Report from the Investigation Commission appointed by Rikshospitalet – Radiumhospitalet mc and the University of Oslo January 18, 2006



Download 0.95 Mb.
Page16/32
Date conversion04.02.2017
Size0.95 Mb.
1   ...   12   13   14   15   16   17   18   19   ...   32

The Cancer Registry’s linkage


In order to have the collected patient data checked and supplemented, they were sent to the Cancer Registry for linkage. A letter of February 20, 1996 from Jon Sudbø to Chief Physician Frøydis Langmark at the Cancer Registry states:

“I refer to our telephone conversation some weeks ago in which we discussed the possibility for a review of the Cancer Registry’s data base as part of a retrospective prognostic study of premalignant changes in the oral mucosa. Regrettably, our data will arrive later than indicated to you. In the meantime, we have collected data relating to persons with the diagnosis oral dysplasia also from Haukeland hospital, and the collection and systematization of these data have taken some time.

As stated, this takes place at the Department of Pathology, section for image analyses, at the Norwegian Radium Hospital, and I am currently working with a project aiming at defining morphological characteristics at premalignant conditions (dysplasia) in oral mucosa. The study is funded over three years by the Norwegian Cancer Society. Supervisor for the project is Professor Albrecht Reith, MD, Department of Pathology, the Norwegian Radium Hospital.

Enclosed is a transcript of the relevant patients. This represents a collection of patients from the entire country, diagnosed from 1984 up to 1995. Personal data are given as far as they are known. I also enclose a disk with the relevant data. Could you forward this to the relevant person.”


The annexed list of patients comprises data from 226 patients. The Cancer Registry has referring letters and similar documentation related to the relevant patients. The Cancer Registry’s subsequent investigations in 2006 show that 63 out of 226 persons originated from the Odontology. This number of patients is then in accordance with the number in the aforementioned letter from Sudbø and Reith dated February 1, 1996. The remainder (163 out of 226) originates from Gade’s Institute (see below).

However, the Cancer Registry has pointed out that the persons who subsequently have been proven to originate from the Odontology did not have personal identification numbers, and that they for that reason were not linked. The Cancer Registry is certain that the only patients that were linked and returned to Sudbø were 160 persons, who all originated from Gade (3 persons were excluded, and of these two were duplicates and one had adenoma in the stomach). Thus Sudbø received linked data for 160 patients from the Cancer Registry on March 22, 1996.

Sudbø has maintained that he was not aware that the 63 patients from the Odontology were not linked. On the other hand, Sudbø states that with the help from Reith and a laboratory technician at the Radiumhospitalet, Ruth Puntervold, he got access to the patients' full personal identification numbers from the Population Registry, and thereby was able to run a link with Cancer Registry data. Puntervold denies that she has assisted Sudbø in this. Also Reith denies that he has helped Sudbø to provide information from the Population Registry or assisted him with linkage. The Commission thus finds it quite unlikely that the patients from the Odontology had person identification numbers and that they were linked with the Cancer Registry data. In light of the information from both Puntervold and the Cancer Registry on this point, the Commission finds no grounds on which to accept Sudbø’s allegations regarding the actual facts.

The cause of death registry has not registered any inquiries from Reith or Sudbø. At that time, Sudbø did not have a clinical position at the Radiumhospitalet, and thus not a free access to the hospital’s case notes. Consequently, the Commission finds it to be not very likely that inadequate sets of data were supplemented in this way.

The Commission thus finds that Sudbø probably had access to human biological material and referring letter information on insufficient patient data (patient data not linked with Cancer Registry data in order to exclude coincidental and previous cancer) from 63 individuals from the Odontology. This means that all the patients from the Odontology ought to have been excluded from the study initially, and that the original number of patients should have been 63 persons less than what was stated in the dissertation and Oncology 2001.

Sudbø’s assertions that the Commission is wrong on this point must be rejected, since the Commission does not find it likely that these 63 persons were linked with Cancer Registry data as alleged by Sudbø. It is also a fact that the so-called Sudbø8 file which formed the basis for New England Journal of Medicine 2004 (and thus in all essentials also New England Journal of Medicine 2001), did not comprise any persons from the material from the Odontology.


The linkage of patient data with Cancer Registry data for research purposes presupposed then, as now, that a participant consent or dispensation from the duty of secrecy existed, see section 4.2.4 and the Cancer Registry’s framework licence dated December 9, 1985, cf item 4.3 of the licence. This was not so.

The Cancer Registry was notified that the Commission was considering expressing a certain criticism on this basis, and made use of its right to comment on an earlier draft report, section 7.3.4. In a letter to the Commission of June 2, 2006, the Cancer Registry submitted that the criticism in the draft is based on an erroneous perception of the Cancer Registry’s different roles. The Cancer Registry alleges that they understood Jon Sudbø’s request of February 20, 1996 as a routine request for follow-up data for patients at the Radiumhospitalet, Department of Pathology. The Cancer Registry alleges that delivery of the required data did not take place as “data supplier or partner in a research project, but as part of the registry role relating to the safeguarding of data completeness and follow-up of patients”, focusing on the central problems which normally are involved in such a follow-up. It is alleged that the data exchange that takes place between the individual hospital and its departments and the Cancer Registry in relation to quality assurance, entails that the Cancer Registry contributes to the diagnosis and treatment departments’ quality control of “their” patient material, at the same time as this normally leads to a quality increase of the Cancer Registry’s main data base. The Cancer Registry alleges that this mutual quality control is “… a most central part of our registry function”, and that this “activity has been ongoing on a routine basis for all years, and that it is considered to be covered by the previous licence as well as the current regulations”.

