Confounded or Confused?


Jacob Schor, ND, FABNO

A study published in May 2022 has left me pondering the difference between being confounded versus being confused.1   Both words are derived from the same Latin verb ‘confundo’, which means “to mix, blend, pour together, disorder, embroil, or confuse.”  Confound and confuse should be synonymous with each other. Yet they aren’t, well not exactly. As ‘confundo’ migrated from Latin through the Old French ‘confondre’ to the Anglo-Norman ‘cunfundre’ to the Middle English ‘confounden’, its meaning shifted in subtle ways.  Apparently by the 13th century, confound took on a slightly different and harsher meaning, as in “to condemn, curse, or to destroy utterly.”

The past participle of the Latin confundo, ‘confusus’, followed a similar journey over time into French and eventually to English so that by the 14th century it became ‘confuse’ as in ‘to perplex the mind’.  By the 1580s confuse meant to ‘treat or regard erroneously as identical.’2

[In British style English, confound is still sometimes considered a swear word as in ‘confound it.’] 

In the paper that instigated my confusion, Axel Möller and Swedish colleagues looked again at whether vaccination against cholera might benefit cancer patients, “Specifically, Ji et al showed that the administration of cholera vaccine after a prostate cancer diagnosis reduced prostate cancer specific mortality rates by almost 50%.”1

Tina Kaczor, ND reported on this potential benefit in the Townsend Letter in the recent past writing, “Since 2017, three studies have been published in peer reviewed journals suggesting that those who took an oral cholera vaccine (OCV) may derive benefit far beyond prevention of Vibrio Cholerae (V. Cholerae) infection. These studies have been retrospective observational studies that have compared the health outcomes of those who have received cholera vaccine with those who haven’t. The studies have focused on specific groups of people, those with colorectal cancer, men with prostate cancer and women with breast cancer. All these studies found that there were dramatically fewer cancer related deaths and overall deaths in those who took OCV versus those who did not.”3

Based on these positive associations, some of us have enthusiastically suggested that patients avail themselves to a specific oral cholera vaccine used in Sweden called Dukoral, which is available OTC in Canada.  Zheng et al’s 2021 report suggests that cholera-vaccinated breast cancer patients had a risk of dying of the disease that was nearly half of those who were not vaccinated [disease-specific survival (HR: 0.53, 95% CI: 0.33-0.84)].4  Ji et al. reported nearly identical benefits in prostate and colorectal cancer patients in 2018: [for prostate (HR, 0.53; 95% CI, 0.41-0.69) and CRC (HR, 0.53; 95% CI, 0.29-0.99)].”5,6

With reported potential benefits such as these from an inexpensive, simple-to-take OTC medication, it is no wonder that many were enthusiastic about cholera, so to speak.  Prior to these cholera vaccines, many naturopathic doctors who focused on oncology care had encouraged patients to get vaccinated routinely, not so much to prevent the vaccine-targeted infections but to stimulate immune function, often suggesting patients go to “travel clinics” to receive a broader range of less common inoculations against diseases that they were unlikely to encounter.  Some colleagues believed that specific vaccines were more effective based on various studies. I recall one colleague who prefers typhus. Our belief was that the process of vaccination might trigger an immune response that spills over against cancer tumors.  Such ‘non-specific effects’ of vaccination have been documented for years.  Opinions regarding which vaccine was superior had varied among practitioners, but Ji et al’s cholera papers created something of a consensus, albeit, a temporary consensus, that patients should make the effort to get the Canadian cholera vaccine.

With this backstory in mind, it caught my attention when Möller et al suggested that Ji’s reported benefits were all the results of confounding. 

In the realm of research, the word “confound” has a specific meaning. Especially in epidemiological investigations of cause-and-effect relationships, it is possible that an unmeasured and unaccounted for third variable may influence the relationship being examined and confound the results. Such confounding variables can make a mockery of one’s conclusions. To be a confounder a variable must be correlated with the independent variable, and it must be causally related to the dependent variable. And of course, generally, the researchers are unaware and thus unable to factor in the effect of the confounding variable. 

Here’s a simple example of a confounder. Imagine that data show eating ice cream is associated with sunburn. One might conclude that eating ice cream increases sensitivity to sun exposure and probability of sunburn.  The confounding variable in this scenario, of course, is ambient temperature.  Warmer temperatures increase ice cream consumption and also increase the likelihood that people will be outdoors and overexpose their skin to sun.7

Let’s look at this Möller study more carefully. 

