Tuesday April 12 2011
The study looked at all bowel cancer operations in England
“The survival rate for bowel cancer surgery varies widely between hospitals,” BBC News has reported. Several other news sources have also reported on the outcomes of colon cancer surgery, which was examined by a major study published today.
The research took a comprehensive look at numerous factors linked to patient death within 30 days of bowel cancer surgery. It looked at the records of all people who had undergone the procedure in England between 1998 and 2006. The study showed a wealth of factors that influenced short-term survival rates, including age, type of colon cancer, patient income and whether people had other medical conditions.
Overall, 6.7% of patients died within 30 days of surgery, with those aged over 80 or with other serious illnesses having the highest risk of death. However, the research did highlight that mortality rates in Scandinavia and Canada were lower, and that some hospital trusts had performance below the national average. Crucially, the research has identified a number of areas where policies could be introduced to lower the risks associated with the surgery, and this will hopefully lead to improvements in survival rates.
Where did the story come from?
The study was carried out by researchers from the University of Leeds and was funded by Cancer Research UK. It was published in the peer-reviewed medical journal Gut.
The study was covered accurately by the newspapers.
What kind of research was this?
This was a retrospective, cross-sectional, population-based study of data from the National Cancer Data Repository (NCDR).
The NCDR is a database provided by the National Cancer Intelligence Network (NCIN), which is a database that links a range of sources of data about cancer and factors related to cancer. For example, it connects the detailed data on tumour incidence and outcome featured in the cancer registry with Hospital Episode Statistics (HES data), which record detailed treatment information but limited detail on the characteristics of tumours. The NCDR allows the treatments and outcomes for each NHS cancer patient in England to be tracked.
The researchers wanted to assess what outcomes occurred for people who had surgery for colorectal cancer across the population. They particularly wanted to monitor the 30-day postoperative mortality rates and to compare the performance of NHS hospital trusts in England.
What did the research involve?
The NCDR consists of pooled data from eight population-based cancer registries covering England, where each individual’s data was linked to their Hospital Episodes Statistics (HES) data (the treatment they received in hospital). To maintain their anonymity the individuals were identified using their NHS number, date of birth, postcode at diagnosis and sex. The individuals had all been in hospital between April 1997 and June 2007 and had HES data with a diagnostic code for cancer.
The researchers extracted data for all individuals who had undergone major surgery for a primary colorectal cancer diagnosed between January 1 1998 and December 31 2006. Data were extracted on their age, sex, how advanced the cancer was, date of diagnosis, date of death (where relevant) and the treatment they received (the type of surgery and which area of the colon had been removed). The researchers also extracted data on which hospital trust the patient had attended for their surgery (or for their first or most extensive surgery if they had multiple operations). They looked at whether the individuals had any other conditions besides their cancer.
The researchers calculated the percentage of patients who had died within 30 days of their operation for each year of diagnosis, age group, sex, stage of tumour at diagnosis, likely income (based on postcode), other diseases and the hospital trust in which they had their surgery.
There were 160,920 cases of colorectal cancer. For 24,434 (15.2%) of individuals, data on the tumour stage at diagnosis were missing, and 404 (0.25%) did not have postcode information, preventing the researchers from estimating their income. However, the researchers made an estimate of these missing values based on statistical calculations.
For their analysis the researchers used a statistical technique called logistic regression to see how various factors were associated with 30-day postoperative mortality. Mortality rates were compared between hospital trusts, taking into account other factors that they had determined to affect these rates, such as the risk of the procedure itself in different patient populations.
What were the basic results?
The 160,920 people diagnosed with colorectal cancer between 1998 and 2006 were treated by 150 different hospital teams across 28 cancer networks. Of these people, 10,704 (6.7%) died within 30 days of surgery. Looking at the death rate over time the researchers calculated that 30-day mortality had declined from 6.9% in 1998 to 5.9% in 2006.
The researchers performed a number of analyses and comparisons, presented below with confidence intervals for their results listed in square brackets. Across England the researchers found that:
- Women were significantly less likely to die postoperatively than men [6.5% (95% CI 6.1% to 6.9%) vs. 6.8% (95% CI 6.4% to 7.3%)].
- Postoperative mortality was significantly associated with age: 1.2% of patients aged less than 50 died within 30 days of surgery compared with 15.0% of people aged over 80 [(95% CI 1.0% to 1.4%) vs. (95% CI 14.1% to 15.9%)].
