Available online at www.sciencedirect.com
British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
Review
Two-week rule in head and neck cancer 2000-14: a systematic review Steve Langton a,∗ , Derrick Siau b , Clare Bankhead c a b c
Royal Blackburn Hospital, Blackburn, UK Wythenshawe Hospital, Manchester, UK Department of Primary Healthcare Sciences, University of Oxford, UK
Accepted 12 September 2015 Available online 12 January 2016
Abstract The fast-track system in the UK for patients with suspected cancer – the two-week rule - states that if cancer is suspected there should be a maximum of 14 days between referral from primary care and consultation with a specialist. This approach is valued by patients, ensures a universal standard of diagnosis, and speeds up the overall management of cancer. However, some say that the rule has had little or no effect on survival, results in a diagnosis of cancer in only a small proportion of patients referred, and is expensive. We have made a systematic review of the effectiveness of the two-week rule in patients with head and neck cancer with the aid of electronic searches of databases. including MEDLINE, EMBASE, the Cochrane Database of Systematic Reviews CINAHL, and CANCERLIT up to the end of 2014. This was supplemented by searching conference proceedings and contacting experts. Retrospective and prospective studies that included either conversion rate (proportion of two-week referrals who were diagnosed with cancer – positive predictive value), or detection rate (proportion of diagnosed cancers referred under the two-week rule - sensitivity), or both, were included. Two reviewers assessed studies for inclusion, and extracted data independently. Heterogeneity was assessed by inspection of the overlap of 95% CI in the forest plot and calculation of I2 . We made a random-effects meta-analysis of 17 studies. All reported the conversion rate, and 10 also reported the detection rate. Meta-analysis indicated an overall pooled conversion rate of 8.8% (95% CI 7.0% to 10.7%) and a pooled detection rate of 40.8% (95% CI 25.7% to 55.8%) Subgroups in which maxillofacial (OFMS) and otolaryngology (ENT) were assessed showed no significant difference in conversion rate (8.3% and 8.8%; p = 0.73). Subgroup analyses of early studies (before the end of 2008) and later studies (2009–14), showed a significant reduction in conversion rates from 10.6% to 6.6%, p = < 0.0001. These early and late subgroups showed a significant increase in detection rate (35.0% to 49.7%, p = 0.0008). The conversion and detection rates were similar to those for a number of other cancer sites that relied on a list of signs and symptoms for referral and were similar in both ENT and OMFS units. There is evidence that two-week referral conversion rates are falling, while detection rates are rising because of an increased number of referrals. The influence of the two-week referrral on outcomes, particularly survival, is not well known. © 2015 The British Association of Oral and Maxillofacial Surgeons. Published by Elsevier Ltd. All rights reserved.
Keywords: Fast-track cancer referral; Two-week rule; Head and neck cancer
Introduction The two-week rule, sometimes referred to as the two-week wait, was introduced by the government in 1999 for breast
∗
Corresponding author.
cancer, and in 2000 for all other cancers. It states that there should be a maximum 14-day wait between the patient’s referral by a primary care practitioner to being seen by a hospital specialist. In the early 1990s inadequacies in cancer care in the UK were being identified and commented on. The media described a so-called “cancer lottery” in which the chances
http://dx.doi.org/10.1016/j.bjoms.2015.09.041 0266-4356/© 2015 The British Association of Oral and Maxillofacial Surgeons. Published by Elsevier Ltd. All rights reserved.
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
of successful treatment depended on being referred to an appropriate specialist rather than a non-specialist.1,2 In 1995 the first EUROCARE studies were published,3 which indicated that outcomes for some cancers were worse in the UK than in some European countries. In response, the government introduced the two-week rule in The New NHS Modern, Dependable4 and the policy was put into effect, beginning in 1999. However, the two-week rule has been the subject of considerable debate. Those in favour say that the rule is valued by patients, it ensures a universal standard of treatment, and is important in speeding up the overall management of cancer. Opponents say that it has had little or no effect on survival, that cancer is diagnosed in only a small proportion of patients referred in this way, and it is expensive to implement. In June 2001, Jones et al.,5 pointed out that there was no good evidence that delays (of weeks) in diagnosing cancer translated into improved staging or survival, and suggested that the predictive values of symptoms of cancer were poorlydefined. Subsequently, Jiwa and Saunders6 reported that the proportion of people referred under the two-week rule who were diagnosed with cancer decreased from 1999-2005, and they suggested that the tick-box system of referral using a list of key symptoms may not, in isolation, be adequate.7 Many studies of a broad range of cancers have examined the conversion rate (positive predictive value – the proportion of patients referred within two weeks who had cancer) and the detection rate (sensitivity – the proportion of patients with cancer who were referred under the two-week rule). For some types of cancer, including breast8 , gastrointestinal,9 and colorectal,10 the conversion rate rarely exceeded 10%. Kumar et al11 reported a short review of patients with head and neck cancer that included six studies before 2008. A number of studies have been published since then, but an updated review is overdue. The principal aim of this systematic review is, therefore, to assess the effectiveness of the two-week rule in the diagnosis of head and neck cancer for the period 2000–14 by collating evidence from all studies that fulfilled the appropriate inclusion criteria, and by examining the conversion rates and the detection rates.
