Screening programmes for the early detection and prevention of oral cancer

Description of the condition

Oral cancer is the sixth most common cancer globally and represents a group of conditions with a range of sites and a varied aetiology. Its annual estimated incidence is approximately 275,000, but unlike many other cancers, its incidence is increasing (Warnakulasuriya 2009). There is a wide geographic variation in the incidence of the disease with two‐thirds of the burden born by low‐income and middle‐income countries from South and South‐East Asia, Latin America and Eastern Europe. However, the incidence continues to rise in the West (IARC 2010) and the age standardised incidence of oral cancer in Western Europe has steadily increased over the past two decades (Boyle 2005). Within the European Union countries, the highest male incidence rates are found in France and Hungary, whilst the lowest rates are found in Greece and Cyprus (IARC 2010). India, Sri Lanka and Pakistan have the highest levels of disease, making it the most common cancer for men in these countries and accounts for up to 30% of all new cases of cancer compared to 3% in the United Kingdom (UK) and 6% in France (Cancer Research UK). The age‐adjusted incidence rate from these countries cancer registries range from 3.4 to 13.8 per 100,000 (Ministry of Health 2005; Warnakulasuriya 2009). The incidence of oral cancer for men in Brazil is second only to France and India with an estimated crude rate of 11 per 100,000. In the UK, the incidence of oral cancer is increasing (Conway 2006; Doobaree 2009) and 6236 cases of oral cancer were diagnosed in 2009 (Cancer Research UK). This represents a doubling of the number of cases seen in 1989 and represents a year on year increase of approximately 2.7% per year (Warnakulasuriya 2009). The incidence of oral cancer is strongly associated with social and economic deprivation (Scully 2009; Conway 2010a), with the highest rates occurring in the most disadvantaged sections of the population. Across Europe, inequalities tend to be observed among men, particularly in the UK and Eastern Europe (Conway 2010b).

Important risk factors in the development of the disease are tobacco, betel quid, alcohol, age, gender and sunlight. More recently, a role for candida and the human papillomavirus (HPV) has been documented (Scully 2009). Epidemiological evidence from US populations indicates a strong association between HPV and oropharyngeal cancers (Cleveland 2011). Incidence of HPV globally is unknown but thought to be rising, but estimates from the United States of America (USA) show a substantial increase of HPV‐positive oropharyngeal cancers, rising by 225% in the period between 1984 and 2004 (Sanders 2011). Increased consumption of alcohol has been implicated in the increasing incidence of the disease in the UK (Hindle 2000) at a time when tobacco use is falling (Ogden 2005), although the precise mechanism remains unclear (Ogden 1998). Heavy drinkers and smokers have 38 times the risk of developing oral cancer compared to abstainers (Blot 1988). This is thought to be due to acetaldehyde, the first metabolite of alcohol, which is classified as a Group 1 carcinogen and is also present in tobacco (Salaspuro 2011). Historically, the risk of developing oral cancer increased with age, however, the age group with the highest incidence (26.8%) in the USA between 2003 and 2007 was between 55 and 64 years of age (SEER 2010). In contrast, many patients from high‐incidence countries are below the age of 40 years of age (Warnakulasuriya 2009).

Of equal concern to the increasing incidence, is the lack of any change in the age‐standardised mortality rates, despite advances in surgical and management techniques. This is unlike the falling rates for cancer of the breast and colon (Cancer Research UK). The five‐year survival rates for oral cancer for most countries is approximately 50% (Warnakulasuriya 2009). These have been estimated at 3 to 4 per 100,000 men and 1.5 to 2.0 per 100,000 for women respectively (Warnakulasuriya 2009). Mortality rates from oral cancer have also increased in certain European countries (La Vecchia 2004). The most important determinant factor in cancer survival is diagnostic delay (Onizawa 2003; McLeod 2005), as over 60% of patients present with stage III and IV disease (Lingen 2008), meaning that their management is complex and multidisciplinary. The stage at diagnosis significantly affects five‐year survival, with survival rates approaching 80% for stage I disease, whilst dropping significantly for stage IV disease (Rusthoven 2010). In addition, the morbidity associated with surgery is high, the rate of second primary tumours is greater than any other type of cancer (3% to 7% per annum) (Day 1992) and is more often the cause of death (Lippman 1989).

