Swedish registries are promising but of limited value without validation

Colonoscopies were previously only done in those with symptoms or signs of disease. Now they are also being performed as screening for colorectal cancer in asymptomatic individuals in many countries. Both patients and screenies need to know how safe colonoscopy is. If complications are frequent then that can undermine screening programs because of a poor cost-benefit ratio.

Sweden is famous for its extensive registries. They could be very useful and have been used in many publications in high-ranked journals. They could help us determine the safety of colonoscopy. However, even though the registries are good and extensive it does not mean that they cannot be improved and need validation.

The present issue contains the article “A register-based study: adverse events in colonoscopies performed in Sweden 2001–2013”. It spans a whole 10-year period for all of Sweden for almost 600,000 colonoscopies in a population of 9.5 million and records the frequency of major adverse events like bleeding, perforation, splenic injuries, and death. Among their findings were that complications were more frequent if associated with polypectomy. This is as expected and consistent with other studies [1]. They were able to extract data from various other national registries. Such an almost complete data collection could in theory be very valuable. However, the authors state in a letter to this journal that “in national Swedish registries only basic data are present, which limits the information one can do research on. Thus, it is not possible to get more detailed data about the procedures. Furthermore, following the linkage between the different registries, which is performed by the Swedish Board of Health and Welfare, all identification of separate patients is withdrawn from the data.”

Why is all identification of individual patients removed? Two reasons are plausible: (1) To maintain patient confidentiality and (2) To protect physicians who contribute to the registries. Complications need to be reported but physicians may not want a large audience to know that it was their patient(s).

What can be done? Reading almost 600,000 patient charts would not be possible. However, adverse events were noted for only a small fraction of the procedures, less than 0.3%. How could researchers access the individual charts while maintaining confidentiality? A small group in each health region in Sweden could be allowed to respond to a request from researchers about detailed data on complications. They could find key reports like colonoscopy, radiology findings, progress notes, and discharge summaries and deidentify them. Then they could send them to the researchers who would not know what hospitals or clinics they came from.

To give an example on how this would advance medical knowledge and improve patient care one can look at the current article:

1. We do not know the specialty of the endoscopists in the current paper. Gastroenterologists tend to have better outcomes than others [2–4]. The authors state in a letter to the editor that in Sweden endoscopy is performed by surgeons, gastroenterologists, or in rare cases by radiologists. Which of them were responsible for the complications? That could tell us who need more training.

2. Bleeding was associated with 0.17% (1:605) of procedures. We need to know if the bleeding was caused by a colonoscopy or if the colonoscopy was done because of bleeding. If it is the latter then colonoscopy is being done to avoid a possible operation. One can accept a higher risk profile in such cases. If the bleeding was caused by a colonoscopy done for screening then it is more serious. The authors only relied on codes to state if the bleeding was caused by colonoscopy or done because of bleeding but it would be better to verify that by reviewing reports from the patients’ charts. We need to know how severe the bleeding was: was it a trivial thing occurring during colonoscopy that was easily controlled by hemostatic therapy; slight bleeding after the procedure; was it easily stopped by a repeat colonoscopy; or was it a severe episode requiring blood transfusions, surgery or even causing death? Only the last instance would be considered so serious that it would be an argument against screening colonoscopies. An admission to the hospital for observation is not necessarily a serious event if no intervention is needed. Did the bleeding happen after removal of a polyp that would have needed surgery had one not tried colonoscopy? Were the bleeders on anticoagulants: aspirin alone is rather safe but that is not the case with warfarin, thienopyridines, or NOACs (novel oral anticoagulants) [5]. Were they discontinued some days before the procedure if the patients were taken them? Were the patients on alternative medicines, which may or may not contribute to bleeding and needs more research [6,7]? Examples of possible concern are garlic, gingko, cranberry juice, vitamin E, ginseng, fish oil, and flaxseed oil. It would probably require a prospective study to get information about alternative medicines so getting more access to the registries might not help.

3. The perforation rate was 0.11% (1:890). The risk of perforation increased with age like bleeding did. Was the reason that older people had more polyps and thus more polypectomies? We should know about the number and size of polyps that were associated with bleeding or perforation. What kind of equipment was used: were the snares used for polypectomy warm (with cautery) or cold (no current used)? It is believed that the use of cold snares decreases the risk of perforation and possibly bleeding [8,9].

4. If a polyp was removed by a simple biopsy, then it is very unlikely to lead to a complication. If older people have more perforations despite controlling for these factors then screening colonoscopies may not be cost-effective after a certain age. How were the perforations treated: by observation, by primary repair, resection or fecal diversion? Was the operation minimal with laparoscopy or was it open and more extensive?

5. Splenic injury. This is an under-recognized complication of colonoscopy [10]. This is the most interesting finding of the study and a large database is needed to learn more about it. The authors state that “There was no way of establishing the cause of the splenic injury from the data provided in the registers, …” Therefore, they had to estimate the cause of the injury indirectly. The rate was very low, 1:20,000 – about one-third of them had a splenectomy. It would be very useful to know the details of this occurrence. Was this a splenic rupture requiring splenectomy or just a subcapsular bleeding, which healed by itself? What were the circumstances in the 11 patients who needed a splenectomy? Was it a difficult colonoscopy requiring excessive force? Had the patients had a previous abdominal operation [11]? Were the scopes used old and stiff or new and flexible? Were the patients so heavily sedated that they could not report pain? Were the patients frail and malnourished? Was a scope guide used? Was room air, CO2 or just water used for insufflation? Was the injury treated by embolization, minimal surgery or an open operation?

6. Was a highly invasive procedure being performed when a complication happened, e.g. placement of a colonic stent for an obstructing cancer? It is associated with a rather high perforation rate. Was mortality due to a procedure or due to natural causes? Example: mortality is high after PEG procedures but hardly ever is it due to the procedure. Very sick people get PEGs and that makes dying more likely.

7. Some endoscopists who do not like to hurt people wish that they had access to propofol anesthesia for all endoscopies. The current study showed that anesthesia was associated with a higher bleeding and perforation rate. The kind of anesthesia used was not known but the authors state that propofol is usually used. If anesthesia was given to a large number of people, then the results would be interpretable. Only 1.6% of the colonoscopies in Sweden got anesthesia. If anesthesia was only given to failed colonoscopies due to difficult colons or to frail multimorbid patients, then a causal link cannot be established.

8. The authors did unfortunately not have information about the frequency of postpolypectomy syndrome. This is not a serious complication but can lead to unnecessary emergency surgery.

If Swedish researchers will be able to get access to individual patient charts where a complication occurs, then they can greatly help the whole world improve colonoscopy quality. If that would be possible we might be able to look to Sweden for guidance. However, as the results are presented there are too many uncertainties, which makes interpretation very difficult and the results are of limited value.