Further, the Cancer Registry submitted that the draft criticism seems to be based on the assumption that the Cancer Registry should have understood that Sudbø applied for data for a research project. If so, this is a basis for criticism that the Cancer Registry considers to be unreasonable. The Cancer Registry in this connection refers to the fact that Sudbø’s approach was accompanied by a detailed list of patients, that the cover letter stated that Sudbø’s project took place at the Radiumhospitalet, Department of Pathology, and that data had also been obtained from Haukeland Hospital. The Cancer Registry submits that in light of all facts relating to Sudbø’s approach in 1996 there was no reason to believe anything else than that the patient list concerned the Radiumhospitalet’s own and called-in biopsy preparations, and that the request was made as a request for follow-up data – a type of requests that the Cancer Registry receives and deals with continuously. It was not until 2006, when this case had come to light, that the Cancer Registry realized that the material which Sudbø in the request from 1996 wished to have quality-assured, did not concern the Radiumhospitalet patients, but material from Gade’s Institute and the Institute of Odontology’s department of pathology. The

Cancer Registry furthermore submits that if the facts had been correctly stated in 1996 Sudbø would not have been given Cancer Registry data. The Cancer Registry in this connection refers to the fact that Sudbø’s request in 1994 was refused exactly because he wanted follow-up on other institutions’ patient material, and also wanted personal access to the Cancer Registry’s data base.

The Commission has considered the remarks made by the Cancer Registry and compared them with the information provided in Jon Sudbø’s letter of February 20, 1996 to the Cancer Registry, quoted above, and other information. On that basis, the Commission finds that it appears from the letter that it is a question of a research project, something which the Cancer Registry ought to have understood. The handing over of the data thus appears to be contrary to the licence conditions, se otherwise section 7.3.4.

The material from Gade


In the aforementioned letter of February 1, 1996 from Sudbø and Reith to professor Bang of Gade, it is further stated:

“… We also have an agreement with chief physician Langmark at the Cancer Registry relating to obtaining data regarding which of these persons later developed oral cancer. However, our existing material of 63 individuals is too small to be able to be used in such a retrospective study and our question to you is therefore whether you could place a corresponding material at our temporary disposition. To begin with, we are interested in personal data from individuals who have been given the diagnosis of mild, moderate or severe squamous epithelium dysplasia from the oral cavity, regardless of localization. If data from the Cancer Registry should show that the percentage that has developed oral cancer is sufficiently large, it would also be of interest to borrow the biopsies for the production of biopsy specimens for coloring and examination.”


A data list with patient data for the period 1981-95 was sent in an undated letter from Bang to Reith and Sudbø. The Commission cannot see that any participant consent or dispensation from the duty of secrecy for the delivery of patient data exist, although this was a requirement when patient information subject to secrecy was to be delivered, see section 4.2.4. Thus, the handing over of the data appears to be contrary to the set of rules applicable at the time, see also section 7.3.3. The University of Bergen has not commented on the Commission’s draft criticism. A letter dated May 15, 1996 from Reith to professor Andreas Myking at Gade (to whom Bang had referred Reith), states:

“We have at our department been in contact with professor Gisle Bang of Haukeland Hospital. He has sent us data from 161 patients diagnosed with dysplasia from mucosa (see the annexed list). We have compared these data with information from the Cancer Registry. It appears from these data that approximately 50% (79/161) of the patients with the diagnosis mild, moderate or severe mucosa dysplasia within a five years’ period

developed squamous cell carcinoma of the oral cavity … Our question to you is therefore whether it is possible to be sent cut blocks and copies of referring letters from the patients in the lists we have enclosed …”
Sudbø and Reith were sent referring letters from Gade, a fact that is confirmed in, i.a., a letter dated December 9, 1996 from Sudbø to Professor Anne Christine Johannesen at Gade. The Gade material originates from clinics spread all over Western Norway.

The patient number of 161 persons is confirmed by the Cancer Registry, which in February 1996 linked 161 persons with the Cancer Registry’s data. (Two out of 163 were duplicates. One more person was according to the Cancer Registry furthermore excluded because of adenoma in the stomach.) Consequently, the list which Sudbø and Reith got from the Cancer Registry comprised linked data from 160 or 161 persons. The Cancer Registry’s investigation in 2006 shows that all 161 persons who were linked originated from Gade.

The aforementioned letter dated December 9, 1996 from Sudbø to Johannesen states:
“Enclosed please find a set of copies of referring letters I have received from Professor Gisle Bang, relating to dysplasia from the oral cavity. You will also find enclosed a Microsoft Excel spreadsheet stating patient data as they are linked from referring letters to data from the Cancer Registry. What is particularly striking with the material, is the high share of dysplasia showing malignant transformation (50%). It therefore seems to be necessary to have verified the original diagnosis (mild, moderate or severe dysplasia) and exclude any CIS. You have been in contact with Professor Albrecht Reith regarding this material, and you probably know the problem from this. Thank you for your interest and will to assist in this!”
The letter of reply from Johannesen to Jon Sudbø dated May 21st, 1997 states:
“… I have looked at all the biopsy specimens from the pile I was sent. I have a series of comments which I shall try to express as clearly and briefly as possible:


  • I have tried to make a list of names, P-numbers (biopsy number) and diagnosis. All this is in normal type.

  • My objections are in italics.

  • Some referring letters are lacking in relation to the list received. These are mentioned by name (italics) without any further text.

  • At many places there is no correspondence between the diagnosis on the list received and the biopsy answer! Here biopsy answer must of course apply.

  • The explanation of why the material from Bergen has such a high frequency of malignant transformation is that you have perhaps not compared the time of their dysplasia diagnosis with the time of the malignant diagnosis. For example, at several places it appears that dysplasia diagnosis lies in a resection border in a carcinoma that has already occurred. This is therefore no

malignant transformation but a dysplasia that has already had a cancer diagnosis. Several of the patients have also previously been operated for squamous cell carcinoma.

  • As regards the diagnoses, I have not changed many. I have accepted a deviation of 1 degree (mild-moderate, moderate- serious) without having corrected the diagnosis, this because the grading to such a relatively large extent is subject to personal assessment. Where there are obvious errors or major deviations, I have corrected the diagnosis.

I hope this is sufficiently clear. Then I want to wish you good luck with your further work!”

It is to be noted that Johannesen’s list comprises data from 144 different persons only. Reith thinks the letter from Johannesen looks “straightforward”, but he wonders why the Cancer Registry only excluded two persons out of 163, whereas Johannesen apparently excluded a further 17 persons. Reith states that he cannot recall having seen/read the letter before.