Methods Papers studied We included prospective and retrospective cohort studies, audits, and abstracts presented at meetings that reported specifically on two-week referral in the UK National Health Service. Studies from other countries were excluded, as were case reports, economic evaluations, and qualitative studies. The single intervention that we studied was fast-track referral through the UK two-week referral from primary to secondary care, using the agreed national guidelines for
121
patients with suspected head and neck cancer.7 This includes referrals from both doctors and dentists in primary care. Outcomes Our main outcome measures were the conversion rate – the proportion of two-week referrals who did have cancer (positive predictive value) and the detection rate – the proportion of cancers diagnosed in patients who had been referred within two weeks (sensitivity). Search We searched MEDLINE (Ovid), EMBASE (Ovid), Scopus, Google Scholar, CANCERLIT, Web of Science (v5.13), CINAHL, CENTRAL (Cochrane Central Register of Controlled Trials), www.clinicaltrialsregister.eu, and www.theses.com from 1999 to June 2014. We also searched the reference lists of identified studies, the authors mentioned in identified studies, a citation index using Thomson Scientific, the “find similar” facility of databases, conference proceedings of the British Association of Oral and Maxillofacial Surgeons (BAOMS) and the British Academic Conference in Otolaryngology (BACO), and consulted known experts. Finally, we looked through back numbers of J Laryngol Otol and Br J Oral Maxillofac Surg. The following terms were used in the search: 2 week, 14 day, fourteen day, rule, wait, referral, standard, urgent referral, fast track, early detection of cancer, head and neck, oral cancer, laryngeal cancer, pharyngeal cancer, and mouth neoplasms. Selection of studies We considered retrospective and prospective studies that included at least the conversion rate or the detection rate. Two reviewers assessed studies for inclusion and extracted data independently. Differences were resolved by discussion. Papers identified from all the sources mentioned were assessed, and duplicates eliminated. Abstracts were studied by two reviewers to shortlist papers that were about the two-week rule in relation to head and neck cancer. Abstracts that clearly showed that the study did not fulfil the inclusion criteria (for example, qualitative studies or commentaries) were eliminated at this stage. We obtained copies of the full papers for those that remained. Extraction of data Data were extracted by two reviewers independently as follows: first author, title of article, date of publication, location of study in the UK, type of study (prospective, retrospective, audit, or abstract presented at a meeting), type of unit in which the study was carried out (OMFS, ENT, head and neck, or other), type of participants (general head and neck, or a
122
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
subgroup such as oral), duration of the study (months), dates between which the study was done, number of referrals under the two-week rule during the study period, number of patients referred under the two-week rule who were diagnosed with cancer, conversion rate (the percentage of two-week referrals who were diagnosed with cancer) and the detection rate (the proportion of all cancers that were referred under the two-week rule).
Quality of each study We used an index for assessment of the quality of nonrandomised studies (MINORS) that was developed by Slim et al12 . On the MINORS scale, items score 2 (reported and adequate), 1 (reported but inadequate), or 0 (not reported). Previous workers10,13 have used the index to calculate an approximate assessment of study quality (0-5 = poor; 6-10 = fair; and 11-16 = good). However, the use of numerical scales for judging the quality of a study has been questioned,14 so we assessed the studies with the MINORS scale and presented the results in Table 1 to give an overall view of the quality of the studies. When we assessed the quality of a study, we also considered its duration. As we know that there are seasonal variations in the presentation of cancers,15–17 we thought that studies that lasted for less than a year were at greater risk of bias.
Analysis of subgroups We subdivided papers in to early (up to and including 2008) and late (2009-14) studies, and in to those done in ENT, and those done in OMFS, units.
Synthesis and analysis of data For data synthesis and analysis we used StatsDirect 3© software. Forest plots were generated for the conversion rates and the detection rates (when reported). The software estimated statistical heterogeneity, and calculated I2 values together with their 95% CI. The software automatically calculated fixed effects and random effects meta-analyses. Random effects analysis does not assume a common effect size and, because the subjects and interventions in the studies differed, we used the random effects method. Publication bias We considered whether papers were more likely to have been published if they reported good or favourable outcomes. Would a paper be more likely to be published (or published rapidly) if it showed that the two-week referral was highly effective or highly ineffective? Given the types of studies included, which were not controlled trials, we thought that publication bias was unlikely to be a factor.
