This article has Open Peer Review reports available.
For which cancers might patients benefit most from expedited symptomatic diagnosis? Construction of a ranking order by a modified Delphi technique
© Hamilton et al. 2015
Received: 5 November 2014
Accepted: 27 October 2015
Published: 30 October 2015
This study aimed to answer the question ‘for which cancers, in a symptomatic patient, does expediting the diagnosis provide an improvement in mortality and/or morbidity?’
An initial ranking was constructed from previous work identifying ‘avoidable deaths’ for 21 common cancers in the UK. In a two-round modified Delphi exercise, 22 experts, all experienced across multiple cancers, used an evidence pack summarising recent relevant publications and their own experience to adjust this ranking. Participants also answered on a Likert scale whether they anticipated mortality or morbidity benefits for each cancer from expedited diagnosis.
Substantial changes in ranking occurred in the Delphi exercise. Finally, expedited diagnosis was judged to provide the greatest mortality benefit in breast cancer, uterine cancer and melanoma, and least in brain and pancreatic cancers. Three cancers, prostate, brain and pancreas, attracted a median answer of ‘disagree’ to whether they expected mortality benefits from expedited diagnosis of symptomatic cancer.
Our results can guide future research, with emphasis given to studying interventions to improve symptomatic diagnosis of those cancers ranked highly. In contrast, research efforts for cancers with the lowest rankings could be re-directed towards alternative avenues more likely to yield benefit, such as screening or treatment.
It is widely believed that the sooner a cancer is diagnosed, the better the outcomes for that patient will be . There is strong evidence from screening trials in some cancers that this is true for the asymptomatic patient. However, once the patient has developed symptoms, the evidence for mortality benefit from expediting the diagnosis is much weaker, and relies on observational data [2–4]. Even so, it is still considered axiomatic that avoiding delay in symptomatic cancer diagnosis is beneficial, in reducing mortality and morbidity, providing symptom relief and possible improvements in psychological well-being. This view underpins the many interventions in recent years aimed at improving survival rates from cancer, especially in those countries with a relatively poor cancer performance, such as the United Kingdom (UK) and Denmark .
Patients certainly value early diagnosis of cancer . However, not all cancers behave in the same way and there are many different types. Survival differs greatly across the common cancers; so does ease of diagnosis –shown by the number of primary care consultations pre-diagnosis , times to diagnosis , and the proportion having an emergency presentation . In some poor-prognosis cancers, there is an element of nihilism with cure considered unlikely whatever the time to diagnosis (though relief of symptoms and prolongation of life remain important) . Such nihilism may pervade the research agenda; for instance, lung cancer has consistently attracted much lower research funding than the other three common cancer sites .
If expediting symptomatic diagnosis is of greater (or lesser) value across different cancer sites, then it would help the research community to know which cancer sites have most (or least) to gain. Those with most to gain might preferentially receive research funding targeted on early cancer diagnosis. In contrast, research effort in cancers with less to gain would focus more on prevention, development of effective screening, or treatment.
This study aimed to answer the question ‘for which cancers, in a symptomatic patient, does expediting the diagnosis provide an improvement in mortality and/or morbidity?’ In the absence of trial evidence, alternative methods had to be employed. Thus, we used a Delphi exercise, combining the available evidence with the pooled responses of experts – an approach which has not been tried before to answer this question.
Phase 1 – Construction of the initial ranking
Derivation of the figures for Delphi Round 1
The second stage adjusted Column A1 to accommodate cancers with high survival, in the belief that the Abdel-Rahman paper would fail to capture fully any survival benefit from expedited diagnosis in this group. For example, testicular cancer has very high survival, and international survival differences are small. However, this could be because the disease is extremely chemo-sensitive (which it is) or because it is a cancer which is diagnosed quickly, or occurs in younger patients – or more than one of these. Therefore, high survival may mask any advantages from rapid diagnosis. For each cancer, UK 5-year survival was identified from the Office for National Statistics publication for England, relating to 2005–9 . For each cancer site, the percentage surviving 5 years was added to the figures in Column A1, giving a half weighting to the survival figures (shown in column Shalf). This process led to the removal of oral cavity cancers from Column A2, as survival statistics separated oral cancers into subgroups; we could identify no composite figure.
