Primary care endorsement letter and a patient leaflet to improve participation in colorectal cancer screening: results of a factorial randomised trial.

BACKGROUND: The trial aimed to investigate whether a general practitioner's (GP) letter encouraging participation and a more explicit leaflet explaining how to complete faecal occult blood test (FOBT) included with the England Bowel Cancer Screening Programme invitation materials would improve uptake.
METHODS: A randomised controlled 2 × 2 factorial trial was conducted in the south of England. Overall, 1288 patients registered with 20 GPs invited for screening in October 2009 participated in the trial. Participants were randomised to either a GP's endorsement letter and/or an enhanced information leaflet with their FOBT kit. The primary outcome was verified with return of the test kit within 20 weeks.
RESULTS: Both the GP's endorsement letter and the enhanced procedural leaflet, each increased participation by ∼6% - the GP's letter by 5.8% (95% CI: 4.1-7.8%) and the leaflet by 6.0% (95% CI: 4.3-8.1%). On the basis of the intention-to-treat analysis, the random effects logistic regression model confirmed that there was no important interaction between the two interventions, and estimated an adjusted rate ratio of 1.11 (P=0.038) for the GP's letter and 1.12 (P=0.029) for the leaflet. In the absence of an interaction, an additive effect for receiving both the GP's letter and leaflet (11.8%, 95% CI: 8.5-16%) was confirmed. The per-protocol analysis indicated that the insertion of an electronic GP's signature on the endorsement letter was associated with increased participation (P=0.039).
CONCLUSION: Including both an endorsement letter from each patient's GP and a more explicit procedural leaflet could increase participation in the English Bowel Cancer Screening Programme by ∼10%, a relative improvement of 20% on current performance.

RCT Entities:   Population Interventions Outcomes

Colorectal cancer (CRC) is the second leading cause of cancer deaths in the United Kingdom (CRUK Cancerstats), and survival rates are lower than in other European countries (Berrino ). Only 13% of UK CRC cases are diagnosed at the earliest stage of the disease (NCIN, 2009). As screening using the faecal occult blood test (FOBT) can increase early-stage diagnosis, significantly reducing bowel cancer mortality (Hewitson ), a national screening programme was introduced in spring 2006 and it currently covers all areas of England.

The England Bowel Cancer Screening Programme invites men and women aged 60–69 years (increased to 60–74 years from April 2010) to participate in faecal occult blood screening in every 2 years. Eligible people receive an invitation letter and information booklet by post. A week later (unless they decline participation during that week), they receive the test kit, which includes brief instructions and a freepost envelope for return of the kit. Currently, around half (49–52%) of the eligible population return their test kits (Weller ; von Wagner ). Although these participation rates are comparable with other European screening programmes (Denis ), increasing informed participation may result in a larger reduction in mortality (Parkin ) and previous research has consistently indicated that the involvement of general practitioners (GPs) may be beneficial (Brawarsky ; Subramanian ; Senore ). At present, GPs in the United Kingdom are not directly involved in bowel cancer screening – they only receive a copy of the results letters sent to their patients.

The simplest method of involving primary care in the screening process is for the invitation to participate to include a personal letter from the GP (at present the invitation comes from a Screening Hub not previously known to the patient). Although increasing participation can be achieved by repeated invitations to non-responders (Steele ), evidence from other countries suggests that an endorsement letter from the patient's GP can significantly improve participation (Cole ; Senore ; Zajac ). However, the GP's endorsement letter seems to affect only a proportion of the invited population. The potential improvement in participation through inclusion of a GP's signature on the endorsement letter was demonstrated by Cole , although other studies have not shown a significant increase in participation using this method (Segnan ; Ling ).

Other patient-specific factors such as poor knowledge of the benefits of bowel cancer screening, concerns about performing the FOBT and perceived self-efficacy for completing the bowel cancer screening procedure can also contribute to poor participation (Seeff ; Subramanian ; Weller ). Another potential difficulty concerns the effect of limited health literacy for the screening participation. People with limited health literacy have lower participation rates and report more barriers to complete FOBT than people with adequate health literacy (Dolan ; Peterson ). Addressing these difficulties, several primary care-based studies suggest providing patients with detailed instructions on the collection, storage and return of test kits, can increase compliance (Miller ; Stokamer ). Coupling this strategy of providing potential screening participants with more detailed instructions based on social cognitive approaches to alleviate misconceptions about screening (Bandura, 2004) and the presentation of risk information in a more effective manner (Rothman and Kiviniemi, 1999; Rothman ) may positively effect FOBT participation.

