A friend of mine pointed me to this rather… interesting (for want of a better word) study the other day. (What can I say? I have interesting friends!) Published in the journal Liver Transplantation (an organ of the American Association for the Study of Liver Diseases), the paper is entitled: Religiosity Associated with Prolonged Survival in Liver Transplant Recipients1 by Bonaguidi et al. of the Institute of Clinical Physiology of the National Research Council of Italy and the University of Pisa.
The study hypothesized that:
Religiosity (i.e., seeking God’s help, having faith in God, trusting in God, and trying to perceive God’s will in the disease) is associated with improved survival in patients with end-stage liver disease who have undergone orthotopic liver transplantation.
This observational cohort study looked at 179 candidates for liver transplantation, who completed a questionnaire on religiosity during the pre-transplant psychological evaluation, and underwent transplantation during 2004-7. The main variable studied was the survival of patients during follow-up, and the influences of demographics, educational level, employment status, clinical data and the results of the questionnaire, on post-transplant survival was analyzed using univariate and multivariate analysis. Eighteen patients (≈10% of the sample population) died during the follow up.
The questionnaire used asked all affirmative questions, meaning that the wording of the questions (see figure below) was designed not to question the participants’ belief in God, but to affirm what they already believed about their personal God. The responses were coded as: 1=not at all, 2=a little, 3=somewhat, 4=considerably, and 5=very much.
Graph generated from table 1 and 2 of the paper
Factorial analysis of the responses to the questionnaire indicated 3 factors or groups in patients: searching for God (active), waiting for God (passive), and fatalism; the study found that the “Seeking God” factor, as well as post-transplant ICU stay, were independently and significantly associated with survival, ever after adjustment of all other covariates, such as the waiting for God factor, fatalism, age, sex, marital status, employment, educational level, viral etiology, Child-Pugh score (score used to assess prognosis of chronic liver disease), serum creatinine level, time from the questionnaire to transplantation, donor age, and intra-operative bleeding.
The study also found that comparatively younger patients did not present the “Seeking God” factor, but had shorter survival times and a 3-fold increased relative risk of dying. It concluded: religiosity is associated with prolonged survival in patients undergoing liver transplantation, thereby affirming its initial hypothesis.
Hallelujah. I hope you all are with me till now, and particularly, the atheist godless bunch amongst my friends is now hanging your head in shame and cowering in fear for your lives. OMG! Should you ever require a liver transplant to recover from the excesses of your heathen lives, you are all going to die, die, I say!
[Silence]
Wait, what? Hold on a god-damned minute! What about the several problems that leapt at my eye at the first reading? Let’s examine them carefully.
The paper starts with grandiose claims in the introduction about how patients diagnosed with serious, end-stage illnesses turn to God, and how!
[…] this religiosity enables patients to better cope with the disease and improves their quality of life […] increased spirituality/religiosity has positive effects on the progression of the disease […]
… citing no less than 10 references, all of which make various kinds of noises about how people become religious after being diagnosed with incurable diseases.
In reality, however, how big of a deal is that? People have different coping mechanisms. Some find solace in delusions, some in shifting responsibility to some unseen, imaginary entity, and some in cold, hard rationality, intellect and grace. People turn to smoking marijuana or partaking of the hallucinogen, psilocybin, after an end-stage cancer diagnosis. Besides, people become religious for various reasons. Many turn religious while in prison, particularly when the newly-found religiosity earns them better stations and rations within the prison system. “Religiosity improving quality of life” is a hackneyed, culturally-conditioned canard, perpetuated by organized religion in order to retain the relevancy of old myths and superstitions in an increasingly skeptical world. Equally meaningless is the ennui-inducing imputation – also made in the introduction – that atheists are more likely to have unhealthy habits, such as smoking, alcohol consumption and sedentary lifestyles, as well as social isolation, all of which affect their lifespans negatively, in comparison to those that attend religious services or engage in religious practices. Notwithstanding the fact that 17% of the 179 patients in the study cohort presented with alcoholic liver disease requiring transplant. Snorgle!
To bring a spurious semblance of balance to their introduction, the authors pay lip service (4-5 scattered references) to studies that have shown that the association between religion/spirituality and health is weak and inconsistent, and that most of the perceived effects of religious practices in reality amounts to social interaction and support for the patients, which understandably prolongs their survival.
But while this expression of religious bias on part of the authors is quite blatant – I have to wonder how this paper ever passed peer review – it is certainly far from being the only problem with the study.
- For a study with 179 participants, look at the answers to the questionnaires and the SD error bars on the scores (the authors do comment upon the “considerable variability in the religiosity of the studied population”).
- In fact, I am surprised to find that the mean scores vary from 1.7 to 3.1, and don’t go closer to 5, which represents absolute belief in god.
- Also, it is not clear why those answers that displayed the Awaiting God factor maximally by factorial analysis scored the lowest in the religiosity scale (perhaps those patients were tired of waiting?).
- Also unclear is why the authors defined a belief in “destiny” (as opposed to in a specific deity) as “fatalism”, especially when the parameters of destiny are as much reasonable (ahem!) as those of “god”.
The authors explain factorial analysis thusly:
The answers to the questionnaire were analyzed by factor analysis, which is a statistical method developed in psychometrics and frequently used in different sciences. This analysis assumes that several observed variables together represent another unobserved variable.
Specifically, factorial analysis evaluates the variability among observed variables (in our case, the scores of the answers to the questionnaire) and searches for possible combinations. The information obtained about the interdependencies between the observed variables is used to reduce their number to a lower number of underlying, unobserved variables called factors.