Phase 2: the Delphi survey – round 1
‘For this particular cancer, in a patient who has symptoms from the cancer, does expediting the diagnosis provide an improvement in mortality?’
The resource pack was derived from a systematic review first carried out in 2010 , and updated in December 2013/January 2014, though published after this study completed . The aim of the review was to determine the impact of different patient, primary care and secondary care intervals (and combinations of these) on stage at diagnosis and/or survival. 121 papers were included. Each was assigned a category based on the study design (see Additional file 1) though most were observational studies of individual patients (category 3). A sample from the resource pack is shown in Additional file 2 (full pack available from authors).
Delphi participants were allocated a ‘purse’ of 40 points to be used to adjust the position of cancers in Column A2 . Positive or negative points could be used - both counting equally to the total spend. All 40 points had to be used, with +/−10 being the maximum for a single cancer site. Delphi participants were asked to use their personal experience, plus the evidence in the resource pack, to make these adjustments. We did not request details of how much participants used the resource pack, and how much they relied on personal experience (though most round 1 replies from participants included comments upon the lengthy time involved in making their judgements, suggesting the resource pack was used considerably).
Weighting of phase 2, round 1 of the Delphi exercise
We weighted the responses to ensure that Round 1 of the Delphi exercise achieved an overall influence of 33 % for the whole study. Column A2 – the initial ranking sent to Delphi participants – totalled 436 points, with a standard deviation of 10.5. The weighting within Phase 2 aimed to give 2/3rd of 436 points to Round 1, i.e. 291 points, to achieve the overall influence of 33 %. The process for calculating the weighting entailed: a) calculation of the mean change in points allocated by Delphi participants to each cancer site (this could be a positive or negative number); b) negative mean changes were temporarily made positive in order to avoid any ‘cancelling out’ effect, and the total of the 21 mean values was calculated to give the overall magnitude of change; c) 291 was divided by this total magnitude of change, to give a conversion factor (8.61); d) originally negative mean changes were restored to negative (reversing stage b); e) individual mean changes for each cancer were multiplied by the conversion factor, and then added to the original points in Column A2 to give Column B.
To estimate how much influence the Delphi participants actually had, the adjusted changes can be expressed as multiples of the standard deviation of the values in Column A2. The greatest adjusted change would have been if every participant had given one particular site the maximum allowed 10 points. Once multiplied by the conversion factor of 8.61, the greatest adjusted change would be 86 points, or 8.2 standard deviations. Thus any cancer could have been promoted from the bottom of the ranking to the top, or vice versa.
Likert rating of mortality and morbidity statements
Delphi participants also rated statements relating to mortality and morbidity benefits for each cancer site. These were: ‘Expedited cancer diagnosis brings mortality benefits in this cancer site’ and ‘Expedited cancer diagnosis brings morbidity benefits such as improved treatment options and psychological benefit in this cancer site.’ Treatment options were described further to include pain or other symptom relief. Responses were requested on a 5-point Likert scale: Strongly disagree; Disagree; Undecided; Agree; Strongly Agree. For analysis, we did not assume the responses were equally spaced; instead we report the median response (modal responses differed in only 7 of the 42 answers, always to an adjacent response).
Phase 2: the Delphi survey – round 2
All 22 responders to Round 1 of Phase 2 were invited to participate in Round 2. The pre hoc principle was that Round 2 should have a total influence on the whole process of 17 %. Each participant was allocated a ‘purse’ of seven points which could be used positively or negatively to one or more cancer sites. Unlike Round 1, participants were not obliged to ‘spend’ all – or any – of their points if they were happy with Column B. Column C was constructed from Column B, plus the mean change in allocated points in Round 2.
To test whether the survival adjustment in the second stage of Phase 1 was reversed in the Delphi process, we calculated the correlation between the change from Columns A1 to A2, and the change from A2 to Column C (the entire Delphi process). Analyses used Stata, version 13 .
No ethical committee review was required for this study, as is the norm with Delphi studies and with research involving staff; participants were deemed to have consented by returning the questionnaire.
