Screening for oesophageal cancer

Abstract

Background

Oesophageal cancer is a global heath problem. The prognosis for advanced oesophageal cancer is generally unfavourable, but early‐stage asymptomatic oesophageal cancer is basically curable and could achieve better survival rates. The two most commonly used tests are cytologic examination and endoscopy with mucosal iodine staining. The efficacy of the screening tests is controversial, and the true benefit and efficacy of screening remains uncertain because of the potential lead‐time and length‐time biases. This review was conducted to examine the evidence for the efficacy of screening for oesophageal cancer (squamous cell carcinoma and adenocarcinoma).

Objectives

To determine the efficacy of early screening, using endoscopy with iodine staining or cytologic examination, in reducing mortality from oesophageal cancer in asymptomatic individuals from high‐risk and general populations.

Search methods

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (2012, Issue 8), The Cochrane Library (2012, Issue 8), MEDLINE (1950 to August 2012), EMBASE (1980 to August 2012), Allied and Complementary Medicine (AMED) (1985 to August 2012), Chinese Biomedical Database (CBM) (January 1975 to August 2012), VIP Database (January 1989 to August 2012), China National Knowledge Infrastructure (CNKI) (January 1979 to August 2012), and the Internet. We also searched reference lists, conference proceedings, and databases of ongoing trials. There was no restriction on language or publication status in the search for trials.

Selection criteria

We included only randomised controlled trials (RCT) of screening versus no screening for oesophageal cancer. Randomisation of groups or clusters of individuals was acceptable.

Data collection and analysis

Two review authors independently scanned the titles and abstracts from the initial search for potential trials for inclusion. We did not find any trials that met the inclusion criteria.

Main results

The electronic search identified 3482 studies. Two authors independently reviewed the references. The reports of 18 studies were retrieved for further investigation. None met the eligibility criteria for a RCT investigation of the effects of screening versus no screening for oesophageal cancer.

Authors' conclusions

There were no RCTs that determined the efficacy of screening for oesophageal cancer. Non‐RCTs showed a high incidence and the reported better survival after screening could be caused by selection bias, lead‐time and length‐time biases. RCTs are needed to determine the efficacy of screening for oesophageal cancer.

Plain language summary

Screening for oesophageal cancer

Oesophageal cancer is a common cause of cancer related death in the world. The prognosis in the advanced stages is unfavourable, but the early oesophageal cancers are asymptomatic and curable, and usually go undetected until they have spread beyond the oesophageal wall. Endoscopy with iodine staining or cytologic examination are two common screening tests for early oesophageal cancer. These screening tests were started in the 1970s, but the true benefit is inconsistent and is uncertain due to lead‐time bias, which is the amount of time by which the diagnosis is advanced by the screening procedure; and length‐time bias, where screening is more likely to detect slow‐growing disease rather than altering the person's duration of life. This review intended to determine the efficacy of the two screening tests for oesophageal cancer. We identified 3482 studies but none were RCTs of screening. Several non‐comparative studies showed that the screening tests may increase the incidence of reported oesophageal cancer and improve the survival results after a screening test, but these results could indicate bias rather than a true causative effect. Therefore, there is a strong need for randomised controlled trials (RCTs), especially long‐term RCTs, to determine the efficacy, cost‐effectiveness, and any adverse effects of screening for oesophageal cancer.

Summary of findings

Summary of findings for the main comparison. Screening for oesophageal cancer.

Screening for oesophageal cancer
Patient or population: patients with oesophageal cancer Settings: Intervention: screening
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Screening
Oesophageal cancer‐specific mortality ‐ not measured	See comment	See comment	Not estimable	‐	See comment	There were no studies eligible for inclusion.
Secondary outcomes ‐ not measured	See comment	See comment	Not estimable	‐	See comment	There were no studies eligible for inclusion.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

Background

Description of the condition

Oesophageal cancer is a global heath problem, and was the sixth most common cancer in 2002 (Parkin 2005). It is characterized by a wide geographical variation at an international level (Corley 2001; Ferlay 2004; Parkin 2005). Oesophageal cancer is prominent in Ethiopia, China, and Mongolia, with male and female standardized incidence rates as high as 28.1 per 100,000 and 19.6 per 100,000, respectively (IACR 2002). This cancer occurs more frequently in males than in females, and the ratio of males to females is from 3 to 5:1 (Helena 2007).

There are two main histopathologic types of oesophageal cancer, squamous cell carcinoma (SCC) and adenocarcinoma (AC). SCC is the dominant histologic type both in Western and Asian countries (Engel 2003). AC is primarily found in developed countries, specifically in the United States of America (US) and Europe. The overall incidence of oesophageal cancer is on the rise, and this increase over the past two decades coincides with a change in the histologic type and the location (Botterweck 2000; Layke 2006). In the 1960s, SCCs comprised more than 90% of all oesophageal tumours (NCI 2008). During the past two decades the incidence of oesophageal AC has risen considerably, such as in the US and Europe, and now the incidence of AC and of SCC are approximately equal (Helena 2007).

The prognosis for patients with oesophageal cancer is generally unfavourable. In China, the overall five‐year survival rate for advanced oesophageal cancer is only around 20% (Zhang 2008). In the US, the overall five‐year relative survival rate for people with oesophageal cancer was the fifth lowest at 15.4% (Wang 2005a). Even among patients with radical oesophagectomies, the five‐year survival rate is only 10% to 20% (Coleman 2003; Crane 2008). Early oesophageal cancer is usually asymptomatic, and it has usually spread beyond the oesophageal wall when the cancer is detected (Dawsey 1998). However, early‐stage asymptomatic oesophageal cancer is basically curable with, for instance, intervention with conventional surgery and endoscopic resection, and studies have shown five‐year survival rates that reach 90% or above (Wang 2003; Wang 2004b; Tachibana 2006). Clearly there is a need for screening strategies for early detection of precursor lesions and early invasive oesophageal cancers.