We, therefore, report here the results of a factorial trial to assess the impact on participation in the English Screening Programme including the test kit: (1) a letter from the patient's GP, recommending participation in the programme and (2) an information leaflet explaining more explicitly how to undertake FOBT.

Materials and methods

A 2 × 2 factorial randomised controlled trial design was adopted to assess the effectiveness of the two interventions – a GP's letter and/or a leaflet giving more explicit information on how to carry out and return the FOBT. Factorial trials have a number of advantages over the standard parallel group design. First, they enable efficient simultaneous investigation of two interventions by including all participants in both the analyses and second, the factorial trial can consider both the separate effects of the intervention and the benefits of receiving both interventions together (McAlister ; Montgomery ). Finally, the design can reduce the total number of participants required to assess the multiple interventions aimed at achieving the same outcome (Montgomery ; Gurusamy ).

Interventions

The GP's endorsement letter was a personally addressed letter from each patient's GP, which: (1) recommended that the patient complete the test; (2), offered support if the patient had any questions about screening; (3) emphasised the importance of being aware of bowel cancer symptoms. We included several key messages in the letter, based on the views of the pilot participants, influential statements reported in previous UK research (Woodrow ) and phrased using a ‘gain-frame’ approach, identified as important when targeting detection behaviours (Rothman ). The key messages included in the endorsement letter indicated the risk of developing bowel cancer was highest in the patients’ age group; there are often no symptoms associated with early bowel cancer and screening can detect bowel cancer at an early stage. In total, 8 of the 20 practices involved provided electronic GP's signatures; the others were sent ‘on behalf of the practice’.

The enhanced procedural leaflet was extensively piloted and revised with 109 people previously invited to bowel cancer screening (see Appendices), and was modified with advice from an expert steering group involved in the trial. Pilot respondents indicated that the leaflet was easy to read (95%), sufficiently detailed (98%), and included ‘very useful’ information for collecting samples (76%), storing samples (75%) and for the decision to participate or not in screening (81%). The leaflet addressed potential barriers to screening identified from the pilot study and previous research (Subramanian ; Weller ; Woodrow ). On the basis of social cognitive theory (Bandura, 2004) and effective methods for improving risk communication (Rothman and Kiviniemi, 1999), the leaflet included an educational or knowledge-building component (reinforcing messages regarding the effectiveness and rationale for screening) and motivational components designed to improve self-efficacy (advice on how to collect samples, concerns about time required, and what people with loose or irregular bowel motions should do).

Recruitment and participants

Approximately 25 000 people from more than 920 GP practices were invited for bowel cancer screening during October 2009 by the Southern Programme Hub. Following a consultation with the National Bowel Cancer Screening Programme, 20 of 88 eligible GPs, in southern England, whose patients were scheduled to receive a screening invitation in the month of October 2009, agreed to participate. We sent patients from these practices a preliminary letter in August 2009 (6 weeks before their scheduled invitation date in October). This letter informed them that they would be receiving an invitation to participate in the near future and that their own GP was involved; in general, it also provided brief information about the trial and bowel cancer screening. They then received a standard invitation letter from the NHS national screening service in October 2009. This letter explained the rationale for bowel screening and why they have been invited; an evidence-based information booklet was also included with the letter.

We included men or women registered with a GP's practice in the south of England and who would be sent an invitation to bowel cancer screening in October 2009. We excluded those who specifically requested to be withdrawn from the programme or who were currently ineligible for invitation (i.e., current bowel cancer patients, people currently in bowel cancer surveillance programmes and so on). The age range for inclusion in the trial was 60–75 years old. The programme normally invites people aged from 60 to 69 years to screening. However, two regions were involved in piloting the age extension to the programme, which meant participants from two GP's practices aged 70–75 years were also included.

