4.2. Decisión de Compra
4.2.2 Tipos de consumidores
4.2.2.3 Según el uso del producto
I found that older patients were more likely to be diagnosed with colon cancers. This proximal shift in CRC site with increasing age has been described in other studies(254,260). This was important as compared with distal CRC proximal CRC tends to present with more advanced disease and has worse outcomes (264). This may be another reason why there are more emergency presentations and worse outcomes in older patients and another reason why flexible sigmoidoscopy alone is not an acceptable way to investigate symptoms in older patients.
Unsurprisingly I found that older patients (over 74 years) were significantly more likely to have one or more co-morbidities than younger patients. Increasing co-morbidity has previously been associated with older patients diagnosed with CRC(254,255). This was unsurprising, as age is commonly perceived to be a surrogate marker for co-morbidity. It needs to be remembered that older patients with CRC are a heterogeneous group that include patients with an otherwise excellent health status. If age is solely used instead of co-morbidity then there is a risk of undertreating older patients, because of biased clinician preferences. I have shown that increasing age is associated with worse outcomes. This is reported in other studies, especially when the patient is unmarried and living alone(256).
I found male patients were diagnosed on average earlier than females. This finding has been replicated in other studies (257,258,261). This has a number of potential implications; firstly it means any data looking to see if outcomes differed by gender would need to be matched for age and secondly that a higher proportion of males are in the BCSP age range. This will be discussed in chapter 4 but could create bias in the potential effect of screening irrespective of different uptake rates of screening based on
gender. Along with gender variation in screening test sensitivity, the age distributions for males and females may be one reason for the apparent superiority of screening for male patients(271). Some authors have even suggested this is should mean CRC screening should be offered at different age ranges depending on gender (261).
In my cohort male patients have a higher rate of rectal cancer than females, with proximal CRC being found more commonly in women. It has been described previously how the proportion of CRC cases that are men increases steadily on moving from the caecum to the rectum (254,257,262,263,272,273). This maybe another reason for the apparent better outcomes for men. Finally, it has been shown that proximal CRC becomes more common as women get older but not for men(272). This again has implications for CRC screening (as screening for proximal cancer is less accurate) and especially for the roll out of one-off flexible sigmoidoscopy screening, which is less effective for older women.
Cardiovascular diseases, hypertension and previous cancers are the main co-morbidities found in patients with CRC. Patients with co-morbidity were more likely to have a proximal CRC in the study cohort(264). It may be that some common co-morbidities such as diabetes and cardiovascular disease are associated with common risk factors or pathophysiological pathways that increase the risk of proximal cancer. These might include factors such as dietary or the presence of micro-vascular disease, which themselves increase the risk of cancer at a proximal site (274).
My cohort also demonstrated a relationship between increasing co-morbidity and deprivation, which has been previously described(265). I also found deprivation was associated with a CRC diagnosis at a younger age. A search of the literature failed to identify any other study from the UK has shown that deprived patients are diagnosed at an earlier age(235). If patients from deprived areas were more likely to be diagnosed at a younger age it is especially important that these patients are targeted for preventative strategies such as the screening programme because this is a difficult group to reach with health promotion strategies and they have a lower uptake of screening (171).
Looking at the associations between patient characteristics and outcomes the main route to long-term survival is through a successful surgical resection of the cancer. If I could identify unexplained low rates of resection for a particular patient characteristic then this opens up the possibility of preventing under- treatment. If the rationale for the under-treatment could not be backed up by clinical evidence of inferior outcomes then the argument could be made that there was some bias with in the healthcare system preventing access to surgery for these patients.
In my cohort I found that increasing age was associated with a reduced crude surgical resection rate and higher one-year mortality, which is backed up by other studies(256). This main reason for this was that the stage of cancer is higher in older patients(8,275). However over the last few decades survival differences based on age has reduced substantially. The predominant reason for this is thought to be the increased resection rates among the elderly(276). Unfortunately, my patient cohort does not have a
record of the patient’s cancer stage so I cannot directly challenge whether there is under treatment based on age bias or simply that older patients had more advanced disease that would not benefit from surgical resection.
In my study cohort there was no significant difference in the resection rate based on gender. This was interesting because while most other studies have reported rates approximately equal between the genders (258,272), a small number of studies have reported more women undergoing an attempt at curative surgery(257). The crude one-year mortality rate was higher among females, but this was likely to be a confounding effect caused by age.
Increasing co-morbidity was associated with reduced surgical resection and higher one-year mortality in my study cohort. In other studies, increased co-morbidity was negatively associated with short-term survival(277). This association was especially strong for older male patients(274). I showed co- morbidity was strongly associated with poor outcomes. This was in agreement with the “competing demands” model, which argues that co-morbidity is associated with delayed diagnosis and also other work showing poorer survival in patients with co-morbidity(250).
Increasing deprivation was associated with reduced surgical resection and higher one-year mortality. This was supported by other studies which found that the most socioeconomically deprived patients had higher mortality rates, largely explained by excess early mortality especially in the first month after diagnosis(251).
I found that a proximal cancer site was associated with reduced surgical resection and higher one-year mortality. Other studies support this finding in particular that outcomes tend to be better in patients with rectal cancer(223).
There are some weaknesses in my methodological approach. Several of these are common to all HES related studies and the most significant of these are associated with missing clinical information. Important omissions from the database include, as already mentioned, a date of diagnosis. Also missing are the stage of cancer, which is of course of great prognostic impact, and patient ethnicity. There will always be some problems associated with the accuracy of clinical recordkeeping and transposition into codes by clinical coders. Other problems include under-reporting of co-morbidity in some hospitals and maybe missing codes for some day case procedures such as colonoscopy.
I acknowledge that the codes included to define the REL1 admission were somewhat subjective, however I mitigated this by using an expert panel to select those included. Only a separate analysis of the patients’ medical records at the time of the REL1 admission could determine for certain if all the included codes were truly relevant.
Finally, while I was able to identify some risk factors, such as increasing age, to explain variation in outcomes. I was not able to exclude all potential cofounding factors. Therefore I could not directly
apply causality to the outcomes. I will however look into the causes for variations in outcome in subsequent chapters and especially the impact of the introduction of the BCSP.
In conclusion, I believe my methodological approach is superior to traditional analytical approaches that focus solely on episodes coded with a cancer diagnosis. This was because it contained more patients with incident CRC, including patients with advanced CRC and those not suitable for surgical resection. These cases are possibly excluded from other cancer registry databases. Furthermore for each patient, I identified more of the relevant episodes of care and therefore increased the completeness of the patients journeys compared to other studies. In particular, I believe I have captured a more accurate start date for the patient’s journey, by identifying the first relevant admission (REL1). Almost a quarter of patients had an earlier relevant admission and this process identified a higher rate of emergency presentation than was previously believed to be the case. This is very important as it creates a target to focus resources on, which will hopefully lead to improvements in CRC outcomes.
I then validated my cohort against external data sources to show already established associations and also describe some new findings. My study cohort is therefore an accurate representation of CRC across England, and forms my main data source in the subsequent chapters of this thesis.