Description of the intervention

A large number of surveillance and screening studies have been reported for oesophageal cancer (Lu 1988; Hardwick 1997; Incarbone 2002; Wang 2003; Lu 2004; Wang 2005). Surveillance is the intermittent examination of a patient with a known precancerous lesion, while screening is the examination of an asymptomatic at risk population to detect a precancerous lesion. Two of the most commonly used screening methods, cytologic examination and endoscopy with mucosal iodine staining, were investigated our study. Both of these can be costly methods. There are two types of cytologic examination. Balloon cytologic examination was developed in China (Shen 1982; Shu 1983; Shen 1984; Shu 1985), which involves inserting a rubber balloon with a cotton mesh down the back of the throat into the stomach to collect the oesophageal cells. The two methods are easy to conduct and can be done in an out‐patient clinic. Sponge cytologic examination was developed in Japan (Nabeya 1979) and is characterized by swallowing a gelatin capsule covered by a sphere of polyurethane mesh. After the gelatin capsule dissolves and the sponge expands in the stomach over a five minute period, the sponge is then gradually pulled out along with oesophageal cells (Roth 1997).

Endoscopy with mucosal iodine staining is a very sensitive screening technique and is effective for early detection of oesophageal cancer (Dawsey 1998). Iodine staining moderately improves the diagnostic accuracy of videoendoscopy. The disadvantage of this method is that it is expensive, inconvenient, and is usually done in a hospital setting (NCI 2008). For oesophageal AC, the incidence of AC is low in the US and European Union (about 4 to 8 per 100,000) (IACR 2002), and the incidence is also low for patients with Barrett's oesophagus (metaplasia in the inferior portion of the oesophagus), about 1/52 to 1/285 patient years (Hameeteman 1989; Caygill 1999; O'Connor 1999; Shaheen 2002; Murray 2003; Basu 2004). Furthermore the screening method, endoscopy, for use in a population to detect the precancerous lesion of AC is controversial. Five‐year survival rates for advanced AC are less than 20%, whereas the survival for early‐stage esophageal AC range from 57% for stage T1sm (invading the submucosa) to more than 80% for T1m (invading the mucosa only) disease (Rice 2003; Epari 2009). The longer survival time may be the result of lead‐time and length‐time biases.

There are other screening methods for oesophageal cancer including radiology, faecal occult blood test, ultra‐thin nasal endoscopy, methylene blue staining at endoscopy, assessing clinical presentations of skin changes in people with an inherited predisposition such as tylosis (horny growths and thickening of skin on palms and soles), and questionnaires about symptoms such as severity of heartburn symptoms. However, these screening methods have poor sensitivity for detecting precursor lesions and early invasive oesophageal cancers, and have little generalisability. Moreover, questionnaires about symptoms may be reliable only when the cancer is relatively advanced. Therefore, we did not include these screening methods in the review.

How the intervention might work

Several prospective cohort studies have shown that early screening using cytology or endoscopy with iodine staining was effective in the detection of early oesophageal cancer. In turn, this could contribute to the early treatment of oesophageal cancer and increase cure rates and reduce the mortality from such a malignancy (Lazarus 1992; Roth 1997; Kouzu 1998; Wang 2003; Lu 2004; Wang 2004a; Guo 2005; Wang 2005b; Moschler 2006). A 24‐year follow‐up cohort early screening trial including 22,285 participants in oesophageal cancer high‐risk areas of China has shown that endoscopy screening followed by early therapy could increase the five‐year and 10‐year survival rates with oesophageal cancer, and the incidence could be decreased by a third (Wang 2001). Another 25‐year follow‐up study in a high‐risk population in China has revealed that balloon cytology and endoscopy with iodine staining may be the best approaches to screening for oesophageal cancer, and patients undergoing early surgical treatment may benefit from increased longer‐term survival with an acceptable quality of life (Wang 2004b). A randomised controlled trial of endoscopy with iodine staining versus no screening for oesophageal cancer is being conducted in Linxian, China (CICAMS 2007).

Why it is important to do this review

The aim of these screening methods for oesophageal cancer is to decrease mortality, improve the chances of a cure, and increase the patient's quality of life. However, the true benefits and efficacy of screening remain uncertain because of the potential lead‐time bias (increasing the proportion of life with known cancer without actually extending the duration of life) and length‐time bias (with screening more likely to detect slow‐growing disease rather than altering the duration of life).

Endoscopy with mucosal iodine staining is widely used for detecting oesophageal cancer in Western countries but it has not decreased the mortality as a result of AC in the general population, so the efficacy of this method for screening is controversial (Gerson 2002; Sharma 2004; NCI 2008; Rubenstein 2008). The use of endoscopy with mucosal iodine staining is limited by the complex manipulation and cost in developing countries such as China. In a high‐risk population of Linxian in China, where age standardized incidence rates of oesophageal cancer for both sexes exceed 100 per 100,000 per year (Ke 2002), using cytologic screening tests as a primary screening test produced high false‐negative results and low accuracy (Guo 2008; Pan 2008). The severe discomfort experienced by patients contributes to low acceptance of this method of screening (Roth 1997; Guo 2008; Pan 2008). However, for AC, although the screening method could improve the survival the strategy is unlikely to be cost‐effective due to its low incidence (Lagergren 1999; Rubenstein 2008; Hirst 2011).

