The study established the surgical volume and incidence of in-hospital complications of RC for bladder cancer in Armenia from 2005 to 2012, and assessed the predictors of these complications. To our knowledge this was the first study conducted in Armenia among patients with RC, with reasonable generalizability to the entire population of patients in Armenia for the specified time period. We found that 273 patients had RC in eight Armenian hospitals, and the incidence of postoperative complications and mortality was quite substantial.
Except age, the baseline profile of Armenian patients was not very different from those in other countries (Konety et al. 2006; Stimson et al. 2010; Kim et al. 2012; van Hemelrijck et al. 2013; Aziz et al. 2014). At the time of surgery Armenian patients were 58.5 years old on average. The average age of 679 RC patients in 18 European centers in 2011, for example, was 70 years (Aziz et al. 2014), and the average age of 50,625 RC patients in the US Nationwide Inpatient Sample from 2001 to 2008 was 68.4 years (Kim et al. 2012). This indicates that bladder cancer may affect Armenian patients at younger ages. Armenian men also get exposed to smoking, the major risk factor of bladder cancer, at quite a young age. In 2001, the average age of smoking initiation was 18.5 in Armenia men and 28.0 in women (Gilmore et al. 2004). In our study, about 74 % of patients reported being current smokers, and smoking continues to be a major public health problem in Armenia.
The observed in-hospital mortality rate in our sample was 4.8 %, somewhat higher than the mortality rates reported in other studies (Konety et al. 2006; Shabsigh et al. 2009; Kim et al. 2012; Maffezzini et al. 2008). However, it is hard to conclude if this is Han indication of a sicker Armenian surgical population or inadequate pre- or postoperative management. Furthermore, the proportion of patients developing at least one postoperative complication was 29 %, well within the range of reported complication rates in other studies that varied from 21.8 to 57 % (Konety et al. 2006).
The most common complications observed in our study were postoperative ileus (7.3 %), followed by wound infections (7.0 %), urinary tract infections (4.8 %), and wound dehiscence (3.3 %). The study in 2538 RC patients by Hollenbeck et al. similarly reported the top four most common complications as ileus (9.7 %), urinary tract infections (7.8 %), wound dehiscence (5.5 %), and wound infections (5.2 %) (Hollenbeck et al. 2005), though this study included a relatively older patient population (65.5 years old) (Hollenbeck et al. 2005). Using a reporting methodology by Shabsigh et al. (Shabsigh et al. 2009), we found that wound complications had the highest incidence in our study population affecting 10 % of patients, followed by GI complications (9 %), and infectious complications (7 %). Shabsigh et al. also reported that these three complication categories were the most frequently occurring in a hospital in New York; however, GI complications were observed in 29 %, infectious in 25 % and wound complications in 15 % of patients (Shabsigh et al. 2009). In the study by Takada et al. among Japanese RC patients the highest incidence was observed for infectious complications (30 %), followed by GI (26 %) and wound (21 %) complications (Takada et al. 2012). These two studies, however, captured 90-day complications (Shabsigh et al. 2009; Takada et al. 2012) while we recorded only in-hospital complications.
Similar to other past studies, we found that postoperative complications significantly increase the hospital LOS of patients (Shabsigh et al. 2009; Konety and Allareddy 2007). Shabsigh et al. reported that LOS is increasing with the increase of the severity of complications (Shabsigh et al. 2009). Another study further reported that the addition of each complication significantly increased the risk of in-hospital mortality, LOS and costs (Konety and Allareddy 2007). The total hospital LOS in our patient sample was higher than in other studies, which could be explained by differences in healthcare systems (Shabsigh et al. 2009; Konety and Allareddy 2007; Maffezzini et al. 2008). Since all centers that conduct RC in Armenia are located in the capital city, patients from remote areas stay longer until they recover and are fit to travel. Furthermore, in Armenia medical costs are mostly covered by patients and the major cost driver is the cost of the surgery and not the length of hospital stay after the procedure.
Not all risk factors identified by our literature review have been proven to be predictive of RC complications in Armenia. For example, several studies reported an increased risk of postoperative complications with increasing age (Hollenbeck et al. 2005; Konety et al. 2006; Kim et al. 2012; Gore et al. 2010). This was not true for the Armenian population, most probably because of a relatively younger patient population. The risk of RC complications by gender has been controversial in the literature with some studies reporting an increased risk for females (Shabsigh et al. 2009; Gore et al. 2010), increased risk for males (Kim et al. 2012; Takada et al. 2012), or no risk difference (Konety et al. 2006; Lee et al. 2004) as in our sample.