Results Missing data Though all studies included the conversion rate as an outcome, they did not all include the detection rate, and it was not possible to obtain the relevant missing data. Outcomes for conversion and detection have therefore been analysed separately.
A total of 1086 papers were identified by the electronic searches for potential inclusion, and four papers by other methods. Thirty-one papers concerned head and neck cancer, two of which were not of the types specified in the protocol. Fourteen papers did not meet the specified inclusion criteria, but two papers each included two separate audits that reported appropriate outcomes.18,19 As a result, 17 studies were included as shown in the PRISMA flow diagram (Fig. 1).
Heterogeneity
Description of studies included
There are a number of potential sources of heterogeneity, both clinically and in the methods chosen. Patients in the different studies will differ. Cancer of the head and neck predominates in the lower social classes, and has strong associations with the use of tobacco and alcohol. In addition, a number of different anatomical sites are classified as head and neck cancer, and different studies may contain different proportions of the various types. To assess heterogeneity we generated forest plots and looked at the 95% CI, and we calculated I2 and the 95% CI for I2 which express, as a percentage, the proportion of variation in the study estimates that may result from heterogeneity.
The details of the characteristics of the studies included are shown as supplementary data, online only. Seventeen studies met the inclusion criteria.18–32 Of these, all reported the conversion rate (positive predictive value) and only 10 reported both conversion rate and detection rate (sensitivity). Twelve studies were full published papers while 5 were abstracts of papers presented at meetings. Fifteen were retrospective and two prospective. Seven studies were done in OMFS units, nine in ENT units, and one was done jointly between the two at the same hospital. Twelve studied head and neck cancer while five of the OMFS units looked at oral cancer. The duration of study varied from 3-36 months (mean 12). The number of two-week referrals examined ranged from
First author and reference no
Clearly stated aim
Inclusion of consecutive patients
Prospective collection of data
Endpoints appropriate to the aim of the study
Unbiased assessment of the study endpoint
Follow-up period appropriate to the aim of the study
Loss to follow up less than 5%
Prospective calculation of study size
MINORS quality assessment
Williams20 Lyons21 Shah22 East23 Singh24 Duvvi25 Hobson26 McKie27 Ahmad28 Haikel (audit 1)18 Haikel (audit 2)18 Miller (audit 1)19 Miller (audit 2)19 Madhvani29 Joshi30 Davey31 Kayhanian32
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
2 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 2 2 2 1 2 2 2 1 2 2 1 1 1 2 1 2
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
1 2 2 1 2 2 2 2 0 0 1 2 2 1 2 2 2
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
12 12 14 11 11 12 12 12 9 10 11 11 11 10 10 11 12
2 = reported and adequate; 1 = reported but inadequate; and 0= not reported. Endpoints: when both conversion and detection rates were included = 2, when only one was reported = 1. Follow-up > 1 year = 2; 6 months to 1 year = 1; less than 6 months = 0.
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
Table 1 Quality of the studies.
123
124
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
Fig. 1. PRISMA flow diagram of the study.
40-1079. Studies were all published in English and were done throughout the UK. Eight studies were published between 2002-8, and nine between 2009-13.
On-going studies No on-going studies were identified. Quality of the studies (Table 1)
Characteristics of studies excluded Fourteen studies were excluded at the final stage of selection when full articles were assessed. Six studies did not report at least one of the specified outcomes,33–38 three were commentaries on other papers,39–41 and one was a short letter with insufficient information.42 Kennedy et al43 reported both the specified outcome measures, but the study was done in Scotland where cancer referral guidelines are different. Hodgson et al44 also reported the specified outcome measures. However, they did not report a consecutive group of two-week referrals to their unit. They examined all referrals for suggestion of malignant disease, and then selected a random subgroup of 241 patients. Because of this unusual method of selection we excluded the study.