Median Likert responses of statements that expedited symptomatic cancer diagnosis brings mortality or morbidity benefits
Expedited symptomatic cancer diagnosis brings mortality benefits
Expedited symptomatic cancer diagnosis brings morbidity benefits
Bladder, breast, cervix, colon, larynx, lung, melanoma, ovary, rectum, testis, uterus
Bladder, brain, colon, kidney, larynx, leukaemia, lung, myeloma, oesophagus, ovary, stomach
Hodgkin’s disease, kidney, leukaemia, myeloma, non-Hodgkin’s lymphoma, oesophagus, ovary, stomach
Breast, cervix, Hodgkin’s disease, melanoma, non-Hodgkin’s lymphoma, pancreas, prostate, rectum, testis, uterus
Brain, pancreas, prostate
Summary of the main findings
This is the first study to try to rank cancers by potential mortality benefit from expedited symptomatic diagnosis. The final ranking, although not definitive, suggests that cancers with the greatest potential benefit are breast, uterus and melanoma; in contrast, pancreas and brain were considered to have the least potential benefit. The Likert responses for whether mortality gains were to be expected from expediting symptomatic diagnosis were concordant with the final rankings. Perceived morbidity benefits were largely unrelated to the mortality ranking for cancer sites.
Strengths and limitations
This study had two distinct phases: construction of the initial ranking and the Delphi process. Both have methodological limitations. The initial Column A1 was predicated upon the view that European cancer survival differences reflected differences in timeliness of diagnosis in part at least. We used the best information available, and considered it preferable to provide an initial ranking rather than offering Delphi participants a starting point of all cancers being equal – while accepting the possibility of an ‘anchoring’ effect . This effect occurs when people are given a specific value for a quantity, before measuring (or in our case, adjusting) the quantity itself. The Abdel-Rahman paper uses data which is now close to twenty years old; although improvements in diagnosis and treatment may have changed the ranking somewhat, the purpose of the initial ranking was to provide a reasonable starting point based – in part at least – on diagnostic differences. More recent data – had it been available – would have reflected the improvements in UK cancer diagnostic pathways in recent years, making international variation less helpful in showing up any effect from improved diagnostics.
The adjustment for survival was also debatable; the rationale was that patients with cancers with high cure rates could still benefit from expedited diagnosis. One internal finding suggests that this process was valuable: the positive correlation between the change in scores introduced by the survival adjustment and the change in scores from the Delphi exercise. In short, the Delphi exercise moved in the same direction as the survival adjustment, rather than reversing it. Furthermore, there was ample latitude in the Delphi exercise to change any misplaced cancer. A cancer ranked last could have become first, and vice versa, though no cancer changed ranking as dramatically as this. Other choices in Phase 1, including different weightings, would have generated different starting points for the Delphi participants; they might not have led to important differences in the final outputs. One unfortunate effect of the survival adjustment was that oral cancer was lost to the Delphi process, as we could not find appropriate survival figures. The initial position of oral cancer was mid-ranking, between rectal cancer and non-Hodgkin lymphoma. We cannot know what its final position would have been had it entered the Delphi exercise.
The Delphi exercise had two rounds. A feature of the Delphi process is reiteration, with structured feedback and subsequent rounds helping to develop consensus. The number of rounds is a balance between maintaining recruitment and allowing change. We saw little alteration in Round 2 – with nearly a third of the panel suggesting no change at all was needed – and hence the value of a third round would probably have been limited. Our number of Delphi participants was within the usual range ; of 189 invited, 22 took part, all of whom completed both rounds. We do not know why most of the non-responders declined. The evidence pack was large, and many participants indicated that the exercise required both considerable time and intellectual engagement. We cannot know if a different panel would have yielded different results. Deliberately, we excluded patients and surgical specialists from the Delphi process: both groups largely have expertise in a single cancer site (or grouped sites, such as urological cancers). We believed that the Delphi participants should be able to comment across a wide range of cancers to allow them to adjust the rankings appropriately. A future exercise could extend to a much wider community, including patients, the public and specialists.