Furthermore, other problems may be associated with these screening tests, such as perforation and bleeding as a result of endoscopic examination, allergic reactions to iodine, false positives contributing to psychological trauma, over‐diagnosis and over‐treatment, potentially false negatives, potentially invasive nature of the intervention, and other complications as a result of these tests. Also, there is little known about the cost‐effectiveness of these screening tests.

In our review, we assessed the evidence regarding the ability of these screening methods to detect oesophageal cancer (squamous cell carcinoma (SCC) and adenocarcinoma (AC)), the impact on mortality, cost‐effectiveness, and quality of life, and the possible harms.

Objectives

Primary objective

To determine the efficacy of early screening, using endoscopy with iodine staining or cytologic examination, in reducing mortality from oesophageal cancer in asymptomatic individuals from high‐risk and general populations.

Secondary objectives

To assess the impact of screening for oesophageal cancer on the incidence of positive cytology or histology, survival, adverse effects, compliance with screening, costs, all cause mortality, and quality of life.

Methods

Criteria for considering studies for this review

Types of studies

We included only randomised controlled trials (RCT) of screening versus no screening for oesophageal cancer. Randomisation of groups or clusters of individuals was acceptable. There were no restrictions on language or publication status.

Types of participants

All participants enrolled in studies of oesophageal cancer screening were eligible for this review, with no restriction on ethnicity or age. Moreover, participants from high‐risk populations and the general population were eligible for inclusion. We excluded individual participants with a previous diagnosis of oesophageal cancer.

Types of interventions

We included mass screening, targeted screening, and opportunistic screening that used endoscopic examination with iodine staining or cytologic examination (including balloon cytologic examination and sponge cytologic examination), individually or in combination.

Types of outcome measures

Primary outcomes

The primary outcome considered in this review was oesophageal cancer specific mortality.

Secondary outcomes

1. Incidence of positive cytology or histology

2. Survival rates (three‐year, five‐year, and 10‐year)

3. Harms of screening (including complications of the screening tests; the adverse outcomes from false‐negative and false‐positive results on initial screening such as delay in clinical diagnosis of 'missed' cases because of false‐negative result, psychological trauma, over‐diagnosis and over‐treatment because of false‐positive result)

4. Compliance with screening

5. Costs

6. All cause mortality

7. Quality of life

Search methods for identification of studies

See: Cochrane Upper GastrointestinaI and Pancreatic Diseases Group methods used in reviews.

Electronic searches

A search was conducted to identify all published and unpublished RCTs.

Trials were identified by searching the following electronic databases:

• Cochrane Central Register of Controlled Trials (CENTRAL) and The Cochrane Library (Issue 8, 2012) (Appendix 1);

• MEDLINE (1950 to week 2 August 2012) (Appendix 2);

• EMBASE (1980 to Week 33 2012) (Appendix 3);

• Allied and Complementary Medicine (AMED) (1985 to August 2012) (Appendix 1);

• Chinese Biomedical Database (CBM) (January 1975 to August 2012) (Appendix 4);

• VIP Database (January 1989 to August 2012) (Appendix 5);

• China National Knowledge Infrastructure (CNKI) (January 1979 to August 2012) (Appendix 5).

We also searched reference lists, conference proceedings, and databases of ongoing trials.

The search strategy was constructed in an attempt to identify all relevant studies regardless of language or publication status (published, unpublished, in press, and in progress) by using a combination of MeSH subject headings and text words related to screening for oesophageal cancer.

We searched the following databases of conference proceedings:

• China Medical Academic Conference Database (1994 to August 2012) (Appendix 1);

• MEDLINE (1950 to August Week 2 2012) (Appendix 2);

• Science Citation Index at Web of Science (1965 to August 2012) (Appendix 1).

We also searched the following databases of ongoing trials:

• Current Controlled Trials (www.controlled‐trials.com);

• The National Research Register (www.update‐software.com/National/nrr‐frame.html);

• World Health Organization (WHO) International Clinical Trials Registry Platform (http://www.who.int/trialsearch/);

• National Institutes of Health (www.clinicaltrials.gov);

• CenterWatch Clinical Trials Listing Service (www.CenterWatch.com/);

• National Cancer Institute (http://www.cancer.gov/clinicaltrials);

• UK National Research Register of all NHS‐funded research (https://portal.nihr.ac.uk/Pages/NRRArchive.aspx).

Searching other resources

Handsearches

We searched the following journals from inception to the latest issue:

• Zhonghua Zhong Liu Za Zhi;

• Cancer;

• Endoscopy;

• Chinese Journal of Clinical Oncology;

• Chinese Journal of Digestive Endoscopy.

Additional searches

We contacted related organizations, authors of primary studies, individual researchers, and experts working in this field to obtain additional relevant unpublished or published studies, studies in progress, confidential reports, and raw data of published studies, if possible.

Data collection and analysis

Selection of studies

To determine the studies to be further assessed, two review authors (Yang SJ, Shao Y) independently scanned the titles and abstracts from the initial search for potential trials. The full text was retrieved for further assessment if the information given suggested that the study:

1. included participants without a previous diagnosis of oesophageal cancer;

2. compared screening versus no screening for oesophageal cancer;

3. assessed one or more relevant outcome measures as specified in this review;

4. included RCTs of screening for oesophageal cancer.