“Because oral cholera vaccine is cheap, widely tested and without serious side effects, it is a perfect candidate for drug repurposing if it is proved to be effective for other diseases.”1

 Möller and colleagues basically repeated and extended the studies by Ji et al to “investigate the effectiveness of oral cholera vaccine after prostate cancer diagnosis on overall mortality. To explore whether any beneficial effect was seen only after cholera vaccination, we also studied the effectiveness of several other travel vaccines and malaria prophylaxis on overall mortality for men with prostate cancer.” Their hypothesis was, “that if oral cholera vaccine has a true effect on overall mortality in men with prostate cancer diagnosis, we should not observe similar effects with other travel vaccines.”

This research was “registry-based” in that all the data were extracted from official government registers.  The researchers used the Stockholm PSA and Biopsy Register (SPBR), a register that included all men living in Stockholm county, Sweden, during 2003 to 2018. The SPBR is linked to the Cause of Death Register, the Cancer Register, the National Prostate Cancer Register, the National Patient Register, the Prescribed Drug Register, and the Longitudinal Integration Database for Health Insurance and Labor Market Studies (LISA). The SPBR register includes more than 400,000 men with data on cause of death, date of death, cancer status (prostate specific antigen (PSA) at diagnosis, Gleason score, TNM stage, date of diagnosis), drugs dispensed by Swedish pharmacies, comorbidities, income, educational status, and primary treatment for prostate cancer.

In Sweden only about 15-20% of cholera vaccinations are obtained from pharmacies and tracked through the Swedish Prescribed Drug Register.  A good percentage are also obtained through a company called Svea Vaccin, a chain of vaccination clinics with 15 clinics in Stockholm county providing about 25% of the vaccines in Stockholm.

Möller’s group also tracked exposure to other vaccines, including for hepatitis (A, B, A+B, or others), Japanese encephalitis, yellow fever, rabies, typhoid fever, tuberculosis, or malaria prophylaxis.

Data were also obtained from national registers that “reflected socioeconomic status,” highest attained education, and country of birth.

Sophisticated Cox models were then used to evaluate the effect of exposure to cholera vaccine, other vaccines and the various other variables on risk of death, both disease specific or all-cause.

A total of 22,738 men with prostate cancer were included in the study. Of those, 4,248 (18.7%) died during follow-up. The median follow-up time in the study was 5 years and the median age at diagnosis of prostate cancer was 68 years.

After being diagnosed with prostate cancer, 717 (3.2%) of these men received oral cholera vaccine (Dukoral), and of these men, 36 died during follow-up. In a crude model (that is, without doing any statistical covariate adjustments), the men who received the vaccine had a 70% reduced risk of dying from all causes (HR 0.30, 95% CI 0.21–0.41, p < 0.0001). After adjustment of all available confounders, the estimated treatment effect of the cholera vaccine remained high, with a 54% reduced mortality (HR 0.46, 95% CI 0.31–0.69, p-value 0.0001).

Exposure to other vaccines was also associated with a reduced risk of dying. Five of the eight vaccines (Malaria prophylaxis, vaccines against any hepatitis, hepatitis A, Japanese encephalitis and typhoid fever) were all associated with statistically significant reductions.

These risk reductions were comparable in effectiveness to the cholera vaccine, ranging from an HR of 0.05 (95% CI 0.01–0.46, p-value < 0.0001) for Japanese encephalitis to 0.59 (95% CI 0.37–0.93, p-value 0.0237) for any hepatitis vaccine. The one exception, interestingly, was the combination vaccine against hepatitis A+B that had a small statistically insignificant effect (HR 0.94, 95% CI 0.45–1.95).

Here’s where it gets confusing.  The authors interpreted their findings to suggest confounding had messed them up, that something else, some other variable that they had not accounted for, was responsible for the benefits rather than the vaccines. 

They put forth several theories of possible confounders.  They suggest what they call the ‘healthy vaccinee effect’, that is that “people who choose to vaccinate are more invested in their health and are therefore healthier.”8

The second possibility is that people in Sweden don’t routinely get vaccinated against cholera.  This oral cholera vaccine is administered to people planning to travel abroad, not as protection against cholera but to reduce risk of diarrhea. Thus, the vaccinated form a select group of people who will “be heavily affected by selection bias.” We might argue that people who plan a foreign trip after being diagnosed with cancer will be healthier than those who feel obligated to stay home.  One might also argue that cancer patients who know their health is deteriorating might decide to make the most of their remaining time while they can. 

