How to deal with science journalists

August 21, 2009

Recently I was contacted for the first time by a journalist who wanted to know more about my research. She had seen our latest paper and wanted to ask me a few questions! Flattering, of course! Communicating science to the non-scientist audience is an enshrined duty of all researchers, although not very many do much of it in reality.

So how would I go about it?

Well, I got some advice.

  1. Give the journalist a brief and non-technical summary. Do not assume that she has read the paper – she is calling you because she wants to know what was in it.
  2. Make certain that you phrase yourself in a way that lends itself to quotes. Metaphors, similes and other rhetorical devices are recommended.
  3. Ask to see the article and correct incorrect quotes and other inaccuracies.

And so I did, while a little strained for time.

The result? Slightly hilarious. You can see it here and here.

Obviously, I didn’t expect that the article would appear on sites that so strongly endorse products and services. Beyond that? I’m a little bit clueless. Will anybody read it? If they do, will it have any impact? Should I have given my answers differently? I don’t know.

But to be honest, the sight of my name in print with “Dr” in front of it is still enough to make my day!


Selenite against mesothelioma – mechanism of action explained

July 17, 2009

Our latest paper is now freely available online as a fully formatted pdf from the Journal of Experimental and Clinical Cancer Research. As I have promised, here is a non-technical summary!

What did we study?

This work is about malignant mesothelioma, an unusual type of cancer that is caused by asbestos. It is always deadly, and current treatment extends life expectancy only by a few months. We have been working for some time on a new experimental drug called selenite – a simple, selenium-based compound.

Interestingly, mesothelioma cells come in two kinds – epithelioid and sarcomatoid. If a tumor contains sarcomatoid cells, the patient will be expected to respond worse to therapy and die sooner. We have previously found that selenite is particularly effective against sarcomatoid cells, and that it is able to induce apoptosis, the “suicide program” of the cancer cells.

In this paper, we studied the apoptosis mechanisms in both epithelioid and sarcomatoid cells, to see if there were any differences that could explain why sarcomatoid cells are more sensitive. Also, very little was known about the apoptotic response to selenite in mesothelioma cells, we wanted to see how they compare to other cells.

What did we find?

Selenite caused the activation of a number of apoptosis signaling molecules. There was a difference between sarcomatoid and epithelioid cells in the activation of two proteins in the so-called Bcl-family. Sarcomatoid cells clearly overexpressed a protein called Bax. Perhaps this is part of the reason why they are more sensitive to selenite.

There is a “master regulator” of apoptosis called p53, and we investigated it rather thoroughly. It turned out that the cells amassed lots of p53 in their nuclei after selenite treatment, but it didn’t do anything! Normally, it would regulate the DNA and determine which genes should be read. But after selenite treatment, p53 became inactive and unable to regulate gene expression.

Cells stained for p53. Brown nuclei contain much p53 that is inactive. A and C are controls, B and D are treated with selenite. Sell the full paper for details (figure 2).

Cells stained for p53. Brown nuclei contain much p53 that is inactive. A and C are controls, B and D are treated with selenite. Sell the full paper for details (figure 2).

My greatest surprise was that the apoptosis signaling network was so robust and redundant. It’s really not a well-defined linear cascade of events, but rather an interlaced network of protein interactions which depend on and modulate each other. In this paper, we found that inhibition of some of the major apoptosis-signalling proteins had virtually no effect at all on the events following selenite treatment, even though we could prove that the inhibitors were effective in themselves.

What are the implications for the future?

We hope that selenite will become a useful drug for the treatment of mesothelioma in the future. If so, part of its mechanism of action has now been established.

Check out the full paper, it’s open access!

Nilsonne, G., Olm, E., Szulkin, A., Mundt, F., Stein, A., Kocic, B., Rundlöf, A., Fernandes, A., Björnstedt, M., & Dobra, K. (2009). Phenotype-dependent apoptosis signalling in mesothelioma cells after selenite exposure Journal of Experimental & Clinical Cancer Research, 28 (1) DOI: 10.1186/1756-9966-28-92

Monstrous effort to map a transcriptional network

July 8, 2009

ResearchBlogging.orgThe FANTOM consortium report in the latest issue of Nature Genetics that they have measured what happens with the entire, total, gene expression during the specific differentiation of a cell line called THP-1. Not the expression of just the 10 000 most important genes, all of them. At the same time.