If there was any doubt regarding these criteria, from the information given in the title and abstract, the full article was retrieved for clarification. Inter‐rater agreement for study selection was measured using the Kappa statistic (Cohen 1960). Where differences in opinion existed, they were resolved by a third party (Chen XY (CXY)). If resolution of the disagreement was not possible, we added the article to 'studies awaiting assessment' and contacted the authors for clarification. If no clarification was provided, we were to consult with the review group editorial base.

Data extraction and management

We planned to extract the data using a piloted data extraction form, including study methods (such as type of design and information on quality such as control of confounders), participants, interventions, follow‐up, drop‐out rate, and outcomes. We resolved any disagreements by referring to the trial report and through discussion, or by consulting with YSJ. If data from the trial reports were insufficient or missing, we would contact the authors for additional information.

If possible, we would extract data to allow an intention‐to‐treat analysis (the analysis should include all the participants in the groups to which they were originally randomly assigned). If the number randomised and the numbers analysed were inconsistent, we would calculate the percentage loss to follow‐up and report this information in an additional table.

Assessment of risk of bias in included studies

We planned to assess the methodological quality as described by the Cochrane Handbook for Systematic Reviews of Interventions Version 5.0.0 (Higgins 2008).

Randomised controlled trials (RCT) would be assessed as follows.

1. Generation of the allocation sequence: adequate (computer‐generated random numbers, table of random numbers, coin tossing or similar); unclear (the trial was described as randomised but the generation of the allocation sequence was not described); inadequate (other methods).

2. Allocation concealment: adequate (central randomisation, sealed envelopes, or similar); unclear (the allocation concealment is not described); inadequate (open table of random numbers or similar).

3. Blinding: adequate (no blinding but the review authors who judged the outcome or outcome measures and participants would not be influenced by lack of blinding; the participants and key study personnel were blinded and the blinding was unlikely to be broken; either participants or some key study personnel were not blinded but outcome assessment was blinded and the non‐blinding of others was unlikely to introduce bias); unclear (the blinding methods were not described); inadequate (no blinding or incomplete blinding, and the outcome or outcome measurement was likely to be influenced by lack of blinding; either participants or some key study personnel were not blinded, and the non‐blinding of others was likely to introduce bias). (The trials we will include are screening versus no screening for oesophageal cancer whereby participants must be aware that they are undergoing an intervention or not and outcome assessors would know whether participants are in the intervention group or control group, so blinding is not possible.)

4. Incomplete outcome data: adequate (no missing outcome data; number and reasons for missing outcome data are unlikely to be related to true outcome; numbers or reasons for missing outcome data were balanced in numbers across groups; missing data have been imputed with appropriate methods); unclear (numbers or reasons for missing outcome data were not described); inadequate (reasons for missing outcome data were likely to be related to true outcome; numbers or reasons for missing outcome data did not balance across groups; potentially inappropriate use of simple imputation for missing outcome data).

5. Selective outcome reporting: adequate (the study protocol is available; all the pre‐specified outcomes of the study have been reported; not available from study protocol, but the expected outcomes including pre‐specified outcomes were clear in the published reports); unclear (the protocol and pre‐specified outcomes were not described); inadequate (not all the pre‐specified primary outcomes have been reported; one or more primary outcomes were measurements, analysis methods or subsets of the data that were not pre‐specified; one or more outcomes of interest in the review were reported incompletely).

Based on these criteria, RCTs would be broadly subdivided into the following three categories:

1. all quality criteria met, low risk of bias;

2. one or more of the quality criteria only partly met, moderate risk of bias;

3. one or more quality criteria not met, high risk of bias.

The results of the assessment of methodological quality would be used for subgroup analysis or sensitivity analysis.

Measures of treatment effect

For dichotomous outcomes, measures of effect would be expressed as relative risk (RR), with 95% confidence interval (CI). Pooled results would be analysed using either a fixed‐effect or random‐effects model, depending on the level of heterogeneity. Measures of effect for continuous outcome scores would be expressed as mean differences with 95% CI.

For continuous outcomes, measures would be analysed using weighted mean difference (WMD) and the 95% CI if studies report the outcome using the same scale. If different scales were used, we would present the data as standardized difference in means (SMD) with 95% CI. Utilizing either a fixed‐effect or random‐effects model would depend on the level of heterogeneity. If there was substantial clinical or statistical heterogeneity, study results would not be combined in a meta‐analysis.

Survival analysis methods would be used to summarize time‐to‐event data and to express the intervention effect as a hazard ratio (HR) for each study, when possible. In this case we would combine data in an overall summary estimate using the methods of Whitehead et al (Whitehead 1991).

Unit of analysis issues

Cluster randomised trials

These are studies employing 'cluster randomisation' (such as randomisation by clinician or practice) where analysis and pooling of clustered data would lead to a 'unit of analysis' error whereby P values are spuriously low, CIs unduly narrow, and statistical significance overestimated. We would combine data from these cluster randomised trials using the generic inverse variance method. If the study has been analysed as if the randomisation was based on individuals rather than on the clusters, we would correct the analyses by extracting the number of clusters or the average size of each cluster, the outcome data ignoring the clustering design for the total number of individuals, and an estimate of the intra‐cluster correlation coefficient (ICC). We would reduce the size of each trial to its 'effective sample size' (Rao 1992). The effective sample size of a single intervention group in a cluster randomised trial is its original sample size divided by a quantity called the 'design effect'. The design effect is 1 + (M – 1) ICC (Donner 2002), where M is the average cluster size and ICC is the intra‐cluster correlation coefficient. If the ICC was not reported it would be assumed to be 0.1 (Ukoumunne 1999).