- People who had an advanced tumour stage (Dukes’ D tumours, where there is spread to other organs in the body) had a 9.9% risk of death compared with a 4.2% risk of death for people with less advanced tumour stage (Dukes' A, the most localised tumour which has not spread beyond the inner lining of the bowel) [(95% CI 9.3% to 10.6%) vs. (95% CI 3.7% to 4.7%)].
- In the most affluent category 5.7% died in this period compared with 7.8% in the poorer areas [(95% CI 5.3% to 6.1%) vs (95% CI 7.2% to 8.4%)].
- There was a 24.3% risk of postoperative death in people who had other medical conditions that themselves carried a high risk of causing death (Charlson comorbidity score greater than 3). In contrast, there was only a 5.4% risk of death in those without comorbid conditions (Charlson score of 0) [(95% CI 22.0% to 26.5%) vs. (95% CI 5.0% to 5.7%)].
- Tumour location within the bowel affected mortality rates: patients with tumours in the colon had higher postoperative mortality than those with tumours of the rectum [7.7% (95% CI 7.3% to 8.2%) vs 4.6% (95% CI 4.3% to 5.0%)].
- Operative urgency was important: 14.9% of patients receiving emergency operations died within 30 days of surgery compared with only 5.8% of those operated on electively (when the date of operation is chosen between the surgeon and the patient) [(95% CI 14.2% to 15.7%) vs. (95% CI 5.4% to 6.2%)].
The researchers then compared the 30-day mortality rates between the hospital trusts in two analyses. One looked at operations on patients diagnosed between 1998 and 2002 and the other looked at cases between 2003 and 2006. In both of these analyses they adjusted for the risk factors of age, sex, year of diagnosis, cancer site, income/deprivation, tumour stage, other conditions (comorbidities) and type of colon cancer/surgery. They took the national average and determined the number of hospital trusts that were statistically better or worse than this. To do this they set confidence intervals of 99.8% around the national average. This gave a range of rates the researchers could be 99.8% sure were the same as the national average. Mortality rates outside the top of this range were considered worse than the national average and mortality rates below the bottom of this range were considered better than the national average.
For patients diagnosed between 1998 and 2002, eight trusts were outside of the 99.8% control confidence limits and performed worse than the national average, while five performed better.
For patients diagnosed between 2003 and 2006, five trusts were outside the 99.8% confidence limits and performed worse than the national average, while three performed better. Across the two time periods three trusts were worse than the national average, indicating consistently worse 30-day postoperative mortality, while one trust performed consistently better.
How did the researchers interpret the results?
The researchers say that their study is the first to provide a comprehensive national perspective on the 30-day operative mortality associated with colorectal cancer surgery. They say that the 6.7% mortality rate is higher than that previously reported for the UK; however, they report that some previous audits have been voluntary so all cases may not have been included in these analyses. The researchers say that 30-day postoperative mortality rates from similar population-based studies in Scandinavia, Canada and the USA ranged from 2.7% to 5.7% and, while there will be differences in how studies are conducted, these rates are lower than that of the UK. They add that further understanding of the risks is needed to minimise these differences and minimise premature deaths in the UK.
This was a well-conducted study that highlighted risk factors associated with 30-day mortality following surgery for colorectal cancer, which the researchers have assessed using comprehensive data compiled by the National Cancer Intelligence Network.
Since this research looked at all national cases of colorectal cancer, the study provides findings that can potentially identify areas where policy changes may improve outcomes. For example, the researchers said in their discussion that socioeconomic deprivation was associated with a higher mortality rate. They call for further evidence to determine whether or not this phenomenon was due to inequalities in care.
The researchers also address some potential limitations to their study. First, they say that the technical coding accuracy in databases has been questioned, but suggest that a recent study of colorectal cancer patients had found “excellent agreement in the information recorded in both datasets [that they had used] with regards to both treatment and outcomes”. They say that a second limitation may be that their database does not contain detailed information about every aspect of a patient or their care that could affect the risk of postoperative death, and so there could have been unmeasured factors that would affect the patients and, therefore, their results.
Also, this study did not look at the cause of death in the patients who had died. Why the risk factors led to a poorer prognosis will have to be further assessed in order to develop policies geared towards lowering death following surgery for colorectal cancer.