Table 1 shows the quality of the studies as assessed by the MINORS scale. Using the criteria previously used by Elkhadem et al,13 four studies were classified as fair and 13 as good. Effects of the intervention Overall conversion rate (positive predictive value): all 17 studies reported the conversion rate. A total of 4028 two-week referrals were made, 360 cancers were diagnosed, and conversion rates ranged from 2.2% to 14.6%. The forest plot shows low to moderate heterogeneity. I2 was calculated to be 72.6% (95% CI 52.1% to 82.0%) and a random effects meta-analysis was done. The pooled conversion rate
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
125
Table 2 Overall conversion rate (positive predictive value) and Fig. 2. forest plot. First author, year, and reference no 20
Williams, 2002 Lyons, 2004 21 Shah, 2005 22 East, 2005 23 Singh, 2006 24 Duvvi, 2006 25 Hobson, 2008 26 McKie, 2008 27 Ahmad, 2011 28 Haikel (audit 1), 2011 18 Haikel (audit 2), 2011 18 Miller (audit 1), 2012 19 Miller (audit 2), 2012 19 Madhvani, 2012 29 Joshi, 2012 30 Davey, 2012 31 Kayhanian, 2013 32
No in study
Conversions
Percentage (%)
95% CI
Weight
100 171 150 48 76 187 177 1079 114 163 542 63 49 252 362 446 50
11 25 9 3 6 19 22 118 6 17 53 7 3 20 8 27 6
11.0 14.6 6.0 6.3 7.9 10.2 12.4 10.9 5.3 10.4 9.8 11.1 6.1 7.9 2.2 6.1 12.0
5.6 to 18.8 9.7 to 20.8 2.8 to 11.1 13.1 to 17.2 3 to 16.3 6.2 to 15.4 8 to 18.2 9.1 to 12.6 2 to 11.1 6.2 to 16.1 7.4 to 12.6 4.6 to 21.6 1.3 to 1.69 4.9 to 12 1 to 4.3 4 to 8.7 4.5 to 24.3
5.1 6.3 6 3.4 4.5 6.5 6.4 8.7 5.4 6.2 8.1 4.0 3.5 7.0 7.6 7.9 3.5
(positive predictive value) was 8.8% (95% CI 7.0% to 10.7%) (Table 2, Fig. 2). Overall detection rate (sensitivity):10 studies reported the detection rate. A total of 912 cancers were diagnosed, 280 of which had been referred under the two-week rule. Detection rates ranged from 13.6% -80.3%. Random effects meta-analysis was done and the pooled detection rate was 40.8% (95% CI 26.7% to 55.8%). (Table 3, Fig. 3). Subgroup analysis: ENT and OMFS units: nine studies were done in ENT units, 2212 two-week referrals were made,
and 183 cancers diagnosed. The conversion rate ranged from 2.2% to 14.6% and the pooled conversion rate was 8.8% (95% CI 6.0% to 12.0%) (Fig. 4). Seven studies were done in OMFS units, there were 738 two-week referrals and 59 cancers were diagnosed. The conversion rate ranged from 6.0% -1.1%, and the pooled conversion rate was 8.3% (95% CI 6.4% to 10.4%) (Fig. 5). The conversion rates in ENT and OMFS units did not differ significantly (p=0.73). We did not analyse the detection rate for OMFS and ENT units as only 2 OMFS studies reported it.
Williams, 2002
0.11 (0.06, 0.19)
Lyons, 2004
0.15 (0.10, 0.21)
Shah, 2005
0.06 (0.03, 0.11)
East, 2005
0.06 (0.01, 0.17)
Singh, 2006
0.08 (0.03, 0.16)
Duvvi, 2006
0.10 (0.06, 0.15)
Hobson, 2008
0.12 (0.08, 0.18)
McKie, 2008
0.11 (0.09, 0.13)
Ahmad, 2011
0.05 (0.02, 0.11)
Haikel (audit 1), 2011
0.10 (0.06, 0.16)
Haikel (audit 2), 2011
0.10 (0.07, 0.13)
Miller (audit 1), 2012
0.11 (0.05, 0.22)
Miller (audit 2), 2012
0.06 (0.01, 0.17)
Madhvani, 2012
0.08 (0.05, 0.12)
Joshi, 2012
0.02 (9.6E-3, 0.04)
Davey, 2012
0.06 (0.04, 0.09)
Kayhanian, 2012
0.12 (0.05, 0.24)
Combined
0.09 (0.07, 0.11) 0.0
10
20
% Conversion rate (95% CI) Fig. 2. Forest plot of overall conversion rate (positive predictive value).
