Screening for reducing morbidity and mortality in malignant melanoma

  • Protocol
  • Intervention

Authors

  • Minna Johansson,

    Corresponding author
    1. FoUU-centrum Fyrbodal, Research Unit, Section for General Practice, Vänersborg, Sweden
    • Minna Johansson, Research Unit, Section for General Practice, FoUU-centrum Fyrbodal, Hedegatan 38, Vänersborg, 45152, Sweden. minna.johansson@vgregion.se.

  • John Brodersen,

    1. University of Copenhagen, The Section of General Practice, Department of Public Health, Faculty of Health Sciences, Center for Health and Society, Copenhagen, Denmark
    2. University of Copenhagen, The Research Unit for General Practice, Department of Public Health, Faculty of Health Sciences, Center for Health and Society, Copenhagen, Denmark
    3. Zealand Region, Primary Healthcare Research Unit, Copenhagen, Denmark
  • Peter C Gøtzsche,

    1. Rigshospitalet, The Nordic Cochrane Centre, Copenhagen, Denmark
  • Karsten Juhl Jørgensen

    1. Rigshospitalet, Nordic Cochrane Centre, Copenhagen, Denmark

Abstract

This is the protocol for a review and there is no abstract. The objectives are as follows:

To assess the effects on morbidity and mortality of screening for malignant melanoma in the general population.

Background

Description of the condition

Malignant melanoma is a lethal cancer, which caused 55,488 deaths globally in 2012 (Globocan 2012a) - and disproportionately more in high-risk countries, such as Australia and New Zealand, where there were 2019 deaths in 2012 (Globocan 2012b). The incidence of malignant melanoma in white populations has risen manyfold over recent decades (Garbe 2009). It has been proposed that this is due to an increase in risk factors, mainly exposure to ultraviolet (UV) radiation from the sun and artificial sources (Waldmann 2012). However, it has also been suggested that this rise in incidence is caused by the overdiagnosis of indolent malignant melanomas, due to increased disease awareness, as increased incidence has not always been followed by increased mortality (Norgaard 2011; Welch 2005).

The most important avoidable risk factor is exposure to UV radiation through sunlight (Gandini 2005). Intermittent sun exposure confers a higher risk, while high continuous exposure seems to be inversely associated with malignant melanoma (Gandini 2005). Exposure in childhood appears to induce a higher risk than exposure later in life (Gruber 2006). An association between artificial sources of UV radiation, such as solariums, and malignant melanoma is likely, but the evidence is limited by inherent biases in the observational studies (Lazovich 2016). Other risk factors include blonde or red hair, green or blue eyes, freckles, an inability to tan, a family history of malignant melanoma, and a large number of naevi and dysplastic naevi (Marks 2000). A randomised controlled trial showed that an effect of sunscreen was to reduce the risk of malignant melanoma, but the evidence is limited by few events in the trial (Green 2011). Educational programmes, including counselling on the avoidance of intense sun exposure and use of sunscreen, have been suggested as a way to reduce mortality of malignant melanoma through primary prevention; a Cochrane Review is currently in progress to evaluate this strategy (Langbecker 2014).

Description of the intervention

Screening for malignant melanoma can be performed through visual self-examination of the skin or visual inspection by a physician, general practitioner, dermatologist, or other health professional. This visual inspection can then be extended to include dermoscopy of identified lesions. Other methods to assist in diagnosing malignant melanomas are evolving and might also be used in screening, for example, teledermatology, mobile phone applications, and spectroscopy-based techniques (Dinnes 2015). The heightened sensitivity that these new methods might confer may increase both the major benefit (a mortality reduction) and the major harm (overdiagnosis) of the intervention. Amongst general practitioners, the sensitivity and specificity of visual inspection has been estimated to be 72% to 84% and 70% to 71%, respectively (Brochez 2001). However, most studies are testing the diagnostic accuracy of inspecting identified pigmented lesions, i.e. not the accuracy of a screening examination of adults with no suspected lesions. Furthermore, sensitivity and specificity do not take overdiagnosis into account; therefore, they are less informative in a screening context than for diagnostic tests. A suite of Cochrane Reviews are currently evaluating the accuracy of tests to assist in diagnosing malignant melanoma (Dinnes 2015).

