Cancer survival rates are heavily dependent on stage at presentation. In turn, stage at presentation is influenced by delay in treatment following first symptoms. The symptom to treatment delay can be divided into two major stages – from first symptom to presentation at the health service, and from presentation to treatment.
Measuring these delays prospectively is difficult or impossible since a study population cannot be identified at the time of symptom recognition. Therefore, the literature on delay is based on retrospective studies, where people with cancer are interviewed to reconstruct their journey from first symptom to diagnosis (or treatment). Guidelines for the conduct of these studies have been promulgated through the Aarhus checklist.[1]
Memory is, of course, fallible. It is, therefore, inherent in the above retrospective method that there will be (considerable) measurement error. Recall can be improved by asking the responder to relate medical events to personal or public events to which a firm date can be ascribed – a wedding or religious holiday, for example. The impacts of measurement error can also be mitigated by selecting large samples. The purpose of this article, however, is to draw attention to another problem: that recall may not just vary at random over the recall period, but may vary systematically. That is to say, there may be a systematic tendency to over- or under-estimate time differences as they recede from the present time. Indeed, there is evidence for such a systematic bias in perception. Memory is ‘telescopic’, such that more distant events appear more recent than the actual occurrence, relative to more recent events.[2]
Previous research into median time delays between events finds that median delay periods from first presentation to treatment are longer than the delay periods from first symptom to presentation.[3] [4] However, if more distant events are perceived, relative to more recent events, as occurring later than they actually occurred, then the above findings might exaggerate the duration of the presentation to treatment epoch, relative to the symptom to presentation epoch.
Is it possible, in the absence of independent objective observations, to design a method to empirically measure (and therefore correct for) ‘telescoping’, with respect to cancer delay interval periods? We propose a method that might support the above conjuncture – that more distant events on the cancer pathway appear more recent on average. In any database of time intervals collected using the Aarhus retrospective method, there will be people who had a short and long ‘second delay’, i.e. the delay from presentation to treatment. Then, under the conjuncture, the first delay should be less ‘telescoped’ in those with short second delays than in those with long second delays. This should show up as a difference in the difference between first and second delays according to the duration of the first delay. We therefore propose to examine this difference in the difference between second and first delays. We expect to find that shorter second delays are associated with a difference in the differences. The greater the inverse correlation between these variables, the greater the degree of telescoping in cancer delay research.
There is, of course, an assumption behind this interpretation. The proposed method assumes that there is no correlation between the lengths of the two delay periods – i.e., people with longer time second delays do not tend to have true longer (or shorter) first delays. However, this assumption might not hold. For example, deprived people with low health literacy may both fail to recognise symptoms and then be treated differently in the service. It would, however, be possible to adjust for this bias in data-sets that recorded the necessary variables. Also, if there is no evidence of telescoping, then we can be somewhat reassured regarding retrospective comparisons of different journey stages on the pathway from first symptom to treatment.
— Richard Lilford, ARC WM Director
References:
- Weller D, Vedsted P, Rubin G, et al. The Aarhus statement: improving design and reporting of studies on early cancer diagnosis. Br J Cancer. 2012; 106(7): 1262-7.
- Rubin DC, & Baddeley AD. Telescoping is not time compression: A model. Mem Cogn. 1989; 17: 653-61.
- Fayehun O, Apenteng P, Umar UA, et al. Diagnosis of cancer in the South and North of Nigeria: duration and causes of delay. BMC Health Serv Res. 2025; 25(1): 738.
- Makene FS, Ngilangwa R, Santos C, et al. Patients’ pathways to cancer care in Tanzania: documenting and addressing social inequalities in reaching a cancer diagnosis. BMC Health Serv Res. 2022; 22(1): 189.