30
126
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
Table 3 Overall detection rate (sensitivity) and Fig. 3 forest plot. First author, year, and reference no 200421
Lyons, Shah, 200522 East, 200523 Duvvi, 200625 Hobson, 200826 McKie, 200827 Haikel (audit 1), 201118 Haikel (audit 2), 201118 Joshi, 201230 Kayhanian, 201332
No. of cancers
Detections by TWR
85 27 22 31 39 552 57 66 17 16
25 9 3 19 22 118 17 53 8 6
Subgroup analysis: early (2001-8) and late (2009-14) studies: 10 studies were done in the early period and seven in the later period. The pooled conversion rate from the fixed effect meta-analysis was 10.6% (95% CI 9.4% to 11.8%) in the early group, and 6.6% (95% CI 4.2% to 9.6%) in the late group, and the difference is significant (p
British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
Review
Two-week rule in head and neck cancer 2000-14: a systematic review Steve Langton a,∗ , Derrick Siau b , Clare Bankhead c a b c
Royal Blackburn Hospital, Blackburn, UK Wythenshawe Hospital, Manchester, UK Department of Primary Healthcare Sciences, University of Oxford, UK
Accepted 12 September 2015 Available online 12 January 2016
Abstract The fast-track system in the UK for patients with suspected cancer – the two-week rule - states that if cancer is suspected there should be a maximum of 14 days between referral from primary care and consultation with a specialist. This approach is valued by patients, ensures a universal standard of diagnosis, and speeds up the overall management of cancer. However, some say that the rule has had little or no effect on survival, results in a diagnosis of cancer in only a small proportion of patients referred, and is expensive. We have made a systematic review of the effectiveness of the two-week rule in patients with head and neck cancer with the aid of electronic searches of databases. including MEDLINE, EMBASE, the Cochrane Database of Systematic Reviews CINAHL, and CANCERLIT up to the end of 2014. This was supplemented by searching conference proceedings and contacting experts. Retrospective and prospective studies that included either conversion rate (proportion of two-week referrals who were diagnosed with cancer – positive predictive value), or detection rate (proportion of diagnosed cancers referred under the two-week rule - sensitivity), or both, were included. Two reviewers assessed studies for inclusion, and extracted data independently. Heterogeneity was assessed by inspection of the overlap of 95% CI in the forest plot and calculation of I2 . We made a random-effects meta-analysis of 17 studies. All reported the conversion rate, and 10 also reported the detection rate. Meta-analysis indicated an overall pooled conversion rate of 8.8% (95% CI 7.0% to 10.7%) and a pooled detection rate of 40.8% (95% CI 25.7% to 55.8%) Subgroups in which maxillofacial (OFMS) and otolaryngology (ENT) were assessed showed no significant difference in conversion rate (8.3% and 8.8%; p = 0.73). Subgroup analyses of early studies (before the end of 2008) and later studies (2009–14), showed a significant reduction in conversion rates from 10.6% to 6.6%, p = < 0.0001. These early and late subgroups showed a significant increase in detection rate (35.0% to 49.7%, p = 0.0008). The conversion and detection rates were similar to those for a number of other cancer sites that relied on a list of signs and symptoms for referral and were similar in both ENT and OMFS units. There is evidence that two-week referral conversion rates are falling, while detection rates are rising because of an increased number of referrals. The influence of the two-week referrral on outcomes, particularly survival, is not well known. © 2015 The British Association of Oral and Maxillofacial Surgeons. Published by Elsevier Ltd. All rights reserved.
Keywords: Fast-track cancer referral; Two-week rule; Head and neck cancer
Introduction The two-week rule, sometimes referred to as the two-week wait, was introduced by the government in 1999 for breast
∗
Corresponding author.
cancer, and in 2000 for all other cancers. It states that there should be a maximum 14-day wait between the patient’s referral by a primary care practitioner to being seen by a hospital specialist. In the early 1990s inadequacies in cancer care in the UK were being identified and commented on. The media described a so-called “cancer lottery” in which the chances
http://dx.doi.org/10.1016/j.bjoms.2015.09.041 0266-4356/© 2015 The British Association of Oral and Maxillofacial Surgeons. Published by Elsevier Ltd. All rights reserved.
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
of successful treatment depended on being referred to an appropriate specialist rather than a non-specialist.1,2 In 1995 the first EUROCARE studies were published,3 which indicated that outcomes for some cancers were worse in the UK than in some European countries. In response, the government introduced the two-week rule in The New NHS Modern, Dependable4 and the policy was put into effect, beginning in 1999. However, the two-week rule has been the subject of considerable debate. Those in favour say that the rule is valued by patients, it ensures a universal standard of treatment, and is important in speeding up the overall management of cancer. Opponents say that it has had little or no effect on survival, that cancer is diagnosed in only a small proportion of patients referred in this way, and it is expensive to implement. In June 2001, Jones et al.,5 pointed out that there was no good evidence that delays (of weeks) in diagnosing cancer translated into improved staging or survival, and suggested that the predictive values of symptoms of cancer were poorlydefined. Subsequently, Jiwa and Saunders6 reported that the proportion of people referred under the two-week rule who were diagnosed with cancer decreased from 1999-2005, and they suggested that the tick-box system of referral using a list of key symptoms may not, in isolation, be adequate.7 Many studies of a broad range of cancers have examined the conversion rate (positive predictive value – the proportion of patients referred within two weeks who had cancer) and the detection rate (sensitivity – the proportion of patients with cancer who were referred under the two-week rule). For some types of cancer, including breast8 , gastrointestinal,9 and colorectal,10 the conversion rate rarely exceeded 10%. Kumar et al11 reported a short review of patients with head and neck cancer that included six studies before 2008. A number of studies have been published since then, but an updated review is overdue. The principal aim of this systematic review is, therefore, to assess the effectiveness of the two-week rule in the diagnosis of head and neck cancer for the period 2000–14 by collating evidence from all studies that fulfilled the appropriate inclusion criteria, and by examining the conversion rates and the detection rates.