Of the people invited for screening, only one from each household was included in the trial. The number eligible participants in each GP's practice ranged from 48 to 115 (median=59), and the total number of people randomised for the trial was 1288. These were randomised to four groups: (1) GP's endorsement letter, (2) enhanced procedural instruction leaflet; (3) GP's letter plus leaflet; and (4) no additional intervention (usual care). The GP's letter and leaflet were sent out with the occult blood test kits after a week of sending the invitation letter.

Randomisation procedure

We randomised using a block randomisation using the ‘ralloc’ command in STATA Version 10 (Timberlake, UK); allocation was concealed to the researchers, practitioners and screening hub staff responsible for recording test-kit returns. There were 644 people randomised to the two factorial groups (322 people allocated to each of the four intervention groups, see Figure 1). As the standard invitation letter allows people to opt not to be sent the test kit, not all those randomised received the intervention.

Figure 1

Trial flowchart.

Sample size

The sample size was based on detecting a difference in main effects between the two factorial groups (e.g., GP's letter vs no GP's letter, or leaflet vs no leaflet), not a difference between the four intervention groups (McAlister ; Montgomery ). The planned sample size of 387 participants per factorial group provided 80% power (α=0.05) to detect an absolute difference between the two factorial interventions of 10% the interventions were thought to function independently, and the study was not powered to detect interaction. It was not thought feasible to conduct a larger study with the resources available, although a difference of ∼10% was recognised to be of potential public health importance, particularly given the low cost (and hence potential for cost-effectiveness) of the interventions.

Main outcome and statistical method

The primary outcome of the trial was verified at return of the test kit to the screening hub within 20 weeks of being invited to screening. We performed data analyses using STATA version 10. The primary analysis was conducted on an intention-to-treat basis; a per-protocol analysis was also performed for people who were actually sent a test kit and hence were subject to the intervention. The statistical significance of the main intervention effects was assessed using a multiple, random effects logistic regression model, adjusting for the other intervention and five covariates (age, gender, GP's practice, previous invitation, inclusion or not of a GP's signature) (Green ; McAlister ; Montgomery ). As the outcome rate was >10%, the adjusted odds ratios from the regression model were used to estimate rate ratios by Zhang's method (Zhang and Kai, 1998).

Results

The characteristics of those participating in the trial, including whether or not they received the intervention and a comparison of the factorial groups, are shown in Table 1. There were no statistically significant differences for gender, age group or previous invitation to screening between the four intervention groups. People randomised to the ‘GP's letter and leaflet’ group (P=0.005) and the ‘GP's letter’ factorial group (P=0.004) were more likely to receive the FOBT kit than other participants.

Table 1

Participant characteristics for the four intervention groups and two factorial trial groups

Intervention groups	Letter and leaflet	GP letter only	Leaflet only	Usual care	Total	P-value
Gender
Male	153	153	154	151	611	0.996
Female	169	169	168	171	677
Age group (in years)
60–64	187	189	183	189	748	0.864
65–69	125	119	131	123	498
70+	10	14	8	10	42
Previous invite
Yes	44	38	41	47	180	0.508
No	278	284	271	275	1108
Sent FOBT
Yes	316	304	299	297	1216	0.005
No	6	18	23	25	72
GP signature
Yes	123	116	—	—	239	0.954
No	199	206	—	—	405
Total	322	322	322	322	1288
Factorial trial groups	Letter	No letter	P-value	Leaflet	No leaflet	P-value
Gender
Male	306	305	0.956	307	304	0.867
Female	338	339		337	340
Age group (in years)
60–64	376	372	0.583	370	378	0.513
65–69	244	254		256	242
70+	24	18		18	24
Previous invite
Yes	82	98	0.199	95	85	0.422
No	562	546		549	559
Sent FOBT
Yes	620	596	0.004	615	601	0.089
No	24	48		29	43
Total	644	644		644	644

Abbreviations: GP=general practitioner; FOBT=faecal occult blood test.

Letter sent with GP's signature rather than signed ‘on behalf of the practice’.