The findings from this study would change if major improvements in treatment for a particular cancer were available. For example, were a new drug for pancreatic cancer to be developed, then expedited symptomatic diagnosis of pancreatic cancer would presumably increase in importance. Therefore, our rankings are of value only for the present time. They also only pertain to the 21 cancers we included. Omission of other cancers from the ranking does not mean no benefit is to be expected, just that any such benefit is unreported.
This is the first exercise of this nature that we know of. Thus we cannot directly compare it with other published studies. The research question offered to Delphi participants made no reference to division of the pre-diagnostic period into patient intervals (before entry into healthcare) and healthcare provider/system intervals . Thus the results may pertain to both, and hence have possible relevance to either awareness campaigns or the design of diagnostic services.
The findings on the potential morbidity benefit were unexpected. Although no cancer sites attracted a median response of ‘disagree’ or ‘strongly disagree,’ ten had a median response of ‘undecided’. This was despite the emphasis given to the possibility of palliation and symptom relief. Three of these cancer sites – breast, melanoma and prostate – have a sizeable literature relating to over-diagnosis, though this mostly relates to screen-detected cancer. It is plausible that, for these sites, Delphi participants believed that the morbidity benefits were outweighed by harms, though it is hard to see how this could extend to the other seven cancers.
Three cancers – breast, uterus and melanoma – were ranked as being most likely to result in improved mortality outcomes through expedited symptomatic diagnosis, and there was a moderate gap between these cancer sites and the next group. In the UK, diagnostic services for breast cancer are highly streamlined, such that in the English National Health Service women with almost any breast complaint (other than pain) are assessed rapidly for the possibility of cancer. This facility – which is the most comprehensive of all the various cancer diagnostic services – is consonant with our findings. We cannot know if the presence of such a scheme influenced Delphi participants, who could reasonably have inferred from the existence of such a service that it was particularly warranted. The generally accepted benefits of breast cancer screening may have encouraged Delphi participants to consider breast to be particularly worthy of expedited diagnosis , even though the screening population is asymptomatic, and were not included in this study.
Uterine cancer has a similarity to breast cancer, in that a single presentation, post-menopausal bleeding, dominates . Services for investigation of post-menopausal bleeding are also well-structured, and timely investigation is the norm. In contrast to breast cancer or melanoma, uterine cancer has not been considered for awareness campaigns encouraging women with post-menopausal bleeding to consult their doctors. Our results suggest this should be considered.
Melanoma also has a characteristic symptom, a pigmented lesion. Times to diagnosis are generally short, emergency admissions are rare, and few patients report seeing their doctor three or more times before diagnosis [7–9]. Some concerns about ‘over-diagnosis’ have been raised for melanoma , based on a stable number of deaths in the face of rising numbers of new diagnoses. An alternative interpretation is that good services for diagnosis of melanoma have led to improved survival, despite an increased incidence.
Sixteen cancers are ranked in the middle of the ranking, though for only one of them, prostate, did the Delphi respondents give a median Likert answer of ‘disagree’ to the statement that expedited symptomatic cancer diagnosis brought mortality benefits. Five cancers within this large grouping had large changes in the Delphi exercise: bladder and lung were promoted, and ovary, prostate and non-Hodgkin’s lymphoma fell. The promotion of bladder may represent the clear survival advantage of early bladder cancer diagnosis when compared with advanced disease. The promotion of lung cancer refutes some of the nihilism attaching to this subject; in the UK at the time of the Delphi exercise there was considerable public publicity of lung cancer symptoms, as part of the Be Clear On Cancer campaign. Furthermore, lung cancer resection rates are rising, and newer treatments, tailored to specific cancer genotypes, have entered mainstream practice. This circumstantial evidence may have encouraged Delphi participants to believe that expedited symptomatic diagnosis has benefits – for some lung cancer patients. Ovary is similar to bladder cancer in its sharp survival difference between early and advanced disease, though, in this cancer, advanced disease is the norm. Diagnostic testing has become much easier, with the widespread use of Ca125 . It is hard to explain why non-Hodgkin’s lymphoma fell. Long-term results from a trial of surgery in prostate cancer have yielded small benefits, though with additional morbidity . This is not the same as demonstrating benefits from expedited diagnosis. Furthermore, prostate screening trials have disappointed, reporting no, or very small benefits. This leaves a confusing picture, of small benefits at best – commensurate with its final place in the ranking.