Their findings are a heap of data which is probably larger than the whole body of research on medicine and biology up until the early 1900’s. If I try to say what their main finding is, I’d lean towards the interconnectedness of the signaling network. It doesn’t have one single weak spot, where you could knock out a certain gene and profoundly change the network dynamics. Knock-out of some genes had effects on many other parts of the network, but in general the system seems to be robust because of redundancy and interconnectedness. I have drawn similar conclusions in my own latest paper, though my methodology is a pair of binoculars compared to their multinational telescope.

Professor Hayashizaki of the RIKEN Omics Science Center was the general organiser of this study.

Professor Hayashizaki of the RIKEN Omics Science Center was the general organiser of this study.

My main thoughts, however, upon reading this paper were not so much about the actual research, but more about the way it was done.

  1.  With the advent of large-scale initiatives like these, we will perhaps have charted most of the “connectome” of the cell within the next decades. This is the map of the decision-making pathways. The neuroanatomy of the cell, if you wish. It has enormous potential to explain how, exactly, things go wrong in diseases such as cancer.
  2.  Biology is starting to resemble some branches of physics, where research advances through large concerted efforts. The author list of this paper is half a page long, with the authors’ affiliations taking up another half page. There will be less space for the nerdy loner scientists and greater demand for the entrepreneurial, outgoing kind of researcher in the future.
  3.  Seventeen figures and fourteen tables, and the whole methods section, have been relegated to the “supplementary material” that is only available online. Reporting on this kind of science in an 8-page article is like writing a short essay on “World War II”. I’m sure the best parts are in there, but you can’t begin to reenact it based on their descriptions. Lots of the interesting sub-analyses, which I presume must have been performed, will never see daylight. This is exactly the sort of science that benefits from the innovation of the online journal. No page limitations are needed there. Just last week, for example, I noticed that PlosOne had published a paper entitled “New Mid-Cretaceous (Latest Albian) Dinosaurs from Winton, Queensland, Australia”, which is 51 pages long and contains 40 illustrations, mainly of various bones photographed from different angles. Try getting that into a conventional journal!

Full reference:
Suzuki, H., Forrest, A., van Nimwegen, E., Daub, C., Balwierz, P., Irvine, K., Lassmann, T., Ravasi, T., Hasegawa, Y., de Hoon, M., Katayama, S., Schroder, K., Carninci, P., Tomaru, Y., Kanamori-Katayama, M., Kubosaki, A., Akalin, A., Ando, Y., Arner, E., Asada, M., Asahara, H., Bailey, T., Bajic, V., Bauer, D., Beckhouse, A., Bertin, N., Björkegren, J., Brombacher, F., Bulger, E., Chalk, A., Chiba, J., Cloonan, N., Dawe, A., Dostie, J., Engström, P., Essack, M., Faulkner, G., Fink, J., Fredman, D., Fujimori, K., Furuno, M., Gojobori, T., Gough, J., Grimmond, S., Gustafsson, M., Hashimoto, M., Hashimoto, T., Hatakeyama, M., Heinzel, S., Hide, W., Hofmann, O., Hörnquist, M., Huminiecki, L., Ikeo, K., Imamoto, N., Inoue, S., Inoue, Y., Ishihara, R., Iwayanagi, T., Jacobsen, A., Kaur, M., Kawaji, H., Kerr, M., Kimura, R., Kimura, S., Kimura, Y., Kitano, H., Koga, H., Kojima, T., Kondo, S., Konno, T., Krogh, A., Kruger, A., Kumar, A., Lenhard, B., Lennartsson, A., Lindow, M., Lizio, M., MacPherson, C., Maeda, N., Maher, C., Maqungo, M., Mar, J., Matigian, N., Matsuda, H., Mattick, J., Meier, S., Miyamoto, S., Miyamoto-Sato, E., Nakabayashi, K., Nakachi, Y., Nakano, M., Nygaard, S., Okayama, T., Okazaki, Y., Okuda-Yabukami, H., Orlando, V., Otomo, J., Pachkov, M., Petrovsky, N., Plessy, C., Quackenbush, J., Radovanovic, A., Rehli, M., Saito, R., Sandelin, A., Schmeier, S., Schönbach, C., Schwartz, A., Semple, C., Sera, M., Severin, J., Shirahige, K., Simons, C., St. Laurent, G., Suzuki, M., Suzuki, T., Sweet, M., Taft, R., Takeda, S., Takenaka, Y., Tan, K., Taylor, M., Teasdale, R., Tegnér, J., Teichmann, S., Valen, E., Wahlestedt, C., Waki, K., Waterhouse, A., Wells, C., Winther, O., Wu, L., Yamaguchi, K., Yanagawa, H., Yasuda, J., Zavolan, M., Hume, D., Arakawa, T., Fukuda, S., Imamura, K., Kai, C., Kaiho, A., Kawashima, T., Kawazu, C., Kitazume, Y., Kojima, M., Miura, H., Murakami, K., Murata, M., Ninomiya, N., Nishiyori, H., Noma, S., Ogawa, C., Sano, T., Simon, C., Tagami, M., Takahashi, Y., Kawai, J., & Hayashizaki, Y. (2009). The transcriptional network that controls growth arrest and differentiation in a human myeloid leukemia cell line Nature Genetics, 41 (5), 553-562 DOI: 10.1038/ng.375