Figure 1 shows that 322 people were allocated to each of the four intervention groups, but 72 people receiving the screening invitation exercised the option of not to receive the kit (and therefore did not receive the intervention).

Table 2 shows the response rate in each of the three intervention groups individually and the comparison between the factorial groups. Both the GP's endorsement letter and the enhanced procedural information leaflet, each increased participation above usual care by about 6% – the GP's endorsement letter from 52.3 to 58.1% (absolute difference 5.8%, 95% CI: 4.1–7.8%); the leaflet from 52.2% to 58.2% (absolute difference 6.0%, 95% CI: 4.3–8.1%). The return rate in people receiving both interventions was 61.2% (absolute difference from usual care 11.8%), suggesting the effect of both interventions is additive (i.e., the absolute difference of GP's letter 5.6% and leaflet 5.9%, together is 11.5%). The absolute difference of ∼10% in the return rate for people receiving both interventions suggests a relative improvement of around 20% on current rates of participation in the English Bowel Cancer Screening Programme. The proportion of people participating in screening was higher for those receiving a signed GP's endorsement letter (64.9%) in comparison with people who received the non-signed (on behalf of the practice) endorsement letter (54.1%); an absolute difference in screening participation of almost 11%.

Table 2

Number of people returning faecal occult blood test kits within 20 weeks according to individual intervention group, factorial group and whether or not the endorsement letter was signed by the patient's general practitioner

	No.	Total	Percentage	95% CI	Difference (in %)	95% CI
Individual groups
Letter+leaflet	197	322	61.2	56–67	11.8	8.5–16
GP letter only	177	322	55.0	49–61	5.6	3.3–8.7
Leaflet only	178	322	55.3	50–61	5.9	3.6–9.1
Usual care	159	322	49.4	44–55	—	—
Factorial groups
Letter	374	644	58.1	54–62	5.8	4.1–7.8
No letter	337	—	52.3	48–56	—	—
Leaflet	375	644	58.2	54–62	6.0	4.3–8.1
No leaflet	336	—	52.2	48–56	—	—
GP letter
Letter signeda	155	239	64.9	58–71	10.8	8.6–14
Letter not signed	219	405	54.1	49–59	—	—

Abbreviations: CI=confidence interval; GP=general practitioner.

Letter sent with GP's signature rather than signed ‘on behalf of the practice’.

This additive effect was confirmed by the random effects logistic regression model (Table 3), which shows there is no suggestion of a significant interaction between the two interventions (odds ratio 1.02, P=0.979).

Table 3

Likelihood of patients returning a faecal occult blood test kit within 20 weeks: results of the logistic regression analysis

Participation (ITT analysis)	Odds ratio	95% CI	Rate ratioa	P-value
Intervention main effects b , c
Letter	1.26	1.01–1.58	1.11	0.038
Leaflet	1.28	1.03–1.59	1.12	0.029
Interaction b
Letter and leaflet	1.02	0.66–1.58	—	0.979
Participation (sent FOBT kit)	Odds ratio	95% CI	Rate ratioa	P-value
Intervention main effects b , c
Letter	1.17	0.93–1.47	1.06	0.186
Leaflet	1.23	0.98–1.56	1.08	0.073
GP signature	1.29	1.01–1.63	1.11	0.039
Interaction b
Letter and leaflet	0.91	0.58–1.44	—	0.697

Abbreviations: CI=confidence interval; FOBT=faecal occult blood test; ITT=intention to treat.

Estimated as OR/(1−P)+(OR × P) where OR is the odds ratio and P is the proportion of kits returned in those not receiving the intervention.

The reference category for each main effect was those not receiving the intervention.

Obtained for model without interaction; each intervention adjusted for the other.

In the intention-to-treat analysis, the logistic regression model indicated that there was no significant effect on the kit return-rates by age (P=0.77), GP's practice (P=0.66), gender (P=0.20), GP's signature (P=0.16) or previous invitation to screening (P=0.56). Both the ‘Leaflet’ (P=0.029) and ‘GP's letter’ (P=0.038) was significantly associated with an increase in participation. In the per-protocol analysis, the insertion of an electronic GP's signature on the endorsement letter was associated with increased participation (P=0.039), and again there was no significant association with other covariates.