The authors want to argue that because most vaccines work equally well, it isn’t the vaccines that provide benefit, but rather some difference in the people who chose to travel outside of Sweden or perhaps some exposure brought about while traveling that provides the benefit. 

Some readers of this publication might have a revelation and jump to the conclusion that this is all about vitamin D.  One could suspect that people are more likely to leave Sweden during the winter and that most destinations would be warmer and sunnier, leading to more sun exposure and increased vitamin D production.  (We might also suspect greater ice cream consumption.) But that feels like a stretch.

All these arguments and theories fail to account for one crucial observation reported by the Möller study. While most vaccinations were associated with similar benefit, recall that there was an exception, the combination hepatitis vaccine made little if any difference in outcome. One would assume that the mystery confounder, if one was present, would have had an equal effect across the board.

Perhaps the simplest explanation is that all the vaccinations help, or at least most of them. Rather than assuming a confounding variable nullifies the consistent benefits seen in these studies, might the actual conclusion be that a range of different vaccinations that provide protection against a wide range of infectious diseases also provide some benefit for cancer patients?  This is after all, what we have predicted for years.  And to this reader, at this time, this appears more plausible than some unknown confounder associated with travel outside of Sweden.

Rather than confuse our patients with discussions of confounding variables, why not instead summarize the research to date simply?  Here’s what we might say:

“Various vaccines seem to be helpful.  The oral cholera vaccine is at this time the most studied and is currently associated with lower risk of death for patients diagnosed with prostate, colon, and breast cancers.  It is also OTC, inexpensive, and quite safe to use.  Other vaccines might eventually prove to be more effective.  For the time being, until there are randomized clinical trials, we can’t say for sure. Oddly, the hepatitis combination vaccine isn’t associated with benefit so skip it.”

Or perhaps we should just encourage patients that after a cancer diagnosis that they vacation in distant places and make sure their vaccinations are up to date.


References

  1. Möller A, Schwamborn K, Spillmann A, et al. Travel vaccines are strongly associated to reduced mortality in prostate cancer patients – a real effect or residual confounding? Vaccine. 2022 Jun 15;40(27):3797-3801. doi: 10.1016/j.vaccine.2022.05.028. Epub 2022 May 21. PMID: 35610103. https://www.sciencedirect.com/science/article/pii/S0264410X22006119?via%3Dihub
  2. Merriam-Webster. (n.d.). Confound. In Merriam-Webster.com dictionary. Retrieved September 30, 2022, from https://www.merriam-webster.com/dictionary/confound
  3. https://www.townsendletter.com/article/460-cholera-vaccine-and-cancer-death-rate-reductions/
  4. Zheng G, Sundquist J, Sundquist K, Ji J. Association of post-diagnostic use of cholera vaccine with survival outcome in breast cancer patients. Br J Cancer. 2021 Jan;124(2):506-512.
  5. Ji J, Sundquist J, Sundquist K. Association between post-diagnostic use of cholera vaccine and risk of death in prostate cancer patients. Nat Commun. 2018 Jun 18;9(1):2367.
  6. Ji J, Sundquist J, Sundquist K. Cholera Vaccine Use Is Associated With a Reduced Risk of Death in Patients With Colorectal Cancer: A Population-Based Study. Gastroenterology. 2018 Jan;154(1):86-92.e1.
  7. [from https://www.scribbr.com/methodology/confounding-variables/ ]
  8. Remschmidt C, Wichmann O, Harder T. Frequency and impact of confounding by indication and healthy vaccinee bias in observational studies assessing influenza vaccine effectiveness: a systematic review. BMC Infect Dis. 2015 Oct 17;15:429.

Published June 3, 2023


About the Author

Jacob Schor, ND, now retired, had a general practice with a focus on naturopathic oncology in Denver, Colorado. He served as Abstract & Commentary Editor for the Natural Medicine Journal for several years (https://www.naturalmedicinejournal.com/) and posts blog articles on natural therapies,  nutrition, and cancer (https://drjacobschor.wordpress.com/). He is a board member of CoAND and past president of OncANP, and someone who is happier outdoors than inside.