Remarkable fitness gain from self-organising properties of cancer cells?

April 8, 2009

ResearchBlogging.orgNeuroscientists have their Aplysia, geneticists have their Drosophila. We, in the field of cancer research, have HeLa – the cervical cancer cells of Henrietta Lacks, probably the most widely used cell line in the world. HeLa, and thousands of cell lines like it, form the bulk of our experimental material. Cell lines are made up of cells that have become immortalised through tumour progression, and can be cultured and passaged indefinitely. Most of them grow nicely on glass or on the coated insides of plastic culture bottles, forming a flat monolayer.

Very much of our knowledge about cancer comes from these model systems. But how well do they resemble an actual tumour?

There is a growing recognition that a tumour is a complex organ made up not only by the tumour cells themselves, but also by many kinds of supporting cells – stromal fibroblasts, blood vessel cells, etc – that enable the cancer cells to grow and proliferate. The 3D interactions between tumour cells and the extracellular matrix (that is, all the stuff that makes up tissues but lies outside of cells) are immensely important in determining the growth patterns of the cancer cells.

The tumour organ, if you wish, is possible only because of the strong self-organising properties of the diverse tissue components. Supporting cells line up next to the cancer cells. Blood vessels grow into areas that are poorly oxygenated thanks to signals that are basically the same as in healthy tissues, for example when muscles grow and need a greater blood flow.

It could be said that the tumours have subverted these normal processes for their own malignant purposes. Or, alternatively, you might argue that it is the normal tissue that drives tumour progression, leaving the cancer cell relatively innocent. (These two viewpoints appear contradictory at first, but a moment’s reflection reveals that it only seems this way because the concepts of purpose and guilt have been introduced in spite of being meaningless in this context.)

A recent paper by J Daubriac et al, in Cell Death and Differentiation, investigates a particular kind of self-organisation: when cancer cells that float freely attach to each other and start to organise into a small tissue.

A very special habitat for cancer cells, which particularly concerns mesothelioma researchers like myself, is in the fluid of the thoracic and abdominal cavities. As you might expect, there is always a small amount of fluid surrounding the lungs and the intestines, to reduce friction when they move. The surfaces are made up of a thin layer of mesothelium, which has a lubricating side facing outwards toward the opposite surface.

Occasionally, mesothelial cells come of the surface and start to live as free-floating cells in the fluid. It is thought that they can spend some time there and then settle again if they find a free spot, thereby contributing to healing any wounds in the mesothelium.