The Commission asked itself the same question. The probable explanation is that the list of 163 observations, which originated from Gade, and which were linked by the Cancer Registry in the beginning of 1996, comprised two duplicates, so that the list comprised 161 persons after linkage. In a list recently prepared by the Cancer Registry there is a column called “Diagnosis on A.C. Johannesen’s quality assurance list”.32 This column contains 17 blank observations (which also are persons). This probably means that Johannesen has quality assured the diagnosis of 161-17=144 persons. This agrees with the figures the Commission’s own investigations have produced, i.a. through comparisons with block numbers, see below. This is also consistent with what Johannesen wrote in a letter of May 21, 1997: “There are some referring letters that are missing in relation to the list sent. These are mentioned by names (italics) without any further text.”

Total number of patients: This means that Sudbø and Reith in the summer 1997 were left with a reclassified data file from Gade’s Institute which was linked with the Cancer Registry data, with data from only 144 persons, i.e. 119 persons less than what is stated in the dissertation and Oncology 2001.

The Commission finds it surprising that in the communication with Gade, after the linkage with the Cancer Registry data, there is no reference to the material from the Odontology. If the 63 persons from the Odontology, which the Cancer Registry has not linked, and which for that reason should have been excluded, nevertheless are added, the data material in the best case consists of data from 144+63=207 different persons.

It should be noted that the number of samples (tissue blocks, biopsy specimens, etc) far exceeds this number, because there are several samples from the same person, something which is totally usual.

Sudbø has expressed that he has no qualification to comment on these figures, stating that it was Reith who was responsible for the handling of data files between Gade’s Institute and the Cancer Registry. However, the letter from Johannesen was addressed to Sudbø alone, although Reith does not exclude in his

statement to the Commission that he has seen the letter dated May 21, 1997. He further states that “the

diagnoses contain so many histopathological terms unknown to me that I would have discussed them with a pathologist at the department, I know I have never done that.” Reith has otherwise referred to a memo written by Sudbø in 2000, in which Sudbø consistently uses the “I” form in his description of the handling of the data material. In light of this, the Commission cannot trust Sudbø’s assertions relating to Reith’s alleged central role in the handling of data files, see also the letter of February 20, 1996 from Sudbø to the Cancer Registry, rendered above, and the letter from the Cancer Registry to Jon Sudbø of March 22, 1996.

The Commission has considered whether Sudbø and/or Reith may have misunderstood and mixed up the number of samples with the number of patients, etc. The Commission finds this not very likely. The Commission here finds reason to refer to Johannesen’s letter of May 21, 1997, which in the Commission’s opinion should not give room for misunderstandings and misinterpretations. In any case, the letter should have caused a thorough reevaluation on their part which could have brought any misunderstandings to light.

The Commission has furthermore questioned whether the material from Gade and, as the case may be, the Odontology has subsequently been supplemented, for example by material from other patients having been collected and brought into the project, for example from Gade, the Odontology, private practices or other channels. The Commission has not found any grounds at all for this being the case. The Cancer Registry drew the same conclusion. This is in fact confirmed also in the dissertation and in the articles. Such a supplement is also contrary to expectations for more practical reasons.

Sudbø’s allegations that he obtained supplementing information as regards complete personal numbers do not, in the light of Puntervold’s, Reith’s and the Cancer Registry’s statements, appear as probable, and they are also somewhat surprising and contradictory in light of his other allegations that it was Reith who was responsible for the handling of data files between Gade and the Cancer Registry and the arrangement of the data material otherwise. On this background the Commission maintains its clear understanding that the patient number stated is not correct.

Accordingly, the Commission finds that there was probably no access to linked data from more than 144 persons, in the best case from 207 persons.


Exclusion of patients with erythroplakias (red patches in the oral cavity)

In the PhD dissertation is stated that 21 out of 263 patients were excluded initially from analyses of survival because they had erythroplakias. Erythroplakias is far more serious than white patches, and also occurs much more seldom. It should be noted that these persons, i.e. persons with erythroplakias, probably were

included in a subsequent study with 37 erythroplakia patients which formed the basis for a scientific publication.33

The Commission finds that the figure 263 cannot be correct. Whether there were 21 persons with erythroplakias in the basic material seems uncertain to the Commission. Johannesen of Gade, however, has stated that erythroplakias were not registered at Gade. If so, this means that patients with erythroplakias must originate from the Odontology. And Sudbø himself stated to the Commission that the erythroplakias originated exclusively from the Odontology. The Commission is in some doubt that 21 of the 61 persons who came from the Odontology had erythroplakias, but this can nevertheless not be excluded due to the missing documentation of these patients. The article that reported the follow-up results from 37 erythroplakia patients does not state from where the patient material (which allegedly was collected in 1988-2000) originates.

However, Sudbø has later on, in comments to the preliminary draft investigation report, maintained that the erythroplakias originate from the Odontology. He has stated that a relatively large material of erythroplakias existed there, because Hanna S. Koppang for many years had taken an interest in these lesions. Sudbø has stated that “the original 21 erythroplakias, originally submitted as leukoplakias, but classified as erythroplakia because they originally had been described as erythroleukoplakia.” As mentioned, the Commission is in some doubt about this explanation. In addition, the Commission has reviewed the list of the 63 persons coming from Oslo, without finding any references to diagnoses of erythroplakia.

Even if the Commission does not manage to document that 21 persons did not have red patches in their oral cavity as stated in the dissertation, there is nevertheless such considerable doubt and uncertainty related to this material that it gives reason to concern for whether it is correct at all. This particularly applies in light of other findings in this case, which will be accounted for below.


Classification and reclassification of patients with white patches in their oral cavity – the inclusion and exclusion process

In figure 1 in New England Journal of Medicine 2001 and figure 5 in the dissertation (se figure 1 in the report) is stated that 242 patients with white patches in their oral cavity (leukoplakia), originally were included in the study after the alleged exclusion of 21 persons with red patches in their oral cavity (erythroplakias).

The Commission refers to the account above in which is determined that it is not likely that this figure is correct, i.e. that one has not had access to 242 persons with dysplasia diagnosis. The real figure is

144 or in the best case 207. The Commission nevertheless finds reason to discuss the actual facts which are alleged by Sudbø et al. in the dissertation and articles.