Dealing with missing data

We planned to acquire any missing or unpublished data by contacting the authors of identified studies by writing standard letters or sending emails. When the data were not missing at random, we would use the principles cited in the Cochrane Handbook for Systematic Reviews of Interventions for dealing with missing data (Higgins 2008), performing intention‐to‐treat (ITT) analysis, or imputing the missing data using the last reported observation carried forward (LOCF). This means that the most recently observed outcome measure is assumed to hold for all subsequent outcome assessment times (Lachin 2000; Unnebrink 2001). If the data were not able to be accessed or were missing at random, we would analyse only the available data and address the potential impact of missing data on the findings of the review in the discussion section.

Assessment of heterogeneity

Heterogeneity would be identified by graphical interpretation and the results for the I² statistic.

Assessment of reporting biases

A comprehensive search for studies that met the eligibility criteria would be conducted, including searching for unpublished studies and the trial registries. Funnel plots would be used to examine the possibility of reporting biases if sufficient studies (15) were found. Asymmetry of a funnel plot could be due to reporting biases, but could also be due to other factors such as true heterogeneity or poor methodological design of small studies. We would analyse the possible reasons for funnel plot asymmetry.

Data synthesis

Data would have been summarized statistically using Review Manager 5.0.0. Statistical analysis would have been performed according to the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions Version 5.0.0 (Higgins 2008).

Subgroup analysis and investigation of heterogeneity

We intended to perform subgroup analyses in order to explore effect size differences according to:

1. screening interventions;

2. the different cytologic and histologic diagnosis of oesophageal cancer;

3. different types of oesophageal cancer;

4. sex, age and ethnicity;

5. duration of follow‐up;

6. frequency of screening;

7. the economic status (developed areas or undeveloped areas).

Sensitivity analysis

We would have performed sensitivity analyses in order to explore the influence of the following factors on effect size:

1. repeating the analysis excluding unpublished studies (if there were any);

2. repeating the analysis taking account study quality, as specified above;

3. repeating the analysis excluding studies using the filters of language of publication, industry funded, country.

The robustness of the results would also be tested by repeating the analysis using different measures of effects size (risk difference, odds ratio. etc) and different statistical models (fixed‐effect and random‐effects models). A funnel plot would have been used to examine the possibility of publication bias (Egger 1997).

Results

Description of studies

Results of the search

The electronic search identified 3482 studies, shown in Figure 1. Two authors independently reviewed the references and, finally, the full texts of 18 studies were retrieved for further investigation. None met the eligibility criteria for an RCT investigation of the efficacy of screening versus no screening for oesophageal cancer. Therefore, it was not possible to include any data for further analysis regarding this topic.

1.

Study flow diagram.

Eight uncontrolled trials described the prevalence of oesophageal cancer or precursor dysplastic changes by screening methods, and four uncontrolled trials focused on the sensitivity and specificity of screening methods for oesophageal cancer. One uncontrolled trial, one retrospective controlled trial, and one retrospective cohort study looked at survival after screening but without a comparison group. One non‐randomised controlled trial and two uncontrolled trials described the incidence of oesophageal cancer after screening. One self‐controlled, single‐blind, crossover study described the accuracy of cytologic examination.

Included studies

No studies were eligible for inclusion.

Excluded studies

See Characteristics of excluded studies.

Risk of bias in included studies

There were no included studies, so bias could not be evaluated.

Effects of interventions

See: Table 1

There were no included studies, so effects could not be evaluated.

Discussion

We expected to find evidence to explore the efficacy of screening for oesophageal cancer. However, no RCTs examining the efficacy of screening by using endoscopy with iodine staining or cytologic examination for oesophageal cancer were found with our literature search. Thus, no judgement on effectiveness can be made due to the lack of methodological rigour in the available studies, which suggest further research is worth doing on this question.

Does screening improve the outcomes of oesophageal cancer?

A successful screening program should have tested whether the screening test improves outcomes (most importantly mortality), yet currently no RCTs have been found to definitively answer this question. We found three trials described the incidence of oesophageal cancer after screening (Lu 1988; Wang 2001; Wang 2005) and three trials focused on the survival of people with oesophageal cancer after screening (Wang 2001, Incarbone 2002, Rubenstein 2008).

Wang 2001 and Wang 2005 are two uncontrolled trials that explore the incidence of oesophageal cancer after a cytological screening test. Because uncontrolled trials can not draw any conclusion about whether an intervention is effective, we could only increase our awareness of the incidence of oesophageal cancer in a specific population. Lu 1988 reported that the incidence of oesophageal cancer was 2.39 times higher than that found in a normal control group after cytological screening. But we also could not come to the conclusion that the cytological screening was effective. Firstly, this was because the high incidence may be the result of selection bias because of the inadequate control. Secondly, it is unclear whether the oesophageal cancer detected by the screening test could have led to a better prognosis (mortality) than in a normal control group.

The three trials on survival results reported that the people with early SCC detected by the screening test had excellent long‐term survival with acceptable quality of life. However, the benefit of screening for AC is not as good as for SCC. The main reason might be the variation in cancer incidence between SCC and AC. The incidence of the precancerous Barrett's oesophagus is about 2% to 3% in the population with a cancer rate of 0.5% to 1% per year, while the incidence of precancerous SCC (severe dysplasia) is about 1% to 2% of the population in a high‐risk area for SCC in China, and the cancer rate of severe dysplasia is about 10% per year. The lower cancer rate of precancerous Barrett's oesophagus would introduce an unsatisfactory survival benefit with screening (Hardwick 1997; Incarbone 2002). Rubenstein 2008 reported that the earlier stage of AC that is detected by endoscopy did not lead to a significant improvement in survival. The survival data from the non‐RCTs could introduce significant lead‐time bias (increasing the proportion of a person's life with known cancer without actually extending the duration of life). Also, the better survival could be caused by length‐time bias (screening being more likely to detect slow‐growing disease rather than altering the duration of life), and Rubenstein 2008 acknowledged that slightly improved survival is more likely to be caused by lead‐time and length‐time effects than a true causative effect.