Description of the intervention

Prevention strategies are important to meet the World Health Organization's (WHO) resolution to incorporate oral cancer into national cancer control programs (Petersen 2009). Although it is important to continue to clarify the public health message and promote primary prevention, determining the feasibility of a national screening programme is an important step in the prevention of the disease. The National Screening Committee define screening as "a process of identifying apparently healthy people who may be at increased risk of a disease or condition" (NSC 2010). Screening can be undertaken across the whole population, opportunistically, when individuals are attending for some other purpose, or selectively, where high‐risk groups are targeted. Programmes for major cancers, such as breast, cervical and bowel cancer have effectively improved the mortality rates and helped to decrease the incidence of these cancers (Gøtzsche 2006; Hewitson 2007). However, it is important to consider both the harms and benefits of any screening programmes. For example, screening for breast cancer has recently been associated with over‐diagnosis and unnecessary treatment causing further physical and psychological harms (Gøtzsche 2013) and so any future programme should always balance these considerations.

How the intervention might work

Screening is predicated on the idea that malignancy is preceded by clinically evident lesions, which if identified early and removed, can either prevent their malignant transformation or reduce their staging. The majority of oral carcinomas are preceded by visible lesions, known as potentially malignant disorders (PMDs) (Warnakulasuriya 2007; van der Waal 2009) that exhibit oral epithelial dysplasia (Scully 2009). A visual screen is not surgically invasive, is painless and has been found to be socially acceptable. Additional Table 1 highlights the different types of PMDs that were considered by the WHO's Working Party on Oral Cancer and Precancer to be important (Warnakulasuriya 2007). The most common form of PMD is leukoplakia (Napier 2008), which has an estimated global prevalence of 2.6% (95% confidence interval (CI) 1.72% to 2.74%) (Petti 2003). However, the extent and rate of progression of dysplasia in leukoplakia is not uniform and can vary from site to site and within the same lesion (Napier 2008).

The overall malignant transformation rate for oral leukoplakia is up to 5% (Scully 2009; van der Waal 2009), but the lack of uniformity in the extent and the rate of dysplastic change in PMDs means that predicting malignant transformation is problematic. However, there remains a consensus in the literature that the majority of cancers are preceded by a detectable pre‐clinical phase (Napier 2008).

Although there have been no randomised controlled trials (RCTs) in any developed or low‐prevalence populations (Brocklehurst 2010a), Speight et al demonstrated using a simulated model that an oral examination of high‐risk individuals may be a cost‐effective screening strategy (Speight 2006). A recent diagnostic test accuracy review looking at the accuracy of conventional oral examination as a screening test in primary settings found sensitivity estimates to range from 0.50 (95% CI 0.07 to 0.93) to 0.99 (95% CI 0.97 to 1.00) with specificity estimates 0.98 (95% CI 0.92 to 1.00) to 0.99 (95% CI 0.99 to 0.99) (Walsh 2013). Positive predictive values ranged from 0.31 to 0.86; negative predictive values ranged from 0.96 to 0.99 (Walsh 2013). Other adjunctive and diagnostic aids can be grouped into visual staining (toluidine blue), oral cytology using brush biopsy and a number of light‐based techniques (e.g. ViziLite (Zila Pharmaceuticals, AZ, USA) and VELscope (LED Dental Inc, BC, Canada)) (Brocklehurst 2010a).

Why it is important to do this review

The Cochrane Oral Health Group undertook an extensive prioritisation exercise in 2014 to identify a core portfolio of titles that were the most clinically important ones to maintain on the Cochrane Library (Worthington 2015). Consequently, this review was identified as a priority title by both the oral medicine and public health expert panels (Cochrane OHG priority review portfolio).