The key item at this stage of the study was to decide which patients met the criteria (the inclusion criteria) to be made part of the planned ploidy study, i.e. a so-called inclusion process. Before starting a research process, it is usual to state inclusion and exclusion criteria for the research participants, i.e. which patients have the qualities to be studied, and which have qualities that mean that these persons cannot be included in the study. It is important that this inclusion process takes place according to certain criteria determined beforehand and that can be checked, in order to avoid inappropriate selection.

To find which patients with dysplasia diagnosis met the inclusion criteria, the patients were checked against three key criteria (see figure 1):



  1. Dysplasia classification: One tissue block from each individual patient had to be classified to see which type of dysplasia the individual patient had. The patients were then to be divided according to the criteria mild, medium and serious dysplasia.

  2. Prior or simultaneous cancer diagnosis: Patients who had or had had oral cancer were to be excluded. This was because it was change/transformation from white patches to cancer over a certain time interval which was to be studied.

  3. Insufficient data material: Persons for whom sufficient data were not available had to be excluded.


1) Dysplasia classification (the reclassification): It appears from the three articles and subsequent articles that the classification was made by four pathologists who reclassified tissue samples from each individual patient, according to guidelines prepared by the World Health Organization (WHO). New England Journal of Medicine 2001 page 1270 states for example: “All histological sections were subsequently reevaluated by four pathologists according to the guidelines of the World Health Organization.” According to how the procedure is described, it must be understood that the reclassification took place blindly (without knowing the patients’ identity, diagnoses and the like), and that the pathologists worked independently of one another. This is because the classification was based on assessment, and because it was a key element in this study that the classification became as correct as possible.

It is therefore correct, as is stated in the articles and the dissertation, that patients in relation to whom one disagreed on the diagnosis grading, had to be excluded. In the articles and the dissertation is stated that altogether 46 patients had to be excluded for this reason. In Oncology 2001 the figure is 45 (see figure 1), but this is explained in the dissertation by an “outlier” having been excluded, but the Commission cannot see that anything was mentioned about this in the article. Sudbø states that it was deleted by the editor.

This reclassification process is described in detail in J Pathol 2001, in which one also directly compared the classifications of the individual pathologist.

It is a fact that the four pathologists referred to are



  • Gisle Bang, Gade’s Institute

  • Hanna Strøm Koppang, the Odontology

  • Anne Christine Johannesen, Gade’s Institute

  • Bjørn Risberg, the Radiumhospitalet

This appears explicitly from Oncology 2001 in which these four persons are listed with names in acknowledgements. There is no doubt that all of them are qualified pathologists. It is also a fact that the original dysplasia classifications in no way were so sure that they could be used in the inclusion process. This means that originally all patients had received a dysplasia diagnosis, but Gade and the Odontology were and are very clear that the grading was flawed and not directly applicable for scientific purposes. In other words, there was an obvious need for a blinded reclassification for scientific purposes. This need was at an early stage in fact underlined explicitly to Jon Sudbø by the department manager, Professor Jahn Nesland, who is a pathologist himself. Nesland states to the Commission that such a reclassification for scientific purposes normally is performed blindly with two independent pathologists who afterwards have a so-called consensus meeting to compare results in order to arrive at an agreed dysplasia classification. Thus, it is a fact that it was a question of reclassification of the entire material these four were to make. One may ask why one allegedly used four pathologists instead of the customary two. A possible answer is that the classification becomes more certain the more independent pathologists are used for classification.

However, the Commission cannot see that the information that the reclassification was made by four pathologists is correct. Admittedly, Gisle Bang was central in the collection of the original material. Nevertheless, it is a fact that Gisle Bang did not make a reclassification of all the material. In the best case, he can be deemed to have participated in the original, but for scientific purposes obviously flawed classification of parts of the material from Gade, but not by far the whole material.

The same applies to Hanna Strøm Koppang. Admittedly, Koppang was central at the collection of the original material, but she has hardly participated in the reclassification of the material. In the best case she can be deemed to have contributed to the original, but for scientific purposes obviously flawed classification of the material from the Odontology, but not by far the whole material.

Anne Christine Johannesen has in a sense participated in a reclassification of material from Gade, i.e. 144 of the patients. However, she has not classified the material from “all” the patients. But Anne Christine Johannesen is unable to understand that she is supposed to have participated in a reclassification also of the Gade material. She states to the Commission that it was a fact that her task was to quality-assure the material they had delivered to Sudbø and Reith, such that her responsibility was limited to doing so. She believes this appears clearly from the aforementioned cover letter dated May 21, 1997 to Jon Sudbø with a copy to Reith. In Johannesen’s letter reference is made i.a. to a letter from Reith to Myking of May 15, 1996, in which as mentioned reference is made to the fact that the linkage with the Cancer Registry data had identified a transformation rate from dysplasia (white patches in the oral cavity) to cancer of as much as 50%. This sensational finding is also emphasized by Sudbø in a letter to Anne Christine Johannesen of December 9, 1996. However, Johannesen points out that the high transformation rate of 50% which Sudbø and Reith had referred to, is based on several obvious errors, for example that those that had received dysplasia diagnosis had a prior or simultaneous cancer diagnosis, such that no transformation can be determined. It further appears that Johannesen did not make a careful reclassification based on the criteria mild, moderate and severe dysplasia.

In this context it should also be noted that the Cancer Registry this year has found it very difficult to comprehend why Sudbø and Reith in December 1996, several months after the Cancer Registry’s linkage, asked Johannesen to revise/reclassify diagnoses relating to patients who should have been excluded based on the information they had received from the Cancer Registry, and based on far more data per patient than Johannesen had any possibility to possess.34

Nor did Bjørn Risberg see himself as one of the four pathologists. Risberg states to the Commission that when he read this section in the article, he assumed that this concerned four pathologists which were unknown to him, and that his task had only been to quality-assure the reclassification performed by others. Risberg cannot recall how many samples he classified, i.e. that he may have classified between 150 and 242 samples. The Commission finds it surprising that Risberg has not been informed that he was one of the four pathologists. In the dissertation it is only Risberg who is thanked for having made a reclassification, whereas Koppang is thanked for histological classification.