Therefore, the difference in histopathology, bias, and the controversial results mean that we are unable to make any judgement on the effectiveness of the screening test, and further RCTs are needed to evaluate the effect of screening on the mortality of people with oesophageal cancer.

Is screening cost‐effective?

This issue is both practical and important for a screening program. Whether screening is cost‐effective could guide public health policy, but no RCTs were found to answer this point. Up to now, three cost‐effectiveness studies indicate that the screening may be cost‐effective (Soni 2000; Inadomi 2003; Liu 2006). Soni 2000 used a decision analysis model to compare the cost‐effectiveness of screening versus no screening, and found that the endoscopic screening cost 24,700 dollars per life‐year saved. Inadomi 2003 used a Markov model to compare the screening with a no screening strategy. The endoscopic screening cost 10,440 dollars per life‐year saved, but follow‐up endoscopic surveillance of patients with Barrett's oesophagus but no dysplasia was highly cost‐ineffective. The results of the two modelling studies are easily affected by the baseline assumptions, that is the cost‐effectiveness result is sensitive to the prevalence, sensitivity and specificity, as well as cost of screening endoscopy, and thus can be easily changed from a life‐saving (refers to lower cost of saving a life year than the value created by a life year) into a life‐losing strategy (refers to higher cost of saving a life year than the value created by a life year). Liu 2006 used an observational study to look at whether the cost‐benefit of screening strategy is effective, but did not consider the long‐term cost and surveillance cost of the screening strategy. These results could not be used to draw conclusions on whether the screening strategy is cost‐effective. Therefore, further RCTs are needed to determine the cost‐effectiveness of screening.

Is screening acceptable?

No RCTs were found on this question. Patients undergoing endoscopy may have a potential risk for perforation, haemorrhage, or death, and the incidence is estimated to be in the range of 0.0013 to 0.00009 (Gerson 2004), which seems low. No RCTs were found investigating the psychological trauma, over‐diagnosis and over‐treatment because of false‐positive results. The implications of adverse effects, for the outcome and cost‐effectiveness of the screening test, need to be determined with further research.

Authors' conclusions

Implications for practice.

On the basis of the current evidence, there are no RCTs to determine the efficacy of screening for oesophageal cancer.

No RCTs are available on whether the screening test could decrease the mortality of individuals with oesophageal cancer. There are studies on the incidence and survival of oesophageal cancer after screening, but the high incidence after screening in the studies could be caused by selection bias, and the excellent survival results may be introduced by lead‐time and length‐time biases. Whether the screening test could improve the outcomes of this cancer can not be concluded.

The true cost‐effectiveness of screening could be judged by long‐term follow‐up RCTs, but this type of study is time consuming and expensive. This is why the economic model is used to evaluate the screening test. The results of modelling studies are easily changed and affected by the baseline assumptions.

Therefore, whether the screening test is effective in the management of oesophageal cancer can not be evaluated and concluded. There is a strong need for RCTs to explore the efficacy of screening methods for oesophageal cancer.

Implications for research.

The efficacy of screening test for oesophageal cancer can only be evaluated and concluded with high quality RCTs, especially a long‐term, long follow‐up study on the mortality.

Appendices

Appendix 1. Cochrane Central Register of Controlled Trials and AMED search strategies

The search strategy for this review has been constructed by using a combination of MeSH subject headings and text words relating to screening for oesophageal cancer. To identify randomised controlled trials, the following search will be combined with the Cochrane highly sensitive search strategy phases one, two and three as contained in the Cochrane Handbook for Systematic Reviews of Interventions Version 5.0.0 (Higgins 2008).

1. esophag$.ab,ti.

2. oesophag$.ab,ti.

3. 1 or 2

4. (carcino$ or cancer$ or neoplasm$ or tumour$ or tumor$ or cyst$ or growth).mp.

5. adenocarcinoma$.ab,ti.

6. Precancerous Conditions/ or Carcinoma, Squamous Cell/ or precursor lesion$.mp. or Esophageal Neoplasms/ or Ulcer/ or Peptic Ulcer/

7. or/4‐6

8. 3 and 7

9. mucosal iodine staining.ab,ti.

10. endoscop$.ab,ti.

11. Iodine/ or "Staining and Labeling"/

12. screening.mp. and 3

13. cytolog$.ab,ti.

14. balloon cytolog$.ab,ti.

15. sponge cytolog$.ab,ti.

16. cytodiagnos$.ab,ti.

17. videoendoscop$.ab,ti.

18. or/9‐17

19. screening compliance.ab,ti.

20. mortality rate.ab,ti.

21. cost‐effectiveness.ab,ti.

22. quality of life.ab,ti.

23. survival rate.ab,ti.

24. adverse effect.ab,ti.

25. or/19‐24

26. 8 and 18

27. 8 and 25

28. 26 or 27

Appendix 2. MEDLINE search strategy

1. randomized controlled trial.pt.

2. controlled clinical trial.pt.

3. randomized.ab.

4. placebo.ab.

5. drug therapy.fs.

6. randomly.ab.

7. trial.ab.

8. groups.ab.

9. or/1‐8

10. (animals not (humans and animals)).sh.

11. 9 not 10

12. esophag$.ab,ti.

13. oesophag$.ab,ti.