Two cancers were deemed to be associated with the smallest gains from expedited diagnosis throughout the whole exercise: brain and pancreas. Both have very poor prognoses, and this may have been uppermost in Delphi participants’ minds . Although expediting symptomatic diagnosis in these cancers may have less effect in improving survival compared with other cancers, it does not mean there is no value in trying to do so. The primary message is that we have to be realistic about what it is possible to achieve. A clear implication can be drawn: that research efforts should concentrate on other aspects, such as prevention, development of novel screening tools, including new biomarkers, and new targeted therapeutic agents. Quite rightly, these cancers with poor outcomes should receive preferential attention and funding, but that this should focus on areas most likely to offer tangible benefits.
This study has produced an initial ranking for the mortality benefit available from expedited symptomatic diagnosis of cancer. It can guide future research directions, to ensure they have the maximum chance of making an impact. It should not be used as an excuse to downgrade research on cancer sites near the bottom of the ranking – but as a pointer towards the optimum research direction.
The authors would like to acknowledge other members of the ABC-DEEP consortium who made a contribution to this work: David Weller, Peter Vedsted, Marie-Louise Torring, Jon Emery, Paul Charlton.
Funded by: EDAG (Cancer Research UK’s ‘Early Diagnosis Advisory Group’). RDN receives funding from Public Health Wales and Betsi Cadwaladr University Health Board. The funding body had no role in study conduct, or the analysis or interpretation of the findings.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Richards MA. The size of the prize for earlier diagnosis of cancer in England. Br J Cancer. 2009;101:S125–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Richards MA, Sainsbury J, Ramirez A, Westcombe A, Haward RA. Influence of delay on survival in patients with breast cancer: a systematic review. Lancet. 1999;353:2155–62.View ArticleGoogle Scholar
- Tørring ML, Frydenberg M, Hamilton W, Hansen RP, Lautrup MD, Vedsted P. Diagnostic interval and mortality in colorectal cancer: U-shaped association demonstrated for three different datasets. J Clin Epidemiol. 2012;65:669–78.View ArticlePubMedGoogle Scholar
- Murchie P, Raja EA, Brewster DH, Campbell NC, Ritchie LD, Robertson R, et al. Time from first presentation in primary care to treatment of symptomatic colorectal cancer: effect on disease stage and survival. Br J Cancer. 2014;111:461–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Berrino F, De Angelis R, Sant M, Rosso S, Lasota MB, Coebergh JW, et al. Survival for eight major cancers and all cancers combined for European adults diagnosed in 1995–99: results of the EUROCARE-4 study. Lancet Oncol. 2007;8:773–83.View ArticlePubMedGoogle Scholar
- Banks J, Hollinghurst S, Bigwood L, Peters TJ, Walter FM, Hamilton W. Preferences for cancer investigation: a vignette-based study of primary-care attendees. Lancet Oncol. 2014;15:232–40.View ArticlePubMedGoogle Scholar
- Lyratzopoulos G, Neal RD, Barbiere JM, Rubin GP, Abel GA. Variation in number of general practitioner consultations before hospital referral for cancer: findings from the 2010 National Cancer Patient Experience Survey in England. Lancet Oncol. 2012;13:353–65.View ArticlePubMedGoogle Scholar
- Neal R, Din N, Hamilton W, Ukoumunne O, Carter B, Stapley S, et al. Comparison of cancer diagnostic intervals before and after implementation of NICE guidelines: analysis of data from the UK General Practice Research Database Br J Cancer. 2013; in press.Google Scholar
- Elliss-Brookes L, McPhail S, Ives A, Greenslade M, Shelton J, Hiom S, et al. Routes to diagnosis for cancer - determining the patient journey using multiple routine data sets. Br J Cancer. 2012;107:1220–6.View ArticlePubMedPubMed CentralGoogle Scholar