Screening can be organised as programmes where all potential participants in a community are individually invited to screening or as campaigns where the eligible population is encouraged to participate, for example, through the mass media. Screening for malignant melanoma is not recommended in the US (USPSTF 2009), Canada (CTFPHC 2013), Australia, or New Zealand (ACNMGRWP 2008). Despite this, non-organised screening is increasingly being adopted in many Western countries: in Australia, the one-year frequency of having skin screening ranges from 10% to 50% of the adult population depending on how skin screening is defined (Balanda 1994; Borland 1995; Girgis 1991; Heywood 1994; Janda 2004), and the corresponding rate in the US is 14% to 20% (Federman 1997; Federman 2006; Ford 2004; Sarayia 2004). In Europe, a campaign involving dermatologists in over 30 countries recommends "visiting your dermatologist regularly for a skin check-up" and conducting self-examination every month (EADO 2016).

How the intervention might work

The prognosis of malignant melanoma is closely correlated to the thickness of the lesion at diagnosis, with thinner lesions having a lower risk of metastases and a better prognosis (Breslow 1970). Therefore, early detection through screening could decrease mortality and morbidity from malignant melanoma. Screening might also result in less invasive surgery and less use of adjuvant therapy if the incidence of late-stage disease is reduced, which would be an important benefit in itself (Welch 2011). However, the vast majority of the substantial increase in the incidence of malignant melanoma in countries where skin screening has been adopted represents a rise in the incidence of thin lesions (Norgaard 2011; Welch 2005).

For early-detection strategies to be effective, they must not only detect more cancers at an early stage, but this must also lead to a lower incidence of late-stage disease over time to result in improved prognosis (Keen 2015; Vainio 2002). If a decrease in late-stage disease does not occur, the increase in early-stage disease diagnosis may simply represent detection of lesions that are histologically malignant but would nonetheless never have caused symptoms or death if they had remained undetected, i.e. overdiagnosis (Welch 2011). Overdiagnosed melanomas differ from false positive findings in that they remain defined as malignant, even after histological examination (Welch 2011). It is not possible to know which specific individuals are overdiagnosed. The same is true for the mortality benefit, i.e. it is not possible to know which specific individuals avoid death from malignant melanoma due to screening (Welch 2011).

However, for both outcomes (i.e. overdiagnosis and mortality benefit), it is possible to estimate the rate of events in the two arms of the trials (deaths ascribed to the disease and incident cases) (Welch 2011). Because screening relies on earlier diagnosis, sufficiently long follow-up after the screening intervention has stopped is necessary to quantify overdiagnosis. It is also necessary that the control group has not been offered screening at the end of the trial. When these requirements are fulfilled, it is possible to quantify overdiagnosis by comparing the number of melanomas diagnosed in the screened versus the non-screened group (Welch 2011), because compensation will have been made for any advance of the time of diagnosis.

Overdiagnosis leads to overtreatment, which means that healthy individuals are exposed to unnecessary surgery and possibly adjuvant therapy. Overdiagnosis also constitutes unnecessary labelling of healthy individuals with a cancer diagnosis, which may result in psychological harm (Welch 2011). Rising incidence at the population level complicates the interpretation of incidence and mortality data available in national registries (Norgaard 2011; Welch 2005). In order to understand these trends, which are observed especially in Western societies, the relative contribution of increased sun exposure and possible overdiagnosis with screening needs to be studied.

As it is not possible to distinguish between benign naevi and malignant melanomas with certainty through visual inspection or dermoscopy, a number of unnecessary biopsies or local excisions of benign lesions will result from screening for malignant melanoma, i.e. false positive findings (Harris 2014; Welch 2005). This may lead to psychological stress in addition to the physical consequences of the excisions. On the other hand, malignant lesions may also be missed at screening (false negatives), which may lead to false reassurance and delayed contact with health professionals and delayed diagnosis (Goldenberg 2016). Screening for malignant melanoma may also lead to the discovery of other skin conditions, non-malignant as well as malignant, and result in treatment for these conditions. This may be either beneficial or harmful to patients. Quantifying the effects on morbidity and mortality of diagnosing such conditions is outside the scope of this Review. However, we will quantify any increase in diagnoses and treatments for such conditions, if possible.