14. 12 or 13

15. (carcino$ or cancer$ or neoplasm$ or tumour$ or tumor$ or cyst$ or growth).mp.

16. adenocarcinoma$.ab,ti.

17. Precancerous Conditions/ or Carcinoma, Squamous Cell/ or precursor lesion$.mp. or Esophageal Neoplasms/ or Ulcer/ or Peptic Ulcer/

18. or/15‐17

19. 14 and 18

20. mucosal iodine staining.ab,ti.

21. endoscop$.ab,ti.

22. Iodine/ or "Staining and Labeling"/

23. screening.mp. and 14

24. cytolog$.ab,ti.

25. balloon cytolog$.ab,ti.

26. sponge cytolog$.ab,ti.

27. cytodiagnos$.ab,ti.

28. videoendoscop$.ab,ti.

29. or/20‐28

30. screening compliance.ab,ti.

31. mortality rate.ab,ti.

32. cost‐effectiveness.ab,ti.

33. quality of life.ab,ti.

34. survival rate.ab,ti.

35. adverse effect.ab,ti.

36. or/30‐35

37. 19 and 29

38. 19 and 36

39. 37 or 38

40. 11 and 39

Appendix 3. EMBASE search strategy

1. (random$ or placebo$).ti,ab.

2. ((single$ or double$ or triple$ or treble$) and (blind$ or mask$)).ti,ab.

3. controlled clinical trial$.ti,ab.

4. RETRACTED ARTICLE/

5. or/1‐4

6. (animal$ not human$).sh,hw.

7. 5 not 6

8. esophag$.ab,ti.

9. oesophag$.ab,ti.

10. 8 or 9

11. (carcino$ or cancer$ or neoplasm$ or tumour$ or tumor$ or cyst$ or growth).mp.

12. adenocarcinoma$.ab,ti.

13. Precancerous Conditions/ or Carcinoma, Squamous Cell/ or precursor lesion$.mp. or Esophageal Neoplasms/ or Ulcer/ or Peptic Ulcer/

14. or/11‐13

15. 10 and 14

16. mucosal iodine staining.ab,ti.

17. endoscop$.ab,ti.

18. Iodine/ or "Staining and Labeling"/

19. screening.mp. and 10

20. cytolog$.ab,ti.

21. balloon cytolog$.ab,ti.

22. sponge cytolog$.ab,ti.

23. cytodiagnos$.ab,ti.

24. videoendoscop$.ab,ti.

25. or/16‐24

26. screening compliance.ab,ti.

27. mortality rate.ab,ti.

28. cost‐effectiveness.ab,ti.

29. quality of life.ab,ti.

30. survival rate.ab,ti.

31. adverse effect.ab,ti.

32. or/26‐31

33. 15 and 25

34. 15 and 32

35. 33 or 34

36. 7 and 35

Appendix 4. CBM search strategy

Translated from Chinese Biomedical Database and Chinese medical subject headings:

1. exp esophageal Neoplasms. ab, ti.

2. exp screening. ab, ti.

3. exp cytodiagnosis. ab, ti.

4. exp endoscopy. ab, ti.

5. #2 or #3 or #4

6. #1 and #5

7. randomized Controlled Trials. pt.

8. #6 and #7

Appendix 5. VIP and CNKI search strategy

Terms translated from Chinese text words as follows:

1. esophageal cancer. ab, ti.

2. screening. ab, ti.

3. mass screening. ab, ti.

4. cytodiagnosis. ab, ti.

5. endoscopy. ab, ti.

6. #2 or #3 or #4 or #5

7. #1 and #6

8. Randomized Controlled Trials. pt.

9. #6 and #7

Characteristics of studies

Characteristics of excluded studies [ordered by year of study]