The RCT provides the strongest level of evidence on which to base clinical decisions (Clarkson 2003) and so represents a level of rigor that is appropriate for assessing the effectiveness of any intervention or programme. As with other cancers, screening for oral cancer and PMDs has potential advantages and disadvantages (Speight 1992). Screening and treatment may offer the opportunity to reduce the incidence of invasive lesions and also could help in decreasing the mortality rates associated with oral cancer. Speight et al demonstrated that targeting high risk groups could result in a pronounced increase in the Quality Adjusted Life Years saved and any associated stage shifts could produce significant cost savings (Speight 2006). However, screening also has the potential to generate false positives and false negatives (Wilson 1968). Earlier versions of this Cochrane review concluded that there was insufficient evidence to support or refute the use of screening for oral cancer in the general population (Kujan 2003; Kujan 2006; Brocklehurst 2010c). The purpose of this latest update was to determine whether the evidence base had changed.

Criteria for considering studies for this review

Search methods for identification of studies

Data collection and analysis

Description of studies

A total of 3239 citations were identified through the MEDLINE searches. The full text of 30 articles were retrieved. Following further screening, 26 of these were excluded with reasons (Characteristics of excluded studies). One of the excluded studies did undertake a community‐based randomised controlled trial to examine the efficacy of toluidine blue in Taiwan (Su 2010). However, the primary aim was to determine whether toluidine blue enhanced the detection rate for potentially malignant disorders (PMD), not mortality rate. Although they did link the examined cohort with the National Cancer Registry to determine five‐year follow‐up, the power calculation was based on the former not the latter. As a result, the study was excluded from this review but included in an accompanying diagnostic test accuracy review (Walsh 2013). Only one study (four reports) met the inclusion criteria (Sankaranarayanan 2000). The principal investigator of the included trial (Dr Sankaranarayanan) was also contacted and no other relevant trials were identified (Figure 1).

The included study (Trivandrum Oral Cancer Screening Study; Sankaranarayanan 2000) was designed to have an 80% power at the 5% significance level to detect a 35% reduction in the cumulative mortality rate of oral cancer in 12 years of enrolment between the intervention and the control groups. The study commenced in October 1995 and three rounds of screening at three‐year intervals were planned for the study. The first round was completed in May 1998 and the second was completed in June 2002. The third round was completed in October 2004. A final round of screening was completed in 2009.

All participants (n = 191,873) were apparently healthy residents aged 35 years or older and lived in 13 rural clusters around Trivandrum City, Kerala, India, The mean number of eligible participants in each cluster was 14,759. These clusters were allocated into an intervention arm (n = 7) and a control arm (n = 6) by a blocked randomisation process. Those residents who were bedridden, suffering from open tuberculosis or other debilitating diseases were excluded alongside those participants who had already been diagnosed with oral cancer prior to entry into the study.

In the intervention arm, non‐medical university graduates were initially trained and were provided with two simple manuals on oral visual examination with colour photographs and descriptions of various oral lesions. Eligible participants were interviewed and information relating to demographic, social and personal habits including the use of paan, tobacco, alcohol and dietary supplements was recorded. Tobacco and alcohol cessation advice was provided as appropriate. Oral visual inspections were performed in daylight with the help of a flashlight. All the intra‐oral sites were carefully examined and palpated and the neck was also palpated to detect enlarged lymph nodes. The findings were recorded as normal, non‐referable lesions and referable lesions.

Participants who had a positive screen were referred for examination by a dentist or physician for confirmation. It is unclear whether these clinicians were also trained in the recognition of oral cancer or PMDs. Oral biopsies were performed in those with clinically confirmed homogeneous leukoplakias, non‐homogeneous leukoplakias, oral submucous fibrosis and oral cancers. Surgical excision was undertaken for leukoplakia wherever possible. All PMDs were reviewed regularly.

In the control arm, participants were visited by a "control health worker" who recorded the same sociodemographic information and measured height, weight, blood pressure and respiratory peak flow measurements. The health workers in the control arm were not trained to undertake a visual oral inspection.

Oral cancer mortality was reported as the main outcome measure.

Of the 96,517 eligible subjects in the intervention arm, 25,144 (26.1%) had one, 22,382 (23.2%) had two, 22,008 (22.8%) had three and 19,288 (20.0%) had four cycles of screening. 49,179 (51.0%) individuals were screened in the first round and 55,993 (58.0%), 64,898 (67.2%) and 43,014 (44.6%) were screened in the second, third and fourth cycles respectively. The participation rate (at least one screen) for this group was 88,822 (92%); males (86%) and females (94%). Of the 95,356 eligible subjects in the control group, 43,992 (46.1%) were screened in the fourth cycle.