In his comments to the preliminary draft investigation report, Sudbø writes among other things that he is “entirely incapable of understanding that the original dysplasia classifications could not be used for scientific purposes. This is the first time that I have heard that this point has been raised. However, it was discussed whether one should have had consensus meetings and calibration by pathologists prior to the classification. I objected strongly to this, because the clinics base their treatment on routine diagnostics, not especially constructed diagnostic procedures and assumptions. Everyone involved in this project,

including Reith, Bryne and A.C. Johannesen, agreed that there was much to say for this view.” Furthermore, Sudbø asserts that the term “reclassification” was consistently used only to designate that different pathologists, independent of one another, had classified and graded the dysplasias.

The Commission does not have confidence in Sudbø’s explanation on this point, since in the articles and the dissertation he so clearly and unambiguously refers to a “reclassification” by four pathologists in line with the WHO’s guidelines. It is also a fact that Johannesen was not “involved in the process” by discussing the planning and the challenges as the impression seems to be by Sudbø’s explanation.

The Commission finds that the reclassification was not performed by four pathologists in the manner described in a series of articles and the PhD dissertation.

In the articles and the dissertation it is alleged that 46 patients were excluded because one did not obtain consensus on the grading of the dysplasia diagnosis. There can hardly have been any consensus meeting or the like, when in the best case it is only one pathologist who classified the entire material. None of the four stated pathologists have been participating in any consensus meeting or the like. Sudbø states that there was no consensus meeting between these pathologists, but that there was consensus in the form of “conformity of opinions”. The Commission also finds such conformity of opinions to be entirely unlikely, as long as the total material hardly was assessed by four pathologists.

The allegation that 46 patients were excluded due to failing agreement among four pathologists regarding dysplasia grading thus appears as unfounded and erroneous.

The Commission has not found anything to underpin that the figure 46 is correct. The Commission has no alternative figure to put forward, however, as long as the material has not been reclassified in a proper manner for scientific purposes, and as long as the Commission believes that the patient basis was far smaller than what is stated.



2) Preceding or simultaneous cancer diagnosis: The articles and the dissertation state that 36 patients were excluded from the study because they had a simultaneous or preceding cancer diagnosis. The point of the study was exactly to study which patients, after having received a dysplasia diagnosis, at a later point in time received a cancer diagnosis. Consequently, persons who had a simultaneous or preceding cancer had to be excluded. The Commission’s review of available data files shows, however, that the number of patients with a simultaneous or preceding cancer diagnosis must have been far higher than what is stated in the articles and the dissertation. Reference is again made to Johannesen’s letter of May 21, 1997 to Sudbø and Reith in which precisely this point is emphasized, i.e. that a far higher number of patients had a preceding or simultaneous cancer diagnosis, and that this was something Sudbø and Reith obviously had overlooked.

Johannesen’s data file annexed to the aforementioned letter shows that at least 47 out of 144 persons should have been excluded for this reason. The Cancer Registry’s own investigations also conclude that this figure should be far higher. The Cancer Registry believes that at least 12 out of 63 persons from the Odontology and 76 out of 156 persons from Gade ought to have been excluded for this reason.

Sudbø states that this information is new to him, and he is not quite able to comprehend it. He raises the question of whether all preceding or simultaneous cancer diagnoses are referring to oral cancer or related (tobacco-conditional) cancer. He points out that it is of limited interest whether a patient has had colorectal cancer, cervical cancer, or melanoma prior to oral cancer. Finally, he raises the question of whether an updating of the Cancer Registry’s data base has taken place as regards registered cancer incidents since 1995/96.

The Commission here refers to the letter from Johannesen that shows that already in 1996/97 one was aware of all cases of preceding or simultaneous cancer. For the avoidance of doubt, the Commission has in addition compared the Cancer Registry’s list from 1996 with the Cancer Registry’s data from this year.35 For the most part there is consistence both as regards cancer dates and localization (the type of cancer diagnosis). There are a few deviations regarding the referring letter date, but nothing of substantial importance. Thereby, it is also clear that this concerns preceding or simultaneous oral cancer, and not other types of cancer. In this context, the Commission has checked the localization code entered for each patient to see if the latter is the case.

Accordingly, the Commission finds that the number of persons with preceding or simultaneous cancer diagnosis was far higher than the number stated in the articles and the dissertation. This is a very serious matter, which clearly and isolated seen entails that the research results cannot be considered as valid.

3) Flawed data material: Finally, in figure 1 is stated that 10 patients were excluded because one did not have appropriate data or material from these patients. Based on the discussions above, this number must be considered as being unlikely low. Reference is here made to the fact that the material from the Odontology (63 patients) was not linked by the Cancer Registry, and therefore should have been excluded for that reason, if not before. Reference is also made to the fact that one only had material from 144 patients from Gade.

It is obvious that far more patients should have been excluded because one did not have sufficient material to include them in the study.

In this context, the Commission finds reason to point out that Johannesen in a letter of May 21, 1997 remarks that “there is no conformity between the original diagnosis and biopsy answer and that the latter must apply”. Furthermore, it is remarkable that Jon Sudbø in an email to Johannesen of December 11, 2001 writes that “The dysplastic material is unadulterated in the sense that it does not include patients with

simultaneous or preceding carcinoma, neither in their oral cavity or UADT otherwise. The not dysplastic material is restricted, 46 cases.”

Based on this, the Commission finds that there are so many important errors and flaws in the inclusion and reclassification process that the resulting outcome is not credible. The Commission is of the opinion that the errors in the reporting are serious.

The inclusion of 150 patients in the ploidy study

According to the articles and the dissertation, altogether 150 patients were included in the study itself. This ploidy study, which is the experiment proper in the PhD project, was made in 1998-99.

At that time Sudbø had used up his four years as a research fellow, and also a last supplementary year. He was then left without any fellowship salary. His closest superior, Professor Håvard Danielsen, PhD, then proposed that Sudbø should use the method developed by Danielsen to analyze the material that Sudbø had in his possession. Sudbø accepted this, and Danielsen arranged for six months of salary funds from the Radiumhospitalet as well as the assistance of a laboratory technician. Sudbø alleges that he himself had taken the initiative to this image analysis already in 1998, but had not got access to the equipment because his oral cavity project was not a prioritized project at the department.