Study	Reason for exclusion
Crespi 1979	Excluded due to uncontrolled trial. Characteristics: Subjects: 430 asymptomatic adults. Intervention: Oesophagoscopy. Results: Prevalence of oesophageal cancer or precursor dysplastic changes.
Crespi 1984	Excluded due to uncontrolled trial. Characteristics: Subjects: 430 asymptomatic adults from Iran and 527 individuals from China. Intervention: Oesophagoscopy. Results: Prevalence of oesophageal cancer or precursor dysplastic changes.
Guanrei 1987	Excluded due to uncontrolled trial. Characteristics: Subjects: 600 asymptomatic adults. Intervention: Endoscopy. Results: Prevalence of oesophageal cancer or precursor dysplastic changes.
Lu 1988	Excluded due to comparison between patients and health control, and no comparison between screening and no screening. Characteristics: Subjects: 958 subjects selected from a prospective screening study. 294 with marked atypia, 328 with mild hyperplasia and 336 normal controls. Intervention: cytologic examinations before selecting subjects, and follow‐up. Results: incidence of oesophageal cancer. Duration of study: followed‐up for 11 years.
Shen 1993	Excluded due to uncontrolled trial. Characteristics: Subjects: 12,649 asymptomatic adults. Intervention: Balloon cytology screening. Results: Prevalence of oesophageal cancer or precursor dysplastic changes.
Roth 1997	Excluded due to a uncontrolled study. Characteristics: An uncontrolled trial, screening with balloon and sponge samplers, followed by endoscopy with mucosal iodine staining and biopsy of all unstained lesions. Subjects: 459 participants without a previous diagnosis of oesophageal cancer. No control. Intervention: Screening with balloon and sponge samplers, followed by endoscopy with mucosal iodine staining and biopsy of all unstained lesions. Results: Sensitivity and specificity of balloon and sponge for detecting oesophageal cancer.
Hardwick 1997	Excluded due to uncontrolled surveillance study, and the cases were not participants with a previous diagnosis of oesophageal cancer. Characteristics: Subjects: 162 patients had EAC. No controls. Intervention: Biopsies and brushings taken for histological and cytological assessment. Results: Sensitivity and specificity of cytology for the detection of oesophageal columnar neoplasia.
Saidi 2000	Excluded due to uncontrolled trial. Characteristics: Uncontrolled trial. Subjects: 4192 asymptomatic adults. Intervention: balloon cytology method, followed by endoscopic examination of suspected cases. Results: Prevalence of oesophageal cancer or precursor dysplastic changes.
Sepehr 2000	Excluded due to a self‐controlled, single‐blind, crossover study, and the comparison was not between screening and no screening, but three diagnostic methods. Characteristics: Subjects: 62 healthy volunteers. Control: Self‐control. Intervention: balloon, sponge and sponge‐mesh cytology samplers. Results: Accuracy of the three types of cytology samplers.
Wang 2001	Excluded due to uncontrolled trial. Characteristics: Uncontrolled trial. Subjects: 22,285 asymptomatic adults. Intervention: Endoscopic examination. Results: Incidence of cancer and survival rate. Followed‐up: 24 years.
Saidi 2002	Excluded due to a uncontrolled study. Characteristics: Subjects: 650 randomly selected asymptomatic adults, and finally 504 healthy volunteers participated. No controls. Intervention: Video‐assisted oesophagi‐gastroscopy. Results: Prevalence of oesophageal cancer or precursor dysplastic changes.
Incarbone 2002	Excluded due to a retrospective surveillance study, and the cases were not participants with a previous diagnosis of oesophageal cancer. Characteristics: A retrospective, controlled trial. Subjects: 12 patients with Barrett's adenocarcinoma with endoscopic surveillance for Barrett's metaplasia. 85 patients with Barrett's adenocarcinoma without endoscopic surveillance. Results: Survival. Follow‐up: 48 months.
Wang 2003	Excluded due to uncontrolled trial. Characteristics: Uncontrolled trial. Subjects: 22,285 asymptomatic adults. Intervention: endoscopic examination. Results: Prevalence of oesophageal cancer or precursor dysplastic changes.
Lu 2004	Excluded due to uncontrolled trial. Characteristics: Subjects: 12,048 asymptomatic adults. Intervention: Endoscopy. Results: Prevalence of oesophageal cancer or precursor dysplastic changes.
Wang 2005	Excluded due to a uncontrolled study. Screening with oesophageal balloon cytological examination. Characteristics: Subjects: 2273 symptom‐free subjects. No controls. Intervention: Screening with oesophageal balloon cytological examination. Results: Incidence of oesophageal cancer. Duration: 15 years for 2199 participants.
Guo 2008	Excluded due to a uncontrolled study. Characteristics: An uncontrolled trial, screening with cytologic examination, followed by endoscopy with iodine staining. Subjects: 710 participants without a previous diagnosis of oesophageal cancer. No control. Intervention: Screening with cytologic examination, followed by endoscopy with iodine staining. Results: Sensitivity and specificity of cytologic examination for detecting oesophageal cancer.
Rubenstein 2008	Excluded due to retrospective cohort study, not a randomised controlled trial, and the cases were not participants with a previous diagnosis of oesophageal cancer. Characteristics: A retrospective, controlled cohort study. Subjects: 155 subjects with EAC and GER identified within the United States Department of Veterans Affairs (VA) National Patient Care Datasets (NPCD). Cases: 25 subjects who had an EGD performed between 1 and five years before the diagnosis of EAC. Controls: 130 subjects without a prior EGD. Intervention:A prior endoscopy was performed between 5 years and 1 year before the diagnosis of EAC. Results:Cases diagnosed at earlier stages than controls but did not experience a significant improvement in survival rate. Duration of the study: Retrospective study, and the data was collected from 1993 to 2003.
Pan 2008	Excluded due to a uncontrolled study. Characteristics: An uncontrolled trial, screening with cytologic samplers, followed by endoscopy with iodine staining. Subjects: 740 participants without a previous diagnosis of oesophageal cancer. No control. Intervention: Screening with cytologic samplers, followed by endoscopy with iodine staining. Results: Sensitivity and specificity of mechanical and inflatable balloons for detecting oesophageal cancer.
Gerson 2009	Excluded due to a uncontrolled study. Characteristics: An uncontrolled trial, screening with endoscopy. Subjects: 126 asymptomatic women screened by endoscopy No control. Intervention: Endoscopy. Results: Prevalence of Barrett's oesophagus.
Lecleire 2011	Excluded due to only oesophageal cancer cases. Characteristics: Subjects: 30 patients had a history of cured ESCC. Intervention: Lugol chromo‐endoscopy versus narrow band imaging. Results: Detection rate of cancer and different grade of dysplasia.

Contributions of authors

This review was written by Shujuan Yang and Chunxia Yang.

The database searches were primarily done by the Cochrane UGPD review group, and other searches were done by Shujuan Yang.

Study selection was done by Shujuan Yang, Niu Jin, Shao Ying and Yuchuan Huang.

Assessment of methodological quality was done by Shujuan Yang, Xiaoyan Chen and Youlin Qiao.

Data extraction was done by Shujuan Yang, Niu Jin, Zheng Jianwei, Hao Wu and Liu Xian.

Sources of support

Internal sources

• No sources of support supplied

External sources

• Chinese Cochrane Centre, West China Hospital of Sichuan University, China.

Declarations of interest

None known

New