Demographic details were not provided for the final (fourth) cycle of screening, but were provided for the first three cycles (Additional Table 2). Across the period of the study (all four cycles), 6.3% (n = 5586) of subjects screened as part of the intervention group had a referable lesion and 59% (n = 3298) of these screen positive subjects complied with referral (Additional Table 3). The control group consisted of 95,356 persons. 46% (n = 43,992) of the control group were screened in the final (fourth) cycle and 2.6% (n = 1163) of these were found to have a referable lesion with 16.3% (n = 189) complying with referral.

Risk of bias in included studies

Effects of interventions

The included study reported data on oral cancer incidence, disease‐specific mortality, and stage at diagnosis after 15‐years follow‐up. Data on quality of life and all cause mortality were not reported.

Summary of main results

The study reported a sensitivity of the visual examination in detecting oral cancer was 67.4%. However, the data for users of tobacco, alcohol or both demonstrated a reduction in mortality rates of 24% after four cycles (15 years), which was statistically significant. In addition, there was a statistically significant difference in the number of stage III cancers between the intervention and control arms, suggesting that the screening programme was identifying cancers at an early stage. When these results are combined with the significant stage shift and survival rate in the intervention group, it would appear that visual examination could be effective at reducing mortality rates for oral cancer when used within a targeted screening programme. However, the included study had a high risk of bias.

Overall completeness and applicability of evidence

The purpose of health care is to improve both the quantity and quality of life (Kaplan 2005). The evidence from the Kerala trial (Sankaranarayanan 2000) is that visual screening can reduce the mortality rate in users of tobacco, alcohol or both and can produce a stage shift. Given that late stage disease is recognised as a major contributory factor for cancer survival (Onizawa 2003; McLeod 2005), it would appear that the screening of high‐risk individuals could be warranted. However, the evidence from this study stems from a population with a high incidence of oral cavity cancer, and its applicability to other countries with lower incidence rates is unknown. In addition, the efficacy of the early management of potentially malignant disorders (PMDs) is a controversial area (Holmstrup 2007; Holmstrup 2009). Holmstrup argues that even if early lesions are surgically removed, the risk of malignant change can remain as a result of "field change" i.e. the lesion represents only a small area of a wider field of damaged mucosa (Holmstrup 2007; Holmstrup 2009).

The trial used non‐medical university graduates, trained specifically to perform visual inspection of the oral mucosa, with the help of a flashlight. The screening was undertaken in individuals own home, with the health workers going 'door‐to‐door'. The ability to translate this screening model to other settings is unclear. However, the Kerala study does demonstrate the potential of allied health professionals to screen for oral cancer and PMDs. This is becoming increasingly relevant as more regulators allow patients to directly access allied providers of dental care (oral health practitioners), in addition to the dentist.

The cost‐effectiveness of the Kerala study was reported after three cycles (nine years) (Subramanian 2009), demonstrating that 1437.64 life‐years could be saved per 100,000 high‐risk individuals, with an incremental cost per life‐year saved of USD 156. According to the authors, this fulfils the target set by the WHO Commission on Macroeconomics and Health, who define an intervention as being cost‐effective when its cost‐effectiveness ratio is less than a country's gross domestic product per capita. However, these results also need to be read in context of the relative prevalence of the condition and account for the potential problem of compliance; out of the 5586 participants that were screened positive, only 59% (n = 3298) complied with referral.

The consideration of both the benefits and harms of screening is an essential component of any programme (Wilson 1968) and is fundamental to the satisfactory introduction of any technology into daily practice (Duffy 2001). The sensitivity reported in the Kerala study was relatively low (67%). In addition, false positives can have unintended psychological consequences, for example, increasing anxiety and exposing the patient to unnecessary further investigations. However, these could be reduced by careful patient management and by educating screened patients about the positive benefits of screening (Speight 1992). Recent studies by Brocklehurst, highlighted the need to train general dental practitioners to discuss positive findings (Brocklehurst 2010b) and the need for standardised criteria to avoid both under and over‐referral in clinical practice (Brocklehurst 2010).