Although it is unlikely, based on the preceding discussions that there were 150 patients that met the inclusion criteria, the Commission has nevertheless found reason to investigate the ploidy analysis in more detail. This is of importance for the clarification of whether the obvious errors that so far have been discovered are due to sloppiness and incompetence or scientific dishonesty.

The Cancer Registry’s investigation: In its investigation, the Cancer Registry refers to Johannesen’s “limited reclassification” showing that one was left with a maximum of 85 patients only who met the inclusion criteria, divided on mild cysplasia (58), moderate dysplasia (18) and severe dysplasia (9). (In addition, 8 persons were given the diagnosis dysplasia, 4 hyperplasia, 4 preceding cancer, 43 simultaneous cancer – in aggregate 144). Correspondingly, table 1 in New England Journal of Medicine 2001 shows a distribution on mild dysplasia (49), moderate dysplasia (57) and severe dysplasia (44) – in aggregate 150. In other words, the numbers stated in the article do not at all correspond with Johannesen’s classification. Nor does the grading stated in the New England Journal of Medicine 2001 correspond to the list Jon Sudbø according to the Cancer Registry received from the Cancer Registry in 1996, and which had the following division: dysplasia 4, mild 38, moderate 22 and severe 99, in aggregate 163.

The Cancer Registry believes that this number of patients that could be included could have been 79 as a maximum, particularly because of preceding or simultaneous cancer diagnosis (77 out of 156). That is to say close to half of what is stated in the articles and the dissertation.


In spite of the unambiguous findings of the Cancer Registry, based on its own investigations and Johannesen’s independent classification, the Commission has nevertheless found reason to make some investigations of its own. This is connected with the fact that the Cancer Registry’s conclusion that no doubt manipulation and fabrication of data was involved, was quite sensational and serious.

The Commission’s own investigations of patient lists and the ploidy analyses: The Commission has been given access to several lists which apparently contain data from 150 patients who probably were included and studied in the ploidy analysis which took place in 1999. It is a fact that the lists comprise 150 observations, i.e. registrations. It is also probable that someone (see in more detail about this below) has analyzed at least 150 blocks/samples/preparations/monolayers. Based on the aforementioned investigations and findings, the Commission raised two entirely central and specific questions:

  1. Was the ploidy analysis performed on 150 different persons, or may it be that several analyses originate from the same person?

  2. Do the persons included and studied really meet the inclusion criteria?

With the help of available data lists and comparisons between them, and comparisons with among other things data from the Cancer Registry, it has been possible to obtain precise documentation of which patients were actually included, including these patients’ disease history (i.e. whether they met the inclusion criteria).

The lists and the patients form the basis for the three mentioned articles, the dissertation and several subsequent publications, i.a. New England Journal of Medicine 2004, and have therefore been in the very center of attention for the Commission, see annex 3.

The Commission will here discuss these lists in more detail. In particular three lists are of interest:


  • L-29. This list is assumed to be the original list used in the study, and which the Commission has received from Danielsen. As head of section, Danielsen obtained it from the archives in 2006 on the Commission’s request. It is noted on the list that it was produced in April 1998. The list is assumed to form the basis for the PhD project, including New England Journal of Medicine 2001.

  • Rawdata. The Commission has received this list from Reith. Reith has stated that he had not seen this list until 2006, when he asked Ruth Puntervold for it in connection with this case. Puntervold is supposed to have received the list from Sudbø in 2005 in connection with Bjørn Risberg’s wish to measure the preparations again, see 5.3. It is in harmony with, and is probably based on, L-29. According to Reith, the data list forms the basis for New England Journal of Medicine 2004, which again in all essentials is based on New England Journal of Medicine 2001.




  • Sudbø8. The Commission has received this list from J. Jack Lee at MD Anderson, who again received it from Sudbø. Lee is the bio statistician who ran the analyses which form the basis for New England Journal of Medicine 2004. According to Lee, this list is the basis for New England Journal of Medicine 2004.

These three lists are in harmony with each other, which means that there is a preponderance of probability that they originate from the same patient material. However, the individual lists contain more or other registrations. The Commission has not found any basis for these lists not forming the basis for the analyses which again form the basis for the publications in i.a. New England Journal of Medicine 2001 and 2004. Nor has Sudbø or others submitted any patient lists that deviate essentially from these lists.

The rawdata list comprises preparation/block/sample numbers which make it possible to obtain information from the Cancer Registry. This list is, apart from block numbers, identical to Sudbø8.

The table in annex 4 shows the 150 observations (records) which the Commission has assumed formed the basis for the article in New England Journal of Medicine 2004. The first column is a continuous numbering of observations as listed in Rawdata and Sudbø8. The second column comprises unique persons. Each observation in Sudbø8 comprises one or several preparation numbers (block numbers). These preparation numbers were linked to the same person apart from two members in observation 51, which proved to belong to two different persons. This observation is therefore listed twice. The third column is the preparation number itself. This is blanked out for reason of personal data protection. There were 8 observations on the rawdata file for which there were no preparation number. For these 8 “missing” is noted in the column. Then follows a column showing the year when the preparation (the sample/biopsy) was taken. The three next columns are from the file which the Cancer Registry delivered to Sudbø in 1996. Then follow three columns from Sudbø8. The column “year leukoplakia” should be corresponding to the column “year preparation”. Finally there is a column showing whether preparations for ploidy classification had been made.

The Commission has had access to dates for leukoplakia diagnoses and dates for cancer. These coincide entirely with Rawdata and Sudbø8. Moreover, it should be mentioned that information on age and tobacco is entirely in conformity on the two files. This documents that these two files must originate from the same patient basis.

This comparison also documents which persons did not meet the inclusion criteria, since the date of oral cancer is before the sample (block/preparation number) was taken. The Commission shows that 69 of 150 observations should have been excluded for this reason. The comparison further documents that there are only 64 different persons on the list which we could document as not being contrary to the inclusion criterion. It should be noted that the year has been removed from the block number in the rawdata list. An indication of

the year could have contributed to someone having discovered the latter error. On the other hand, the year is stated on L29, without anyone having discovered discrepancies with the inclusion criterion.

