This study has shown that people with AML from more deprived socioeconomic classes are less likely to undergo bone marrow transplantation than their counter-parts from more advantaged social classes, even after adjusting for the presence of recorded co-existing disease.
The main strength of this study is the large size of the study population. We were able to study over 23 000 incident cases of AML in the UK using data derived from hospital records. Data also included the co-existing medical conditions of AML patients, which allowed us to adjust for co-morbidity. One drawback, however, was that our data on co-morbidity was based on hospital admission records for those conditions. This means that whilst we should have captured the most severe co-morbid illness we will have missed more minor disease. As a result it is possible that there is some residual confounding by co-morbid illness in this study and that some of the gradient in socioeconomic status and access to bone marrow transplantation is still due to co-morbid disease.
By using routinely collected and available data rather than questionnaires or interviews, we have eliminated any bias in the reporting of socioeconomic class and any social class bias in participation in the study. Bone marrow transplantation recording is also likely to be accurate in hospital records given the highly specialised nature of the procedure. It seems likely to us therefore that the validity of this information is good.
One potential weakness of this study is that we were unable to adjust for the cytogenetic risk group of our AML patients. Based on cytogenetics at presentation, AML patients are classified into good, intermediate, and adverse risk groups, each with very different long term outcomes[8]. Good risk patients in their first remission are not transplanted, whereas adverse risk patients are almost always transplanted (subject to fitness and donor availability). Any bias introduced by this in our results, however, is likely to apply across all social classes, unless patients from lower social classes are more likely to be in the good risk group than those from higher socioeconomic classes, for which there is no evidence. Although there is evidence that patients from lower social classes present later with disease symptoms in general, it is less likely that late presentation is an important factor in AML survival given the acute presentation of the disease and its relatively poor prognosis. For intermediate risk patients, however, it is possible that later presentation may have an impact on the treatment administered and its outcome.
The accuracy of social class classification is imperfect given that Townsend Score is not an individual measure of deprivation. This will have introduced a non-differential bias into our results, if any, i.e. both patients who had had a bone marrow transplant and those who had not will have been similarly affected. Such a bias will have moved odds ratios closer to '1'. It seems then that if we had been able to perfectly adjust for socio-economic deprivation, our results may have shown an even greater class bias.
The validity of co-morbidity recorded in HES data may also be imperfect. Any inaccuracies would, however, apply equally across social class strata and so is unlikely to have introduced bias into these results. Furthermore these results showed no difference in recorded comorbidity across the social classes. Residual confounding cannot be ruled out completely, however, since only comorbidity recorded in the hospital episode data have been taken into account. Other comorbidities not related to hospital admission or not recorded during the admission may have existed which would have resulted in incomplete adjustment for comorbidity.
To our knowledge no studies examining the association between bone marrow transplantation and socioeconomic class have previously been published. Studies have, however, examined the associations between social class and chemotherapy in a number of cancers. Several studies found that lower socioeconomic class predicted under-use of chemotherapy in colorectal cancer (CRC)[3, 4], breast cancer[2] and lung cancer[3]. Two North American studies found that low socioeconomic status was associated with under-use of adjuvant chemotherapy in both breast[2] and colorectal cancers[4], and postulated that this was in part due to a combination of poor access to care, financial barriers and physicians' assumptions and biases regarding patients from lower socioeconomic classes, such as the availability of social and monetary support, their expectations of treatment and their likely compliance with treatment, for example. A further study concluded that lower incomes, absent or limited insurance cover and poorer education reduced access to high-quality adjuvant chemotherapy, which in turn reduced survival in breast cancer[9].
In the UK, a Scottish study showed that patients from the poorest deprivation quintile were less likely to receive chemotherapy for lung cancer and colorectal cancer than the most advantaged patients after adjusting for age, tumour stage at diagnosis, health authority and distance from oncology centre[3]. Delay between referral and treatment was similar across all social classes and so did not explain the findings. Although this study did not adjust for comorbidity, another Scottish study which had done so also found poorer survival in colorectal cancer patients from the most deprived socioeconomic quintiles, in a study population which showed no correlation between socioeconomic deprivation and co-morbidity[10]. The findings of these studies, in the UK healthcare setting where access to treatment is equal and free, suggest that decision-making (by both physicians and patients) regarding chemotherapy may be influenced by non-clinical factors.