The adjusted odds ratios associated with each intervention are also reported in Table 3. On the basis of intention-to-treat, the adjusted rate ratios calculated from these odds ratios were 1.11 for the GP's letter (P=0.038) and 1.12 for the enhanced leaflet (P=0.029), and slightly lower (1.06 and 1.08, respectively) when the analysis was restricted to those who received the intervention (per-protocol analysis); reflecting the fact that the return rate in the no-treatment groups (61% and 62%, respectively) is higher when patients who opt out of screening are excluded from the denominator (number of people receiving the FOBT kit, rather than total number invited).

Discussion

Main findings

A letter of endorsement from the GP and a how-to-do-it procedural leaflet sent with the FOBT, each appears able to achieve a small, but important, increase in participation in the National Bowel Cancer Screening Programme. The effects appear to act independently of each other and would be additive in practice. The non-randomised comparison of the effect of a GP's signature also suggests that it is better if the endorsement letter is signed by the patient's own GP rather than by the more impersonal ‘on behalf of the practice’.

Relation to previous findings

The findings are consistent with previous research from other countries, which have demonstrated that personalised invitations improve screening test-return rates (Brawarsky ; Segnan ; Cole ; Senore ; Zajac ) and emphasise the important role that invitation materials can have on participation in screening (Steele ). Furthermore, the per-protocol analysis revealed this effect is mediated by the inclusion of a GP's signature on the endorsement letter. Similar to previous research (Miller ; Stokamer ), the intervention would seem to enhance participants’ perceived self-efficacy to complete the FOBT kit.

The additive effect observed in the trial may reflect the influence that each separate intervention had on two distinct groups of potential non-participants. For example, Senore found those with a higher education, generally based their decision to participate in screening after reading the information materials, whereas people with a lower education tended not to read the information materials and rely on their GP's advice instead. Alternatively, people with limited health literacy (and associated lower education status) report more barriers to complete FOBT testing (Peterson ). The brief enhanced procedural leaflet may have been sufficient for people with limited health literacy to address these barriers and engage in bowel cancer screening. It is, therefore, plausible that the GP's endorsement letter was influential in those who preferred receiving health advice from their GP, whereas the procedural leaflet helped people overcome perceived barriers to complete the FOBT kits. Further research directly evaluating process variables such as health literacy, the effects of gain-framing vs loss-framing, and the importance of GP's recommendations for bowel cancer screening are required.

Limitations

The main design weakness was the potential for patients to opt-out of the trial after randomisation, but before the intervention was delivered. This led to a slight imbalance in the intervention groups, but there is no suggestion that this had an important effect on the results. Similarly, although the power of the study to demonstrate interaction was limited, there was no suggestion of important interaction between the two interventions. The lack of an individual GP effect on test-return in the regression analyses suggests the findings are likely to be generalisable across practices. The relatively low GP participation rate and the inability for a number of GP practices to provide an electronic signature may have implications for the widespread adoption of the endorsement letter for bowel cancer screening.

A further concern was the possibility of a ‘priming effect’ as participants were told about the trial and informed they would receive an invitation to screening 6 weeks in advance. In one previous study, participation in screening was increased when people received an advanced notification letter before their invitation (Cole ). However, as the usual care group in our trial did not participate in screening at a rate much above the national average for participation in the National Programme, priming did not appear to have an important effect.

Implications and conclusion

The results emphasise that a minor amendment to the way screening is conducted can have important effects on uptake rates. Adding a GP's letter and a more explicit instruction leaflet appears able to increase participation by at ∼10% (potentially providing a 20% relative improvement in the current participation rate). However, as less than half the GP practices recruited to this trial provided an electronic signature to the screening hub on request, this suggests there is a lack of GP engagement with the programme. Including bowel cancer screening uptake as a QOF indicator may provide the necessary incentive to remedy this lack of engagement.

Given the low cost of including a GP's endorsement letter and more explicit how-to-do-it leaflet, it is very likely that the two interventions would be cost-effective. Sending the initial invitation from GPs might also reduce the level of initial opt-out, but the potential gain in participation would have to be weighed against the increased cost and administrative complexity.