Life as a “floater” is quite different from life on the surface, because on the surface it is necessary for the cells to have polarity. Polarity is the essence of epitheliality, and means that there is a distinct surface facing downwards and to the sides (basolateral), and an apical surface with completely different characteristics facing towards the lumen, or hollow space, of whatever the epithelium in question is lining, be it an airway, a milk duct, or an intestine. The basolateral side is characterised by specialised adhesion structures; desmosomes and adherens junctions to other cells, and integrins and cadherins to the underlying connective tissue, usually a specialised basal lamina. The apical side is where secretion and absorption take place, and where specialised structures like cilia and microvilli are found.

Polarity is lost when the cell floats alone.

The cell must have contact with other cells in order to establish the adhesion structures that define its polarity. Normal “floaters” never achieve this. The presence in the thoracic fluid of a group of cells that adhere to each other is an ominous sign of cancer. Most cells, in fact, are programmed to commit suicide if they lose attachment, in a particular type of apoptosis termed anoikis. Cell attachment is a bit like the safety handle on a garden shredder; the moment you stop pressing it, the machine shuts off.

Dr Daubriac has investigated this behaviour by cultivating mesothelioma cell lines in flasks where they could not attach to the bottom. Instead, they started attaching to each other. They quickly reestablished their polarity, forming spheroids in the µm size range. Some of the cells even constructed a basal lamina in the centre of the spheroid, thus recapitulating the normal tissue features rather fully.

Two particularly interesting findings appeared when the spheroids were compared to normal monolayer culture. Firstly, their growth rate was lower. Cells in spheroids appeared to become rather content with sitting there, with neighbours on all sides, and stopped proliferating. Secondly, they were also much less likely to undergo anoikis than monolayer cells. Daubriac and his colleagues investigated the intracellular signalling pathways in some detail and were able to say that those signals that normally lead to anoikis were shut down in spheroid-growing cells.

Cytological specimens of mesothelioma cells growing in a spheroid pattern, termed papillary structures. The image is from

Cytological specimens of mesothelioma cells growing in a spheroid pattern, termed papillary structures. The image is from

So what is it with these tumour cells that makes them think that they have to build a new tissue? Have they acquired genetic changes that turn this program on under inappropriate conditions?

The authors’ idea is that spheroid growth is motivated by the protection against anoikis that it brings. This implies that there is a selection effect. The cells that could not form spheroids may have died and disappeared much faster, leaving the spheroids to make up a growing proportion of the floating cells. In this case, the decreased death rate may offset the decreased proliferation rate of spheroid-growing cells, leading to a net fitness gain for these cells. Furthermore, the spheroids were able to start proliferating again when they were returned to a normal culture flask and could grow out as a monolayer. It brings an image to one’s mind where spheroid cell groups are the agents of metastasis. If that is so, then a treatment directed against the spheroids might be highly valuable!

Again, we can witness the enormous explanatory potential of the evolutionary paradigm. And it always helps to keep in mind that tumours are among the most dynamic evolutionary systems that exist, because of their widespread genetic instability.

If self-organisation in spheroid growth contributes to metastasis, then it’s a bit revolutionary, because self-organisation usually works the other way! The more the cancer cells try to build a functioning tissue, the less malignant will the tumour be. Cancer cells resembling the original tissue are called well-differentiated, and then there is a whole scale with anaplastic at the bottom, which means that the cancer cells look pretty much like stem cells with no particular tissue allegiance. The differentiation state is closely liked to the patient’s prognosis.

This paper leads to a clear hypothesis that mesothelioma cells have an evolutionary fitness gain from spheroid growth. In principle, that should not be too difficult to prove or refute. I look forward to more research on the topic!

Full reference:

Daubriac, J., Fleury-Feith, J., Kheuang, L., Galipon, J., Saint-Albin, A., Renier, A., Giovannini, M., Galateau-Sallé, F., & Jaurand, M. (2009). Malignant pleural mesothelioma cells resist anoikis as quiescent pluricellular aggregates Cell Death and Differentiation DOI: 10.1038/cdd.2009.32

Surely you’re joking, mr Ernberg!