Disease-specific mortality in cancer screening trials is an outcome prone to bias from misclassification of the cause of death (Prasad 2016), mainly due to 'sticky-diagnosis' bias and 'slippery-linkage' bias (Black 2002). Sticky-diagnosis bias refers to the risk that the cause of death is falsely attributed to the disease in question, even if this is not the case (Black 2002). Slippery-linkage bias refers to the opposite: that death is attributed to another cause, usually because some time has elapsed since diagnosis. These biases work in opposite directions, and it is not possible to know which one will dominate (Black 2002). Total mortality as an outcome is free of these and other biases and is therefore the most reliable outcome when evaluating the effect of a screening programme. Total mortality has the further advantage that it includes treatment-related mortality, which is of particular concern for screening interventions where overdiagnosis and overtreatment is suspected. It thus captures both the main benefit and the main harms of the intervention. The downside of total mortality as an outcome is that large populations are needed in the trials to reliably detect a difference (Prasad 2016).

Why it is important to do this review

Screening for malignant melanoma is currently practised in many countries, apparently without support from rigorously conducted randomised controlled trials. This is a noteworthy drawback since evidence from randomised controlled trials is considered mandatory before the introduction of screening programmes for cancer (UKNSC 2015; WHO 2008). There is evidence that screening for malignant melanoma may cause overdiagnosis of harmless malignant melanomas and consequently overtreatment (Norgaard 2011; Welch 2005); this is in addition to false positive findings, which in breast cancer screening, for example, are known to cause substantial long-lasting psychological stress (Heleno 2015).

In the light of these concerns, our aim is to assess the evidence for screening the general population to see if it can reduce morbidity and mortality due to malignant melanoma.

Objectives

To assess the effects on morbidity and mortality of screening for malignant melanoma in the general population.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials, including cluster-randomised trials, that compare screening for malignant melanoma with no screening, regardless of screening modality or setting.

Types of participants

We will include studies in any type of population and any age group that is not suspected of having malignant melanoma, including high-risk populations, such as people with light skin living in countries with high sun exposure. We will include studies that do not explicitly exclude a subgroup of participants with previous melanomas.

We will not include studies performed exclusively in participants with previous melanomas, however, as we consider this control monitoring rather than screening.

As participants should not be selected for suspicion of malignant melanoma, we will not include studies of diagnostic tests or of symptomatic individuals who seek medical attention.

Types of interventions

Screening for malignant melanoma using any type of screening modality and in any setting. We will include studies of any screening frequency, including once-only, that are performed by any type of health professional. The control intervention will be no screening.

Types of outcome measures

Primary outcomes
  1. Total mortality.

  2. Overdiagnosis of malignant melanoma (i.e. excess number of malignant melanomas diagnosed in the screening group).

  3. Quality of life/psychosocial consequences (short-term (postintervention up to six months), medium-term (six to 12 months), and long-term (> 1 year) effects).

Secondary outcomes
  1. Mortality specific to malignant melanoma.

  2. False positive rates (skin biopsies/excisions with benign outcome).

  3. False negative rates (malignant melanomas diagnosed between screening rounds and up to one year after the last round).

  4. Use of surgery defined as more than local excision (includes surgery with use of lymphadenectomy).

  5. Use of surgery defined as local excision.

  6. Use of adjuvant therapy.

  7. Incidental findings of other skin conditions (benign or malignant).

  8. Use of health services for any reason.

We will include studies regardless of whether they quantify our prespecified outcomes or not. We will include at least all primary outcomes in our 'Summary of findings' tables.

We will quantify total and disease-specific mortality at five years, 10 years, and the longest follow-up period available.

Search methods for identification of studies

We aim to identify all relevant randomised controlled trials (RCTs) regardless of language or publication status (published, unpublished, in press, or in progress).

Electronic searches

We will search the following databases for relevant trials:

  • the Cochrane Skin Group Specialised Register;

  • the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library;

  • MEDLINE via Ovid (from 1946 onwards);

  • Embase via Ovid (from 1974 onwards); and

  • LILACS (Latin American and Caribbean Health Science Information database, from 1982 onwards).

We have devised a draft search strategy for RCTs for MEDLINE (Ovid), which is displayed in Appendix 1. This will be used as the basis for search strategies for the other databases listed.

Trials registers

We will search the following trials registers using the subsequent terms: melanoma, skin cancer, skin neoplasm, screening, early detection.

Searching other resources

References from included studies

We will check the bibliographies of included and relevant studies and reviews for further references to relevant trials. We will use Web of Science for citation tracking of key articles.

Searching by contacting relevant individuals

We will ask the lead authors of included studies if they are aware of any other published, unpublished, or ongoing studies, or results of studies, that would meet our inclusion criteria.

Adverse effects

We will not perform a separate search for adverse effects of the target intervention. However, we will examine data on adverse effects from the included studies that we identify.