Methods Papers studied We included prospective and retrospective cohort studies, audits, and abstracts presented at meetings that reported specifically on two-week referral in the UK National Health Service. Studies from other countries were excluded, as were case reports, economic evaluations, and qualitative studies. The single intervention that we studied was fast-track referral through the UK two-week referral from primary to secondary care, using the agreed national guidelines for
121
patients with suspected head and neck cancer.7 This includes referrals from both doctors and dentists in primary care. Outcomes Our main outcome measures were the conversion rate – the proportion of two-week referrals who did have cancer (positive predictive value) and the detection rate – the proportion of cancers diagnosed in patients who had been referred within two weeks (sensitivity). Search We searched MEDLINE (Ovid), EMBASE (Ovid), Scopus, Google Scholar, CANCERLIT, Web of Science (v5.13), CINAHL, CENTRAL (Cochrane Central Register of Controlled Trials), www.clinicaltrialsregister.eu, and www.theses.com from 1999 to June 2014. We also searched the reference lists of identified studies, the authors mentioned in identified studies, a citation index using Thomson Scientific, the “find similar” facility of databases, conference proceedings of the British Association of Oral and Maxillofacial Surgeons (BAOMS) and the British Academic Conference in Otolaryngology (BACO), and consulted known experts. Finally, we looked through back numbers of J Laryngol Otol and Br J Oral Maxillofac Surg. The following terms were used in the search: 2 week, 14 day, fourteen day, rule, wait, referral, standard, urgent referral, fast track, early detection of cancer, head and neck, oral cancer, laryngeal cancer, pharyngeal cancer, and mouth neoplasms. Selection of studies We considered retrospective and prospective studies that included at least the conversion rate or the detection rate. Two reviewers assessed studies for inclusion and extracted data independently. Differences were resolved by discussion. Papers identified from all the sources mentioned were assessed, and duplicates eliminated. Abstracts were studied by two reviewers to shortlist papers that were about the two-week rule in relation to head and neck cancer. Abstracts that clearly showed that the study did not fulfil the inclusion criteria (for example, qualitative studies or commentaries) were eliminated at this stage. We obtained copies of the full papers for those that remained. Extraction of data Data were extracted by two reviewers independently as follows: first author, title of article, date of publication, location of study in the UK, type of study (prospective, retrospective, audit, or abstract presented at a meeting), type of unit in which the study was carried out (OMFS, ENT, head and neck, or other), type of participants (general head and neck, or a
122
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
subgroup such as oral), duration of the study (months), dates between which the study was done, number of referrals under the two-week rule during the study period, number of patients referred under the two-week rule who were diagnosed with cancer, conversion rate (the percentage of two-week referrals who were diagnosed with cancer) and the detection rate (the proportion of all cancers that were referred under the two-week rule).
Quality of each study We used an index for assessment of the quality of nonrandomised studies (MINORS) that was developed by Slim et al12 . On the MINORS scale, items score 2 (reported and adequate), 1 (reported but inadequate), or 0 (not reported). Previous workers10,13 have used the index to calculate an approximate assessment of study quality (0-5 = poor; 6-10 = fair; and 11-16 = good). However, the use of numerical scales for judging the quality of a study has been questioned,14 so we assessed the studies with the MINORS scale and presented the results in Table 1 to give an overall view of the quality of the studies. When we assessed the quality of a study, we also considered its duration. As we know that there are seasonal variations in the presentation of cancers,15–17 we thought that studies that lasted for less than a year were at greater risk of bias.
Analysis of subgroups We subdivided papers in to early (up to and including 2008) and late (2009-14) studies, and in to those done in ENT, and those done in OMFS, units.
Synthesis and analysis of data For data synthesis and analysis we used StatsDirect 3© software. Forest plots were generated for the conversion rates and the detection rates (when reported). The software estimated statistical heterogeneity, and calculated I2 values together with their 95% CI. The software automatically calculated fixed effects and random effects meta-analyses. Random effects analysis does not assume a common effect size and, because the subjects and interventions in the studies differed, we used the random effects method. Publication bias We considered whether papers were more likely to have been published if they reported good or favourable outcomes. Would a paper be more likely to be published (or published rapidly) if it showed that the two-week referral was highly effective or highly ineffective? Given the types of studies included, which were not controlled trials, we thought that publication bias was unlikely to be a factor.