The comparison shows that none of the dates in Sudbø8 agrees with the data from the Cancer Registry. This means that the dates in Sudbø8 are fictitious. The Commission’s comparison of the Cancer Registry’s list from 1996 and the Cancer Registry’s list from 2006 are in all essentials concurrent, so that an error at the Cancer Registry is excluded. And the Cancer Registry has also made a thorough investigation of i.a. all relevant referring letters, etc.

Totally there is reference to ploidy preparations for 69 out of 150 observations (65 different persons) in the rawdata list. Observation number 51 in the rawdata (patnid_re=51) is listed twice since this observation in rawdata had two preparation numbers which proved to be two persons.

The Commission’s comparison shows the following:



  • The rawdata list and Sudbø8 comprise 150 observations made up of a maximum of 140 persons

  • Of those 150 observations only maximum 81 observations meet the inclusion criterion

  • Of the 150 observations there are only 69 observations in which there is a reference to a ploidy preparation

  • Of the 150 observations there are 23 observations in which the year of death is prior to the year of leukoplakia (the year in dateopl). This means that the patient was dead before the diagnosis allegedly was made.

  • No observations with block numbers originate from the Odontology in Oslo.

The list “All original blocks and HE biopsy specimens linked to ploidprep and L29 series” moreover shows a connection between block numbers and list numbers (L31 etc) for the observations in which there is a ploidy preparation (probably ploidy classification). In all there are 167 observations. The Commission has a list with a variable/column, “place”, which shows whether the block is from Gade or from the Odontology.36 This is defined based on block number or list number. As regards the list L47 there is verification with block numbers. The Commission’s list shows the following:




Observations

Unique block numbers

Missing

4

4

Gade

97

80

The Odontology

66

42

In total

167

126

Consequently, there are 167 ploidy preparations, of which 126 only are unique block numbers. This means that there are unique and thus valid ploidy preparations for a maximum of 126 persons. This can be compared with that, according to Sudbø himself, 196 persons were processed statistically where ploidy classification existed in J Pathol 2001. In other words, this statement hardly agrees with the actual facts.

It is worth noting that only 69 out of 167 ploidy numbers can be linked to L29/Rawdata/Subø8. This means that a ploidy analysis has only been made on 69 (and not 150) of the block numbers existing in the rawdata list, which is the basis for the New England Journal of Medicine 2004.

For this reason, the Commission has seen no point in making a new ploidy analysis of the raw data, since the raw data is so obviously flawed.

Sudbø has reacted to this, and has among other things referred to one of the persons who classified remembering to have received about 150 blocks. Danielsen also believes to have seen a tray with approximately 150 blocks. The Commission would remark to this that one has probably classified approximately 150 blocks, such that those who classified, i.a. Wanja Kildal, probably believed that it concerned the number of persons stated in the articles. But the fact is that it involved many duplicates and many persons who should have been excluded, i.a. due to preceding and simultaneous cancer diagnosis.

The Commission has calculated an age distribution from the file Sudbø8 (which form the basis for New England Journal of Medicine 2004) and compared this with the age distribution for the original data from Gade and the Odontology. For Gade two schedules have been made: 1) Based on a list produced by A.C. Johannesen, which via block numbers is linked with Cancer Registry data, and 2) Based on the file that the Cancer Registry delivered to Sudbø in 1996. Table 2 shows the result for three age groups. Age for Gade and the Odontology is age when the biopsy was taken. Age in Sudbø8 is not defined in more detail on the file itself, but in an email to J. Jack Lee of MD Anderson (who made the analyses) Sudbø writes that this is age at “time at initial diagnosis”. Thus we can assume that there are the same age definitions in the files when age for diagnosis is stipulated as age when the biopsy was taken. Moreover, it is difficult to see which other age it could be in Sudbø8. The table shows that there are far more persons in the age group 65-78 in Sudbø8 than in the other files, also when they are joined. Thus there is a distinct discrepancy between the file which forms the basis for New England Journal of Medicine 2004 and the files which form the basis for this.




Table 2: Age distribution in data file for New England Journal of Medicine 2004 (Sudbø8) compared with the age distribution in data file delivered from the Cancer Registry in 1996 (the Oslo Odontology [62 persons] and Gade [163 persons] and lists from Gade [142 persons].






















Basis

Canc.Reg.06

Canc.Reg.96

(Canc.Reg.06/

Gade list 98)









JJL06




Oslo

Gade1

Gade2

Oslo+Gade1

Oslo+Gade2

Sudbø8

Age*



















30-64

40

79

68

108

119

32

65-78

13

59

53

66

72

112

79-95

4

25

19

23

29

6

Total

57

163

140

197

220

150

Lacking

5




2

7

6

0

Total

62

163

142

204

226

150






















Average age

56.8

63.5

63.3

61.4

61.8

68.9






















New England Journal of Medicine 2001
















68.9

New England Journal of Medicine 2004
















68.9






















*Age for Oslo and Gade: Referring letter year minus year of birth; Age for Sudbø8:age in file

The Commission has also had access to the first file which was sent to the USA for New England Journal of Medicine 2004 and compared this with the last file. Table 3 renders the results for some observations.



Table 3. The last 28 records from the first and last file which formed the basis for the article in NEJM 2004.

Both files contain 150 records (lines) numbered continuously at patnid from 1 to 150.




Updated rawdata (first file)

Sudbø8 (last file)

patntid age tobacco yr leukoplakia yr cancer patnid age tobacco yr leukoplakia yr cancer