March 20, 2009

Have we really solved the riddle of cancer? Yes, says Ingemar Ernberg, the venerable professor who has written the foreword to ”Prostatacancer”, hot off the presses of the Karolinska University Press.

I was somewhat surprised by his argument, which runs something like this: If there ever were a riddle of cancer, we have solved it by showing how the cell’s actions are controlled by gene regulatory networks. With ceaseless environmental perturbations of these networks, coupled with the powerful organizing principle of evolution, nothing mysterious remains.

Certainly, the advances in tumor biology have been tremendous over the past decades. And it is no coincidence that much of what we have learned about genetics and cell signaling has been discovered in the context of cancer. But can we really say that we understand these processes because we have identified the constituent parts and some of their connections?

If this were true, the riddle of consciousness was really solved in the 1800:s, when Golgi invented the silver staining that for the first time enabled us to see how neurons connect with axons and dendrites.

What professor Ernberg does not consider is the complexity that arises through the dynamic information transfer of the network. On this higher-order level, in cells just as in the nervous system, behavior emerges that cannot meaningfully be accounted for by cataloging the interactions of the component parts.

If this is not immediately obvious, consider the following. Certain genes, when upregulated, cause cells to proliferate a lot. An example is the c-myc gene. This gene can be accidentally moved to the place for the immunoglobulin gene in certain lymphocytes when they are infected with the Epstein-Barr virus. As a result, the lymphocytes proliferate enormously, and we have leukemia. Other genes, which sometimes cause cells to proliferate a lot, can also sometimes cause them to die a lot. An example is th JNK gene. There has been much controversy over whether JNK is pro- or antiproliferative. Now, it is generally accepted that it is both.

In total, we humans have around 20 000 genes. Even if each gene only interacts with 10% of the other genes, and the interaction is always linear, a model to explain the cell’s behaviour would be totally intractable even with enormous computing power. When many of the interactions are non-linear, it becomes clear that a successful description of this system, with the power to predict what it will do, must consider a higher level of organisation. Analogies abound; reading the Pickwick papers by Charles Dickens letter by letter vs. by the meaning of phrases and their conjunctions (D. Hofstadter, in Gödel, Escher, Bach), or understanding a city by copying the telephone directory vs. actually finding out where people are going every day and why (Sidney Brenner).

The riddle of cancer remains. The most important discovery we have made so far is that the riddle of cancer is identical the to riddle of Life itself; namely how the genes and proteins that are the basic units of biological information, as well as the basic operators on this information, together determine the fate of the cells which are the smallest units of life as we know it.

(I am indebted to professor Ernberg for having created much of the intellectual arena where I have encountered several of the more groundbreaking recent advances of thought in tumor biology, and I argue against him safe in the knowledge that he will only be pleased that his ideas are debated.)

Will your cell phone give you CANCER?

March 17, 2009

The world is full of hidden dangers, in some people’s view. Some things that were thought harmless at some point in history have turned out to be genuine killers – lead, tobacco smoke, and asbestos, to mention only a few examples.

Our historical success in identifying unobvious dangers has lead some people to develop mental models where there is always a widespread behaviour or exposure that is next in line to be uncovered as a cause for illness and mortality. Others believe (as I do) that with time, the remaining environmental hazards to be discovered will be less and less important, because the size of the adverse effect correlates strongly to how easy it is to detect, and therefore most of the really dangerous stuff is probably already known.

Whichever attitude people take, most agree that it is best to base specific recommendations on a dispassionate reading of the scientific data. However, opinions differ greatly about how strong the evidence has to be in order to restrict something that might be dangerous. “The principle of caution” states that it’s better to regulate when we are uncertain. But how uncertain?

A useful guide for determining the strength of evidence has been proposed by Austin Bradford Hill (1965). Its criteria have been summarised as follows: Causality becomes knowable when scientific experiments demonstrate, in a strong, consistent (repeatable), specific, dose-dependent, coherent, temporal and predictive manner that a change in a stimulus determines an asymmetric, directional change in the effect.