Quality of the evidence

The Kerala study had a number of methodological weaknesses that may have introduced bias. These included a lack of detail about the process of sequence generation to ensure random assignment, no analysis of the impact of clustering on the results and no detail about allocation concealment. In addition, there was no blinding of the outcome assessment and withdrawals and drop‐outs were not described. More importantly, only 59% of individuals with screen‐positive lesions in the intervention arm complied with referral and it was unclear whether the clinicians who saw these patients followed any standardised criteria. In addition, only 26.4%, 26.0%, and 21.4% of subjects had a biopsy in the second, third cycle or fourth cycle respectively, with no detail being provided from the first cycle. As it is not possible to diagnose early malignancy by visual appearance alone, this may have led to substantial under‐reporting of oral cancer and the lack of a histological diagnosis makes it difficult to accurately assess the correct diagnosis and true prevalence of these disorders. It has been argued that the small number of randomised clusters could also produce statistical heterogeneity and the close geographical proximity of the clusters may have led to contamination (Kujan 2005).

Potential biases in the review process

We conducted a broad search of several databases and placed no restrictions on the language of publication when searching the electronic databases or reviewing reference lists of included studies. All data extraction and risk of bias assessment was conducted in duplicate.

Agreements and disagreements with other studies or reviews

A meta‐analysis of visual screening found a weighted and pooled sensitivity of 84.8% (95% confidence interval (CI) 73.0 to 91.9) and specificity of 96.5% (95% CI 93.0 to 98.2) (Downer 2004). However, there was considerable heterogeneity in the studies pooled due to differences in the size of the target populations and it is arguable that a meta‐analysis was inappropriate. A more recent review has evaluated the diagnostic accuracy of non‐specialist conventional oral examination, vital rinsing, light‐based detection, biomarkers and mouth self examination for the identification of individuals with suspected oral squamous cell carcinoma or PMD. Sensitivity values for studies for all tests were varied and relatively imprecise with sensitivity values ranging from 0.59 (0.39 to 0.78) to 0.99 (95% CI 0.97 to 1.00) for the conventional oral examination (Walsh 2013).

These values of sensitivity and specificity for visual examination have not been surpassed by any other type of method, such as self examination, vital staining (toluidine blue), oral cytology or light‐based techniques (Lingen 2008; Patton 2008; Walsh 2013). The review by Walsh et al showed sensitivity values ranging from 0.18 (95% CI 0.13 to 0.24) to 0.33 (95% CI 0.10 to 0.65) for mouth self examination and a sensitivity value of 0.20 (95% 0.01 to 0.72) for the addition of toluidine blue (Walsh 2013). In Patton's systematic review, 23 studies met the inclusion criteria, yet there remained insufficient evidence to support or refute the use of adjunctive techniques to the visual examination (Patton 2008). In another review, Lingen found that the majority of the published studies had employed these techniques on patients who had already received a diagnosis and that they did not improve upon the sensitivity or specificity of the visual examination (Lingen 2008).

Characteristics of included studies [ordered by study ID]

Characteristics of excluded studies [ordered by study ID]

Internal sources

• The University of Manchester, UK

• AlBaath University, Syrian Arab Republic

• Manchester Academic Health Sciences Centre (MAHSC), UK

  The Cochrane Oral Health Group is supported by MAHSC and the NIHR Manchester Biomedical Research Centre

External sources

• Cochrane Oral Health Group Global Alliance, UK

  All reviews in the Cochrane Oral Health Group are supported by Global Alliance member organisations (British Association of Oral Surgeons, UK; British Orthodontic Society, UK; British Society of Paediatric Dentistry, UK; British Society of Periodontology, UK; Canadian Dental Hygienists Association, Canada; National Center for Dental Hygiene Research & Practice, USA; Mayo Clinic, USA; New York University College of Dentistry, USA; and Royal College of Surgeons of Edinburgh, UK) providing funding for the editorial process (http://ohg.cochrane.org/)

• National Institute for Health Research (NIHR), UK

  CRG funding acknowledgement:
The NIHR is the largest single funder of the Cochrane Oral Health Group
Disclaimer:
The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the NIHR, NHS or the Department of Health