Results Missing data Though all studies included the conversion rate as an outcome, they did not all include the detection rate, and it was not possible to obtain the relevant missing data. Outcomes for conversion and detection have therefore been analysed separately.
A total of 1086 papers were identified by the electronic searches for potential inclusion, and four papers by other methods. Thirty-one papers concerned head and neck cancer, two of which were not of the types specified in the protocol. Fourteen papers did not meet the specified inclusion criteria, but two papers each included two separate audits that reported appropriate outcomes.18,19 As a result, 17 studies were included as shown in the PRISMA flow diagram (Fig. 1).
Heterogeneity
Description of studies included
There are a number of potential sources of heterogeneity, both clinically and in the methods chosen. Patients in the different studies will differ. Cancer of the head and neck predominates in the lower social classes, and has strong associations with the use of tobacco and alcohol. In addition, a number of different anatomical sites are classified as head and neck cancer, and different studies may contain different proportions of the various types. To assess heterogeneity we generated forest plots and looked at the 95% CI, and we calculated I2 and the 95% CI for I2 which express, as a percentage, the proportion of variation in the study estimates that may result from heterogeneity.
The details of the characteristics of the studies included are shown as supplementary data, online only. Seventeen studies met the inclusion criteria.18–32 Of these, all reported the conversion rate (positive predictive value) and only 10 reported both conversion rate and detection rate (sensitivity). Twelve studies were full published papers while 5 were abstracts of papers presented at meetings. Fifteen were retrospective and two prospective. Seven studies were done in OMFS units, nine in ENT units, and one was done jointly between the two at the same hospital. Twelve studied head and neck cancer while five of the OMFS units looked at oral cancer. The duration of study varied from 3-36 months (mean 12). The number of two-week referrals examined ranged from
First author and reference no
Clearly stated aim
Inclusion of consecutive patients
Prospective collection of data
Endpoints appropriate to the aim of the study
Unbiased assessment of the study endpoint
Follow-up period appropriate to the aim of the study
Loss to follow up less than 5%
Prospective calculation of study size
MINORS quality assessment
Williams20 Lyons21 Shah22 East23 Singh24 Duvvi25 Hobson26 McKie27 Ahmad28 Haikel (audit 1)18 Haikel (audit 2)18 Miller (audit 1)19 Miller (audit 2)19 Madhvani29 Joshi30 Davey31 Kayhanian32
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
2 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 2 2 2 1 2 2 2 1 2 2 1 1 1 2 1 2
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
1 2 2 1 2 2 2 2 0 0 1 2 2 1 2 2 2
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
12 12 14 11 11 12 12 12 9 10 11 11 11 10 10 11 12
2 = reported and adequate; 1 = reported but inadequate; and 0= not reported. Endpoints: when both conversion and detection rates were included = 2, when only one was reported = 1. Follow-up > 1 year = 2; 6 months to 1 year = 1; less than 6 months = 0.
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
Table 1 Quality of the studies.
123
124
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
Fig. 1. PRISMA flow diagram of the study.
40-1079. Studies were all published in English and were done throughout the UK. Eight studies were published between 2002-8, and nine between 2009-13.
On-going studies No on-going studies were identified. Quality of the studies (Table 1)
Characteristics of studies excluded Fourteen studies were excluded at the final stage of selection when full articles were assessed. Six studies did not report at least one of the specified outcomes,33–38 three were commentaries on other papers,39–41 and one was a short letter with insufficient information.42 Kennedy et al43 reported both the specified outcome measures, but the study was done in Scotland where cancer referral guidelines are different. Hodgson et al44 also reported the specified outcome measures. However, they did not report a consecutive group of two-week referrals to their unit. They examined all referrals for suggestion of malignant disease, and then selected a random subgroup of 241 patients. Because of this unusual method of selection we excluded the study.