In this post, which I fear will be rather lengthy, I will try to review the evidence for and against dangers of mobile telephones. The positive health effects are significant (easier access to medical advice and emergency services, in particular), but we will only look at the possible risks here.


There is a debate over the health risks of electromagnetic radiation in general. When we use a mobile telephone, we expose ourselves to an electromagnetic field which penetrates about 4 cm into the head. Some national authorities have defined exposure limits. For example, the Swedish Radiation Safety Authority requires that all phones cause less energy absorption than 2 W/kg in human tissues, a limit that has been more or less arbitrarily chosen on the base of acute radiation effects. Absorption of radiation causes tissues to become warmer, and 2 W/kg is far below any measurable thermic effect. (This is the general principle of the microwave oven, and the only completely indisputable biological effect of electromagnetic fields.)

Other dangers have been seriously investigated, particularly the risk of cancer and of neurological disorders such as headaches, dizziness, and dementia.

Loud activist groups are lobbying for more restrictions. The most authoritative critics are probably the BioInitiative Group. They released a report in 2007 urging the prohibition of strong electromagnetic fields. I purports to “document serious scientific concerns about current limits regulating how much EMF is allowable”. But although scientists are among the authors, the document is not a piece of scientific criticism. It is a jumble of cherry-picked research reports and unfounded claims. The authors are quite consistent in only mentioning research that supports their own point of view. Furthermore, they accept controversial data regardless of prior probability.

Let’s say that I were to investigate whether my zodiac sign is a strong risk factor for Parkinson’s disease. If I were to choose a confidence level of 95%, I would end up being wrong 5% of the time, which is the usual cutoff in science. Now, let’s say I get a positive result. Does that mean that sagittarians are at increased risk for Parkinson’s? No, because I have to take into consideration the prior probability that my result is correct. And our current understanding of disease physiology and the differences between people born at different times in the year makes such a result utterly implausible, at best. This is, incidentally, why it is impossible to prove that homeopathy works without first changing our understanding of the laws of chemistry.

The BioInitiative report, fearless of ridicule, is thus forced to conclude that since adverse effects of very low intensity electromagnetic radiation have been seen in cells in certain experiments, it must be the information conveyed by the radiation rather than the heat in the tissue that causes damage. It seems to be envisioning “death rays” that can be specifically tailored to destroy tissues, and that coincidentally are similar to the electromagnetic fields that surround us every day.

No, the BioInitiative report is not serious. What, then, do the big public agencies have to say?

The WHO has concluded that there is no increased risk for cancer or any other diseases, but that cell phones may lead to traffic accidents and may interfere with pacemakers. The EU:s expert group has concluded that there is no increased risk for cancer except possibly for a benign tumour called acoustic neuroma, and only for these who have used the cell phone for more than ten years, with almost a doubling of the incidence rate.

Here is where risk communication comes into the picture. About 80-100 cases of acoustic neuroma are diagnosed every year in Sweden, with a population around 9 million people. The risk is consequently about 1/100 000 for a person in a year. An increase in the relative risk by 100% corresponds, then, to an increase in the absolute risk from 0.001% to 0.002%.

MRI image showing an acoustic neuroma

MRI image showing an acoustic neuroma

And this is the most tangible risk. With regards to other types of cancer and neurological diseases, risks completely fail to pass the Hill criteria. The effects, if any, are weak, inconsistent, unspecific, weakly dose-dependent if at all, incoherent and unpredictive. Sadly, space does not permit me to go through the entire literature in this post, but good summaries are available from the WHO and the EU.

In spite of the lack of evidence, the campaigners are making headway. In September 2008, MEPs voted 522 to 16 to urge ministers across Europe to bring in stricter radiation limits and said: “The limits on exposure to electromagnetic fields (EMFs) which have been set for the general public are obsolete. The European Parliament is greatly concerned at the Bio-Initiative international report which points in its conclusions to the health risks posed by emissions from devices such as mobile telephones, UMTS, WiFi, WiMax and Bluetooth, and also DECT landline telephones. The plenary therefore calls on the Council to […] set stricter exposure limits.”