Data collection and analysis

Some parts of the methods section of this protocol uses text that was originally published in other Cochrane Reviews co-authored by KJJ and PCG, predominantly Krogsbøll 2012.

Selection of studies

Two authors (MJ and KJJ) will independently assess the potential relevance of all titles and abstracts that are identified through the searches, and they will assess full-text copies of potentially eligible articles. When necessary, the other authors (JB and PCG) will resolve disagreements through discussion. Two authors (MJ and KJJ) will independently search reference lists, and one author will undertake citation tracking (Web of Science) of included articles.

We will use Covidence to assess the titles and abstracts that are identified in our search of the listed databases, provide reasons for exclusions, and generate a flow chart. Covidence is an online systematic review platform provided by Veritas Health Innovation Ltd, an Australian not-for-profit company.

Data extraction and management

Two authors (MJ and KJJ) will independently extract data from the included trials and enter them into a data extraction form using Covidence. We will evaluate the data extraction form by pilot testing it using a representative sample of the studies. When relevant information is missing from the reports, we will contact the authors.

We will extract the following data from all of the included trials: study design, type of screening test used, total study duration, number of participants allocated to each arm, gender of participants, number lost to follow up for each outcome, baseline comparability, setting, age, country, and date of study start.

We will extract the number of events or rates for total mortality, mortality specific to malignant melanoma, overdiagnosis, false positives, false negatives, surgical interventions defined as more than local excision, surgical interventions defined as local excision, and adjuvant therapy. For psychosocial consequences or quality of life (QoL) outcomes, we will extract the mean; standard deviation or standard error; and name, range, and direction of the scale used.

Assessment of risk of bias in included studies

We will use Cochrane's 'Risk of bias' tool to formally assess the following domains, as described in Higgins 2011: sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other biases, including the degree of contamination of the control group by searching studies for data on the rate of opportunistic screening in the control group.

We will also use the Outcome Reporting Bias in Trials (ORBIT) tool to assess outcome reporting bias (Kirkham 2010). We will assess the randomised groups for baseline comparability. We will use Grading of Recommendations Assessment, Development and Evaluation (GRADE) to assess the level of confidence in individual outcomes.

Measures of treatment effect

For total mortality, mortality specific to malignant melanoma, overdiagnosis, false positives, false negatives, surgical interventions defined as more than local excision, surgical interventions defined as local excision, and adjuvant therapy, we will calculate the risk ratios (RR) and the risk differences (absolute risks). We will calculate standardised mean differences (SMD) for QoL outcomes if different scales have been used and the scales are comparable. If the same scale has been used in all studies, we will calculate the mean difference (MD). For all measures, we will calculate 95% confidence intervals (CI). We will define false positives as the rate of biopsies and local excisions with benign results in the intervention arm in the included trials.

Unit of analysis issues

For cluster-randomised trials, we will use effect estimates and standard errors from analyses that take the clustering into account. When such estimates are not available, we will explore the possible effect of clustering in a sensitivity analysis.

Dealing with missing data

When possible, we will conduct analyses as intention-to-treat (ITT). We will contact authors if the reports do not contain sufficient data for this. If ITT analyses are not possible, we will undertake available case analyses and assess the possible bias resulting from dropouts and losses to follow up in best-case or worst-case analyses for all primary outcomes.

Assessment of heterogeneity

We will assess clinical and methodological differences between the trials before any meta-analyses are done, and we will judge whether we can pool trial results. We will explore statistical heterogeneity using the I² statistic. If we find values of I² statistic above 30%, we will explore causes of heterogeneity in sensitivity analyses and subgroup analyses, and we will not pool results if we encounter unexplained heterogeneity that would render the pooling of results uninformative.

Assessment of reporting biases

Funnel plots are unlikely to be useful in this review to assess publication bias because we do not expect to include enough studies for a valid funnel plot (more than 10 trials is generally recommended). We will use a funnel plot if more than 10 trials are found; otherwise, we will narratively evaluate outcome reporting bias for individual outcomes in any of the included trials and explore this using the ORBIT tool (Kirkham 2010).

Data synthesis

If we judge meta-analyses to be appropriate, we will use random-effects analyses if the populations included, trial design, and frequency and number of screens offered differ substantially.