- Chambers S, Dunn J, Occhipinti S, Hughes S, Baade P, Sinclair S, et al. A systematic review of the impact of stigma and nihilism on lung cancer outcomes. BMC Cancer. 2012;12:183.View ArticleGoogle Scholar
- National Cancer Research Institute. The National Cancer Research Institute (NCRI) Cancer research in the UK 2002–2011: an overview of the research funded by NCRI Partners.; 2013.Google Scholar
- Abdel-Rahman M, Stockton D, Rachet B, Hakulinen T, Coleman MP. What if cancer survival in Britain were the same as in Europe: how many deaths are avoidable? Br J Cancer. 2009;101:S115–24.View ArticlePubMedPubMed CentralGoogle Scholar
- Office for National Statistics: Cancer survival in England: Patients diagnosed 2005–2009 and followed up to 2010. London; 2011.Google Scholar
- Murphy E, Black N, Lamping D, McKee C, Sanderson C. Consensus development methods, and their use in clinical guideline development: a review. Health Technol Assess. 1998;2:88.Google Scholar
- Black N, Murphy M, Lamping D, McKee M, Sanderson C, Askham J, et al. Consensus development methods: a review of best practice in creating clinical guidelines. J Health Serv Res Policy. 1999;4:236–48.PubMedGoogle Scholar
- Neal R, Tharmanathan P, France B, Din N, Peters T, Hamilton W, et al. Time to diagnosis in symptomatic cancer – does it have an effect on clinical outcomes? Systematic review and metaanalysis. In 2nd NAEDI research conference. Cambridge, UK 2013.Google Scholar
- Neal RD, Tharmanathan P, France B, Din NU, Cotton S, Fallon-Ferguson J, et al. Is increased time to diagnosis and treatment in symptomatic cancer associated with poorer outcomes? Systematic review. Br J Cancer. 2015;112:S92–107.View ArticlePubMedPubMed CentralGoogle Scholar
- Davino C, Fabbris L. Survey data collection and integration. Springer: New York; 2012. ISBN:978-3-642-21307-6.Google Scholar
- Corp S. Stata statistical software: release 13. College Station, TX: StataCorp LP; 2013.Google Scholar
- Atkin WS, Edwards R, Kralj-Hans I, Wooldrage K, Hart AR, Northover JMA, et al. Once-only flexible sigmoidoscopy screening in prevention of colorectal cancer: a multicentre randomised controlled trial. Lancet. 2010;375:1624–33.View ArticlePubMedGoogle Scholar
- Linertová R, Serrano-Aguilar P, Posada-de-la-Paz M, Hens-Pérez M, Kanavos P, Taruscio D, et al. Delphi approach to select rare diseases for a European representative survey. The BURQOL-RD study. Health Policy. 2012;108:19–26.View ArticlePubMedGoogle Scholar
- Weller D, Vedsted P, Rubin G, Walter F, Emery J, Scott S, et al. The Aarhus statement: improving design and reporting of studies on early cancer diagnosis. Br J Cancer. 2012;106:1262–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Marmot MG, Altman DG, Cameron DA, Dewar JA, Thompson SG, Wilcox M. The benefits and harms of breast cancer screening: an independent review. Br J Cancer. 2013;108:2205–40.View ArticlePubMedPubMed CentralGoogle Scholar
- Walker S, Hyde C, Hamilton W. Risk of uterine cancer in symptomatic women in primary care: case–control study using electronic records. BJGP. 2013; in press.Google Scholar
- Weyers W. The ‘epidemic’ of melanoma between under- and overdiagnosis. J Cutan Pathol. 2012;39:9–16.View ArticlePubMedGoogle Scholar
- NICE: Ovarian cancer. The recognition and initial management of ovarian cancer. (NICE ed., vol. CG122: NICE; 2011.Google Scholar
- Bill-Axelson A, Holmberg L, Garmo H, Rider JR, Taari K, Busch C, et al. Radical prostatectomy or watchful waiting in early prostate cancer. New England Journal of Medicine. 2014;370:932–42.View ArticlePubMedPubMed CentralGoogle Scholar
- Quaresma M, Coleman MP, Rachet B. 40-year trends in an index of survival for all cancers combined and survival adjusted for age and sex for each cancer in England and Wales, 1971–2011: a population-based study. Lancet. 2015;385:1206–18.View ArticlePubMedGoogle Scholar