The politicians have to listen to the people, after all, even when the people are wrong. The whole conundrum is hilariously exposed in this episode of “Yes Minister”. (Go to YouTube for the remaining parts of the episode.)

In my opinion, the moral of the story is not that our elected leaders must stand up for science (although that is also true), but that we must all aid and assist our decisionmakers to behave rationally. If they think people want homeopathy and prohibition of GMO-crops, then that is what we will end up with.

When we have superior knowledge, the democratic right to share it is also a duty.

Why do female doctors have more fun with mammograms?

March 2, 2009
Cancer deaths could be reduced by one third, or more, according to expert estimates. One of the strategies to reach this goal is the use of screening programmes. Women in most high-income countries will participate during their lifetime in a screening programme for cervical cancer, and another for breast cancer. Both types of screening have an essential component of image interpretation by a doctor – either a mammogram, i.e. a special x-ray picture of the breast, or a Pap smear, which is a sample of cells from the vagina that is examined under the microscope. These tests are like windows into the magnificent palace of the female body, through which doctors peer like nightwatchmen to try to make out whether any of the shadows in the darkness inside are dangerous lurkers.

Millions of such pictures are examined every year, by doctors suffering under the same unfavourable circumstances as pilots: everything is routine, but the consequences of even a slight error could lead to disaster and death. Decision-making is a challenge. Every doctor must calibrate him- or herself to a set-point where he detects as many as possible of the real positive cases without over diagnosing and creating false positives, which will cause patients to needlessly undergo maiming surgery. Doctors’ decisions, like anybody else’s, are affected by such mundane factors as their alertness, their mood, their serum caffeine levels, and the weather.

This subjectivity needs to be managed.

Of course, doctors have lots of strategies to do that. Most of the time, it works very well. Doctors meet, discuss cases in rounds, and recalibrate their set-points. The keep the images and monitor the patients for decades to see whether any tumours develop that they could have detected earlier. And they do research into the predictors of accurate diagnosis.

One such paper appeared last week in the American Journal of Roentgenology. It investigates the very plausible hypothesis that doctors who enjoy their work also diagnose more accurately. Berta Geller and her colleagues sent a questionnaire to 131 breast radiologists, and then linked their answers to known performance on 700 000 mammograms in a database. In the end, there was no significant connection between the reported enjoyment of interpreting mammograms and the performance of the doctors. This finding is very reassuring and in my opinion a bit surprising.

But these is more: The authors have investigated what factors are most likely to predict doctors’ enjoyment of interpreting mammograms, and these are the results:

Predictors of enjoyment

The y axis shows the odds ratio, which can be understood as “fold change in likelihood”. As the diagram shows, doctors were roughly eight times more likely to enjoy interpreting mammograms if they are women! This was a much stronger predictor than the feeling of competence, which comes next. The most important negative predictor was the fear of malpractice suits. (This is a US study.) The “Non-salaried” bar in the middle compares doctors who are paid per case to those paid a fixed salary. Not much difference there.

Why on earth do women find the work so much more enjoyable? There were 101 men and 29 women among the doctors, which should be enough to keep random variation out of question. (That adds up to 130 doctors – one is missing, who perhaps did not state his/her gender.)

Are there hidden confounders? Perhaps the women were younger on average in this sample, and more enthusiastic? No, because older doctors enjoyed their job more than younger doctors did. Could it be that women feel disproportionately skilled at interpreting mammograms? Or is it in fact a real difference, due to some other factor such as that all the doctors identify more with female patients, or that women enjoy image interpretation more in general?

All rather unlikely explanations I think. But science is at its best when it reveals unlikely chains of causation to be true! I hope someone will make the effort to find out whether this is a consistent finding and, in that case, what causes it.

Full reference: B. M. Geller, E. J. A. Bowles, H. Y. Sohng, R. J. Brenner, D. L. Miglioretti, P. A. Carney, J. G. Elmore (2009). Radiologists’ Performance and Their Enjoyment of Interpreting Screening Mammograms American Journal of Roentgenology, 192 (2), 361-369 DOI: 10.2214/AJR.08.1647