123 80 3 1993 2000 123 80 3 1993 2000

124 74 3 1993 2001 124 74 3 1993 2001

125 74 3 1993 1994 125 74 3 1990 1991

126 65 2 1994 2000 126 65 2 1994 2000

127 76 3 1991 1993 127 76 3 1989 1990

128 55 3 1985 1989 128 55 3 1985 1986

129 66 2 1982 129 66 2 1982 -------

130 72 3 1994 1995 130 72 3 1991 1992

131 75 4 1989 2000 131 75 4 1989 2000

132 76 3 1985 1986 132 76 3 1983 1984

133 65 1 1990 1993 133 65 1 1987 1988

134 73 3 1997 1998 134 73 3 1994 1995

135 66 4 1987 1988 135 66 4 1984 1985

136 51 3 1995 1997 136 51 3 1990 1994

137 69 3 1989 1990 137 69 3 1986 1987

138 69 1 1991 1992 138 69 1 1983 1988

139 78 3 1994 1996 139 78 3 1991 1993

140 67 3 1997 1998 140 67 3 1994 1995

141 78 2 1989 1989 141 78 2 1986 1988

142 63 3 1995 1997 142 63 3 1993 1994

143 78 3 1991 1995 143 78 3 1988 1990

144 64 2 1994 2002 144 64 2 1994 2002

145 63 3 1993 1995 145 63 3 1988 1991

146 74 4 1993 2002 146 74 4 1993 1995

146 67 3 1996 1998 147 67 3 1994 1997

148 74 2 1985 1989 148 74 2 1983 1985

149 65 1 1990 1992 149 65 1 1987 1988

150 81 2 1994 2000 150 81 2 1994 1994
Conspicuously many dates have been changed from the first to the last file. Even more conspicuous is that the year for the leukoplakia diagnosis has been changed, whereas the age for the leukoplakia diagnosis is unchanged.

Accordingly, the Commission finds that the lists to which the Commission has had access and which form the basis for the PhD project, are not correct. Neither dates, number of patients, nor other checkable observations agree to any reasonable degree with the published data and results. The Commission has tried different approaches and has made a series of other comparisons of lists, sample numbers and patient identities, etc., but the conclusion has always been the same. Neither dates for the leukoplakia diagnosis nor

the date for cancer agree with the corresponding dates in the Cancer Registry, and it is difficult to find any other explanation than that dates and lists to a large extent have been fabricated. This finding is in harmony with the Cancer Registry’s independent internal investigation. This finding is also in harmony with the Commission’s pointing out of flaws in the patient basis.

The ploidy analysis


After having decided which patients should be included in the study (allegedly 150) and which had to be excluded (allegedly 113) via the dysplasia classification, the next step was to carry out the experiment itself. This consisted of classifying the samples/blocks/monolayers from patients included in the study, to see which degree of dysplasia (mild/moderate/severe) and which type of lesion the patient had by the help of a genetic analysis – a so-called ploidy analysis; graded according to diploid/tetraploid/aneuploid. The point was to determine whether a relatively simple DNA analysis of the white patches could predict the likelihood of subsequent development of cancer.

At this time a ploidy classification was made at the Radiumhospitalet, i.e. a measurement of the amount of DNA (hereditary material) by an image-analytic machine. The analysis machine then makes a DNA histogram which draws up a person classification. The classification is subjective, even if the purpose of the machine analysis is to make it as reliable as possible and by that objective.

New England Journal of Medicine 2001 states on page 1272 that “all specimens were coded, and DNA histograms were classified in a blinded manner by four observers.” In J Pathol 2001 there are three. This is in spite of the routine at the hospital being that the classification was to be made blindly by only two independent persons.

To the Commission it has been somewhat unclear who these four independent persons were. Jon Sudbø explains that it was Wanja Kildal, Håvard Danielsen, Jon Sudbø himself, and partly Albrecth Reith. Bjørn Risberg did not take part in this classification, although he was obviously qualified for it. Risberg himself has been somewhat surprised that he was not included in this classification. Reith explains that he understood that Wanja Kildal and Håvard Danielsen made the classification. Reith further states that he himself some time later made an analysis of the histograms to see if he “was in line with WK and HED’s analyses”, but obviously he does not see himself as one of four observers, in that he refers to the fact the dissertation does not state anything about “four” observers.

It must be assumed that Wanja Kildal, who had been trained by Håvard Danielsen, was qualified to classify the samples, in the same way as Håvard Danielsen. Wanja Kildal states that she classified all the samples she received from Jon Sudbø. After Wanja Kildal had classified the samples, she showed Håvard Danielsen the first 30 classifications, in order that he could check if she had done it correctly. This

was done by Danielsen. Sudbø firmly believes that Danielsen also classified the remainder of the samples. He also alleges that a consensus meeting took place between Danielsen, Kildal and himself. Danielsen on his part is certain that he did not classify the other 120 samples, and that no consensus meeting ever took place. Based on Reith’s own statement, the Commission finds that Reith can hardly be considered as one of the four alleged observers. Whether and to which extent Jon Sudbø classified the material, seems rather unclear to the Commission. If Sudbø did classify, it is doubtful whether the classification was blinded inasmuch as Jon Sudbø and Albrecth Reith probably had access to and knew the patients’ identity and diagnosis. Principally, Sudbø denies this about the non-existing blinding. Alternatively, he alleges that since both he himself and Reith “had had access to and good knowledge of the contents of the background file from the Cancer Registry … the non-existing blinding should in such case apply also for him [Reith].” However, Reith denies that he had such access to the data material, and has accounted for this in a way the Commission finds credible. On this point, the Commission will comment that in most observation studies a “blinded” classification may be and is performed even if those who carry out the classification could have cheated by opening the blinding. In normal circumstances, one has sufficient trust in the researchers who carry out the study.

Accordingly, the Commission finds that the ploidy analysis hardly took place as described in New England Journal of Medicine 2001. On the other hand, this is not an item of crucial importance to the validity of the results, since the Commission is convinced that the samples which were analyzed comprised several duplicates, and that several blocks originated from patients who should have been excluded from the study. The ploidy analysis itself then appears to the Commission as pseudo valid, since the results were not linked with the correct number of patients that could be included.

The research result


The results of the ploidy analysis compared with cancer development, i.e. the research result itself, was astounding. The question was to which extent the ploidy in cells from white patches could be used as a sign (a predicative indicator) of future oral cancer, which is a very serious form of cancer.

Sudbø et al could show that patients with aneuploid lesions had a particularly poor prognosis, by approximately 90% developing oral cancer during a five years’ follow-up. At the same time patients with diploid lesions had a very good prognosis, by only 5% subsequently developing cancer. For patients with tetraploid lesions the probability of transition to cancer was a little above 50%. Thereby Sudbø had confirmed his hypothesis and arrived at a very good method to predict oral cancer for persons with white patches.


Based on the above, this sensational research result can no longer have credibility.
1   ...   12   13   14   15   16   17   18   19   ...   32


The database is protected by copyright ©dentisty.org 2016
send message

    Main page