The evidence is controversial regarding the issue of surgical volume and patient outcomes. While some studies supported the positive impact of higher surgical caseload and hospital volume on RC outcomes (Kim et al. 2012; McCabe et al. 2007), others showed no difference (Konety et al. 2006; Takada et al. 2012; Gore et al. 2010). There is no unified approach to define high-, medium- or low-volume hospital categories for RC. The study by Kim et al. in 1173 US hospitals found that patients in high volume hospitals (defined as ≥5RC/year) had 33 % less chance of in-hospital complications and 40 % less chance of in-hospital mortality than patients treated in low volume hospitals (defined as <1.5 RC/year) (Kim et al. 2012). Another study that defined >10 RC/year as high, 5–10 as medium and ≤5 as low-volume hospitals, found no association between volume and risk of complications (Takada et al. 2012). A recent US study reported that surgeons who perform ≥7 RC per year had 44 % lower odds of major complications at 90 days than those performing 1 RC per year (Leow et al. 2015). We, however, did not capture per surgeon surgical volume. In our sample, the majority of patients underwent surgery in one of the hospitals, hospital A, where the RC volume was on average 22/year. The RC volume was 6, 4 and <1 per year in hospital B, C, and in ‘other’ category. The odds of complications were almost twice higher in patients in hospital C than in hospital A.
ASA scores above 2 have been shown to increase the risk of RC complications (Hollenbeck et al. 2005; Novotny et al. 2011; Shabsigh et al. 2009; Takada et al. 2012). In our study, ASA score was not an independent predictor of complications. The comorbidities that have been shown to increase the risk of complications are also variable between published studies (Hollenbeck et al. 2005; Novotny et al. 2011; Kim et al. 2012; Takada et al. 2012; Gore et al. 2010). Such differences can be explained by differences in the prevalence in the index population, in over- or under-reporting, and in peri-operative management. In our sample, for example, although about 74 % of patients (with complete information) reported current smoking history, the recorded prevalence of COPD was only 7 %. In our study IHD was one of the strongest predictors, increasing the odds of developing in-hospital complications by three times. Most past studies either did not capture IHD as a separate patient baseline factor (Konety et al. 2006; Kim et al. 2012), or captured other cardiovascular disease entities (Hollenbeck et al. 2005; Gore et al. 2010). Considering our finding, we think that future studies should report the prevalence of IHD in this population and further investigate its effect on outcomes.
With detailed review of medical records we captured several pre-operative factors that are not available through administrative databases. In bivariate analysis, we found that patients without complications had significantly higher glucose levels than those who developed in-hospital complications. In multivariable analysis, higher glucose levels were associated with lower odds of complications. Our review of past studies did not find any study in RC patient population that evaluated baseline glucose levels and their effect on surgical outcomes. Considering the retrospective nature of the study, it is impossible to conclude if the low glucose levels are a result of advance cancer stages and derangements in carbohydrate metabolism. We were also unable to fully investigate the perioperative management of patients with low or high glucose levels.
Based on our findings, any intraoperative and/or postoperative transfusion of fresh frozen plasma or red blood cells was another independent predictor, almost doubling the odds of post-operative complications. Intraoperative blood transfusion was a significant predictor of in-hospital complications in a study of 2538 RC patients by Hollenbeck et al. (OR = 1.4, 95 % CI 1.2–1.7) (Hollenbeck et al. 2005). While several past studies did not report patient blood loss or transfusion of blood products (most likely because of the use of administrative databases that generally lack such data), transfusion of blood products has been shown to increase the risk of adverse surgical outcomes in cardiac surgery (Horvath et al. 2013), knee arthroplasty (Hart et al. 2014), and neurosurgery (Rolston et al. 2014).
Potential limitations of our study warrant discussion. There is always some degree of variability in medical records and data capture methods between different hospitals. To minimize this, we used a standardized data collection tool. Smoking history, which was a variable of high interest, was not recorded in almost one third of medical records, and had to be excluded from adjusted analysis. As the information was collected retrospectively, it was impossible to accurately evaluate the amount of blood loss, the appropriateness of indications for transfusions, and reasons and management of low glucose levels. At the same time, one of the major strengths of our study was the use of detailed clinical information from medical records that was somewhat limited in other studies that used administrative databases (Hollenbeck et al. 2005; Konety et al. 2006; Kim et al. 2012; Gore et al. 2010). This allowed us to generate new hypotheses regarding the role of glucose levels and IHD in RC complications. We also used a relatively recent cohort of patients, and obtained a national sample of RC patients. The results of our study can be used for comparisons with studies from other low or middle income countries.