Table 1 shows the quality of the studies as assessed by the MINORS scale. Using the criteria previously used by Elkhadem et al,13 four studies were classified as fair and 13 as good. Effects of the intervention Overall conversion rate (positive predictive value): all 17 studies reported the conversion rate. A total of 4028 two-week referrals were made, 360 cancers were diagnosed, and conversion rates ranged from 2.2% to 14.6%. The forest plot shows low to moderate heterogeneity. I2 was calculated to be 72.6% (95% CI 52.1% to 82.0%) and a random effects meta-analysis was done. The pooled conversion rate
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
125
Table 2 Overall conversion rate (positive predictive value) and Fig. 2. forest plot. First author, year, and reference no 20
Williams, 2002 Lyons, 2004 21 Shah, 2005 22 East, 2005 23 Singh, 2006 24 Duvvi, 2006 25 Hobson, 2008 26 McKie, 2008 27 Ahmad, 2011 28 Haikel (audit 1), 2011 18 Haikel (audit 2), 2011 18 Miller (audit 1), 2012 19 Miller (audit 2), 2012 19 Madhvani, 2012 29 Joshi, 2012 30 Davey, 2012 31 Kayhanian, 2013 32
No in study
Conversions
Percentage (%)
95% CI
Weight
100 171 150 48 76 187 177 1079 114 163 542 63 49 252 362 446 50
11 25 9 3 6 19 22 118 6 17 53 7 3 20 8 27 6
11.0 14.6 6.0 6.3 7.9 10.2 12.4 10.9 5.3 10.4 9.8 11.1 6.1 7.9 2.2 6.1 12.0
5.6 to 18.8 9.7 to 20.8 2.8 to 11.1 13.1 to 17.2 3 to 16.3 6.2 to 15.4 8 to 18.2 9.1 to 12.6 2 to 11.1 6.2 to 16.1 7.4 to 12.6 4.6 to 21.6 1.3 to 1.69 4.9 to 12 1 to 4.3 4 to 8.7 4.5 to 24.3
5.1 6.3 6 3.4 4.5 6.5 6.4 8.7 5.4 6.2 8.1 4.0 3.5 7.0 7.6 7.9 3.5
(positive predictive value) was 8.8% (95% CI 7.0% to 10.7%) (Table 2, Fig. 2). Overall detection rate (sensitivity):10 studies reported the detection rate. A total of 912 cancers were diagnosed, 280 of which had been referred under the two-week rule. Detection rates ranged from 13.6% -80.3%. Random effects meta-analysis was done and the pooled detection rate was 40.8% (95% CI 26.7% to 55.8%). (Table 3, Fig. 3). Subgroup analysis: ENT and OMFS units: nine studies were done in ENT units, 2212 two-week referrals were made,
and 183 cancers diagnosed. The conversion rate ranged from 2.2% to 14.6% and the pooled conversion rate was 8.8% (95% CI 6.0% to 12.0%) (Fig. 4). Seven studies were done in OMFS units, there were 738 two-week referrals and 59 cancers were diagnosed. The conversion rate ranged from 6.0% -1.1%, and the pooled conversion rate was 8.3% (95% CI 6.4% to 10.4%) (Fig. 5). The conversion rates in ENT and OMFS units did not differ significantly (p=0.73). We did not analyse the detection rate for OMFS and ENT units as only 2 OMFS studies reported it.
Williams, 2002
0.11 (0.06, 0.19)
Lyons, 2004
0.15 (0.10, 0.21)
Shah, 2005
0.06 (0.03, 0.11)
East, 2005
0.06 (0.01, 0.17)
Singh, 2006
0.08 (0.03, 0.16)
Duvvi, 2006
0.10 (0.06, 0.15)
Hobson, 2008
0.12 (0.08, 0.18)
McKie, 2008
0.11 (0.09, 0.13)
Ahmad, 2011
0.05 (0.02, 0.11)
Haikel (audit 1), 2011
0.10 (0.06, 0.16)
Haikel (audit 2), 2011
0.10 (0.07, 0.13)
Miller (audit 1), 2012
0.11 (0.05, 0.22)
Miller (audit 2), 2012
0.06 (0.01, 0.17)
Madhvani, 2012
0.08 (0.05, 0.12)
Joshi, 2012
0.02 (9.6E-3, 0.04)
Davey, 2012
0.06 (0.04, 0.09)
Kayhanian, 2012
0.12 (0.05, 0.24)
Combined
0.09 (0.07, 0.11) 0.0
10
20
% Conversion rate (95% CI) Fig. 2. Forest plot of overall conversion rate (positive predictive value).
30
126
S. Langton et al. / British Journal of Oral and Maxillofacial Surgery 54 (2016) 120–131
Table 3 Overall detection rate (sensitivity) and Fig. 3 forest plot. First author, year, and reference no 200421
Lyons, Shah, 200522 East, 200523 Duvvi, 200625 Hobson, 200826 McKie, 200827 Haikel (audit 1), 201118 Haikel (audit 2), 201118 Joshi, 201230 Kayhanian, 201332
No. of cancers
Detections by TWR
85 27 22 31 39 552 57 66 17 16
25 9 3 19 22 118 17 53 8 6
Subgroup analysis: early (2001-8) and late (2009-14) studies: 10 studies were done in the early period and seven in the later period. The pooled conversion rate from the fixed effect meta-analysis was 10.6% (95% CI 9.4% to 11.8%) in the early group, and 6.6% (95% CI 4.2% to 9.6%) in the late group, and the difference is significant (p