We will apply the trial sequential analysis model to the dichotomous outcomes (Brok 2008), but not to the continuous outcomes because the trial sequential analysis currently assumes MDs and not SMDs. It is a statistical model, similar to interim analyses in clinical trials, used to quantify the reliability of data in cumulative meta-analyses, adjusting the P values for sparse data and multiplicity. The required information size (the number of participants required to accept or reject the hypothesis of a certain a priori anticipated effect) is calculated using the following five components.

  1. Alpha = 0.05 (type 1 error).

  2. Power = 0.90 (type 2 error 0.10).

  3. Proportion (frequency) of participants experiencing serious adverse events and adverse events (based on observations).

  4. Relative risk reduction (RRR) or increase of 20%.

  5. Diversity (heterogeneity based on our observations).

Preferably, this model should be applied to trials with a low risk of bias only, but we will conduct analyses that also include trials with high risk of bias.

Where results are estimated for individual studies with low numbers of outcomes (less than 10 in total) or where the total sample size is less than 30 participants and a risk ratio is used, we will report the proportion of outcomes in each group together with a P value from a Fisher's Exact test.

Subgroup analysis and investigation of heterogeneity

We will perform subgroup analyses for the following groups:

  • light versus dark skin;

  • young versus old individuals;

  • women versus men;

  • high- versus low-risk countries;

  • screening by specialists (i.e. dermatologists or in screening units by specially trained staff) versus usual care (for example, general practitioners); and

  • high- versus low-intensity screening.

Sensitivity analysis

We will perform sensitivity analysis for studies with high versus low overall risk of bias concerning the randomisation process and blinded outcome assessment. If results differ between studies with high and low risk of bias, we will rely on studies with low risk of bias.

We will conduct a sensitivity analysis of any included studies that were prospectively registered in trials registers.

'Summary of findings' tables

We plan to include a 'Summary of findings' table in our review to summarise at least all primary outcomes (Higgins 2011), and we will assess the level of confidence in the evidence using the five GRADE domains (risk of bias, inconsistency, imprecision, indirectness, and publication bias).

Acknowledgements

The Cochrane Skin Group editorial base wishes to thank Robert Dellavalle, who was the Dermatology Editor for this protocol; Matthew Grainge and Sonia Ratib who were the Statistical Editors; Ching-Chi Chi, who was Methods Editor; the clinical referee, An-Wen Chan; and the consumer referee, Kathie Godfrey.

Appendices

Appendix 1. Draft search strategy MEDLINE (Ovid)

1. exp Melanoma/
2. melanoma$.ti,ab.
3. 1 or 2
4. exp Skin/
5. (skin or epiderm$ or derm$ or cutaneous).ti,ab.
6. 4 or 5
7. 3 and 6
8. malignant melanoma$.ti,ab.
9. exp Skin Neoplasms/
10. skin cancer$.ti,ab.
11. skin neoplas$.ti,ab.
12. or/7-11
13. Mass Screening/
14. screening.ti,ab.
15. "Early Detection of Cancer"/
16. Early detection.ti,ab.
17. or/13-16
18. 12 and 17
19. randomized controlled trial.pt.
20. controlled clinical trial.pt.
21. randomized.ab.
22. placebo.ab.
23. clinical trials as topic.sh.
24. randomly.ab.
25. trial.ti.
26. 19 or 20 or 21 or 22 or 23 or 24 or 25
27. exp animals/ not humans.sh.
28. 26 not 27
29. 18 and 28

[Lines 19-28: Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity- and precision-maximizing version (2008 revision)]

Contributions of authors

MJ was the contact person with the editorial base.
MJ co-ordinated the contributions from the co-authors and wrote the final draft of the protocol.
MJ, KJ, JB, and PCG worked on the methods sections.
MJ, KJ, JB, and PCG drafted the clinical sections of the background and responded to the clinical comments of the referees.
MJ and KJ responded to the methodology and statistics comments of the referees.
MJ and KJ contributed to writing the protocol.
KJ is the guarantor of the final review.

Disclaimer

This project was supported by the National Institute for Health Research, via Cochrane Infrastructure funding to the Cochrane Skin Group. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the Systematic Reviews Programme, NIHR, NHS or the Department of Health.

Declarations of interest

Minna Johansson: nothing to declare.
John Brodersen: nothing to declare.
Peter C Gøtzsche: nothing to declare.
Karsten Juhl Jørgensen: nothing to declare.

Sources of support

Internal sources

  • No sources of support supplied

External sources

  • The National Institute for Health Research (NIHR), UK.

    The NIHR, UK, is the largest single funder of the Cochrane Skin Group.

Ancillary