changed slightly when using the fixed effects model which indicates reliable pooled results.
In addition, sensitivity analysis were performed to test whether the effects size of LOS varied by the countries were the studies were carried out. Subsequently, we tested the hypothesis that different market forces are possibly confounding the conclusions of our meta-analysis. After stepwise exclusion of the studies carried out in North America, the pooled LOS effect increased, while the statistical heterogeneity reduced (WMD: –2.75; I2 = 59%).
Sensitivity analysis were also performed to analyze the variation in the year of publication in order to test our hypothesis that pathways which were developed and implemented in more recent years could have had more success in reducing the LOS than less recent studies (or vice versa). After stepwise exclusion of all studies published before 2003, one study with two subgroups showed no statistical heterogeneity and a statistical significant pooled LOS effect of –2.03 days.
Results of studies conducted in the primary and secondary healthcare (hospital care) setting
The study about cancer pathways for patients with colorectal cancer, lung cancer, breast cancer, prostate cancer, melanoma and other types of tumor covers both the primary setting as well as the secondary setting . This study represented 3,292 patients.
This study was based on the implementation of standardized cancer patients pathways in 2008. Therefore, dissatisfaction of cancer patients with long waiting times was investigated. Although the cancer pathways did not include the diagnostic workup performed in general practice, the study focused on the time between referral by patients general practitioner (GP) to the first consultation at the hospital. Besides using registered data, patients were questioned about their satisfaction with the waiting times and GP’s were questioned about their involvement in diagnosing the cancer. In the Danish healthcare system the GP serves as a gatekeeper to secondary care and GP’s refer patients to other clinicians when there is a reasonable suspicion of cancer. The authors concluded that the waiting time during the diagnostic process was reduced and patients were more satisfied after implementation of cancer pathways .
More detailed information about the specific pathway which was studied, was found an article included in the list of references . In this publication it was stated that a pathway in the Danish context is a standardized pathway that most patients suspected of cancer will be able to follow. It describes the patient’s pathway from clinical suspicion of a certain cancer through diagnostic procedures and treatment. The pathway describes the medical procedures, the necessary organization encompassing both primary and secondary sectors of the health system, and timeframes in accordance with the political agreement. Main emphasis in pathways is placed on information to be given to the patient, explicit identification of the responsible healthcare professional or department in all phases, procedures for referral, description of multidisciplinary teams in each pathway as a forum for making decisions on diagnosis and recommended treatment, and timeframes of all phases . In addition, the framework of the Danish cancer pathways includes three different descriptions of the pathway: a flowchart, a narrative text and a table providing an organizational overview. An example of a Danish cancer pathway is presented .
The included study used a non-randomized controlled trial study design. A control group was compared to two post-pathway groups, i.e. a pathway group and a non-pathway group .
Types of tumor
Pathways for multiple types of tumor were studied: colorectal cancer, lung cancer, breast cancer, prostate cancer, melanoma and other types of tumors .
This study was performed in Europe .
Patient satisfaction was the reported measured, which was measured using a patient questionnaire and a general practitioner questionnaire. This information was supplemented with register data .
To assess the quality of this study conducted in both primary and secondary care, we adhered to the validated criteria suggested by the Effective Organisation of Care Group (EPOC), see ‘results studies in the secondary care setting, quality assessment’ in the result section. See for the results of the risk of bias assessment Additional file 6.
Implementation of cancer care pathways
The included study described the development of cancer care pathways rather specific, but little information was given about the implementation process. Based on the description regarding the development, we may assume there was multidisciplinary involvement of clinicians and other healthcare professionals. But it remains unclear which role the regional representatives and other relevant healthcare professionals had in the implementation process of the pathway.
Using the taxonomy for guideline implementation , the organizational domain was addressed for at least the development of the presented pathways.
Creation of an multidisciplinary team: in this study the development of cancer care pathways was based on a common framework. Subsequently, healthcare professionals formulated the medical content in accordance with the consensus-based framework and finally the pathways were approved by a two-step process involving all stakeholders. This framework ensured that all stakeholders were able to influence the process which could be characterized as a ‘bottom-up and top-down’ approach with involvement of both local and central actors, and in which administrators, healthcare professionals, and politicians cooperated to strike a balance. Further, agreeing on a framework and the integration of needs from the view of various professional disciplines, created a common understanding on how the best possible pathway was acquired. The framework was used for all cancer types to ensure consistency and ease the implementation of the various pathways.
In addition, for the development of these pathways working groups developed the content using a consensus-making process where all stakeholders participated actively and contributed to the final product. The clinical working groups were asked to describe standard timeframes for the various elements involved in each pathway. These timeframes were further estimated without consideration of existing capacity and resources and thus indicate the minimum time needed to treat an “ideal patient” in an “ideal health system”.
Once a pathway had formally been approved, the five health regions had three months to ensure implementation at the local level. The regional representatives in the working groups knew the pathways in details which was an important factor in ensuring the implementation process .