In other words, this technique prepares artificial constructs, called factors, secondary to observed variables; it reduces the variability amongst observed overlapping variables by creating a new set of discrete variables, each of which contains a majority part of the observed variables. But for this study, it also means that, for example:
- Not everyone who considers the disease a lesson from god is actually seeking god.
- Some of those who waited for god to solve their problems were not just awaiting, but actively seeking, god, or believed in destiny also.
This inevitable overlap of belief, and the consequent fuzziness of the boundaries of the respective factors, immediately pulls a shroud of doubt on the interpretation of data, particularly when the authors claim that the “Seeking God factor was associated with survival.”
The conclusions are further clouded when “Seeking God” as a predictor of survival in univariate analysis and the so-called “increased risk of dying” in younger patients not seeking god end up with huge confidence intervals (CI) of respectively 1.05-8.32 and 1.07-8.45. CI is a range of values that quantifies the uncertainty derived from drawing the sample from a small representative population. A narrow confidence interval implies high precision, with plausible values lying within a tiny range. A wide interval implies poor precision — and implies that authors are trying to hedge their bets.
The authors claim that younger patients (50±9 yrs., as opposed to the “not-so-young” at 53±8 yrs.) belonged less to the “Seeking God” factor, when the age difference between their contrived grouping is merely 3 years, with >10% error bars. This seems to me to be an example of post-hoc rationalization of observed data; they appear not to have defined the age bar (“young” vs. “not young”) at the study initiation.
A major issue with the study – that occurred to me – veers around the single endpoint of patient survival. For a study that purports to show that religiosity prolongs survival, only 10% of the patients died in the 4 year follow-up period. So unless it is shown that those 18 individuals were by and large godless – okay, I am kidding! – say, less religious compared to their co-patients in the cohort, the conclusions of the study shall remain suspect, notwithstanding the statistical legerdemain.
The authors engage in prodigious hand-waving in the discussion, bemoaning the lack of knowledge of underlying physiological mechanisms linking religiosity with survival, and complaining at one point that “resources offered by spiritual factors constitute a topic that is poorly investigated and used in modern medicine, as if a sort of collective amnesia existed…” However, instead of offering conclusive proof of their hypothesis through their study, they state their “belief” that:
[…] the relationship between religiosity and prognosis is not one of mere association; instead, faith as a way of coping is a real resource for seriously ill patients and helps to improve their prognosis.
Sadly, wishful thinking is not science, and if this observational study honestly sought to demonstrate specific health benefits of willful suspension of disbelief and engagement in delusional thinking, this wasn’t it. In other words, U R doin’ it rong!
Can we chant now, all together, “Correlation does not presuppose causation!” — for, if one is to glorify correlation in so blatant a manner, how about I propose a parallel study, titled: “religiosity is associated with prolonged physical and emotional abuse of victims by their religious leaders” for which there is, at least, ample prima facie evidence in the contemporary world?
My friend who alerted me to the paper is a Biostatistician by profession; he, too, jotted down his reactions to the paper in his blog; do read. In addition, in the same journal, Frank de Vocht of University of Manchester wrote a rebuttal letter2 to the editor, raising several important design and analysis flaws in this paper, including the authors’ inability to recognize and address several confounding factors: (I arranged them into points for easier comprehension.)
- Limiting the study population to patients that actually responded to a questionnaire on religiosity has most likely introduced a selection bias in this study […] this should at the very least have been acknowledged by the authors as a limitation of the study while preferably its impact should have been assessed by comparing participants to nonresponders.
- […] positive thinking and a positive attitude, with and without the influence of religion, has been suggested to improve survival rates of patients. […] (authors) failed to recognize that this “positive attitude” can be related to a wide variety of factors of which religiosity might be one.
- […] in specific populations religiosity could be correlated to attitude and serve as a proxy variable without implying causality […] authors refer to a study showing that the effects of spirituality can already be mediated by better social support alone […] authors did not acknowledge this and did not collect information on general attitude and social environment to be able to assess whether any additional influence of religiosity did exist.
- […] (data) suggest a different baseline risk between patients “Seeking God” and those “Not Seeking God” expressed by a 3-fold higher risk of death from all causes in the latter group. This indicates important differences in other, most likely lifestyle, factors related to pretransplant health between both groups. When not adequately addressed, these baseline differences prohibit any conclusions on attitude, with or without the influence of a God, and survival in these patients.
Why is it important to pick up such study papers, however scientifically flawed, and critically analyze them? The answer, which has to do with “good science-communication practice”, is given astutely by de Vocht: [I quote, because I can’t improve upon it] Given the methodological flaws described here but also mentioned, but not recognized as important, by the authors, they should have adopted an epistemological more modest and careful attitude with regard to their conclusion. Even more so because studies addressing these types of hypotheses tend to get picked up in mainstream media without the nuances of a peer-reviewed scientific article. [End quote]
This post was originally published on October 23, 2010. However, during transfer from the old to new system, the formatting went awry and the post got buried. I have updated it and added references. ~ February 18, 2014.
Papers discussed:
- Bonaguidi F, Michelassi C, Filipponi F, & Rovai D (2010). Religiosity associated with prolonged survival in liver transplant recipients. Liver Transplantation, 16 (10), 1158-63 PMID: 20818656
- de Vocht F (2011). The influence of seeking God in the association between religiosity and prolonged survival in liver transplant recipients. Liver Transplantation, 17 (2) PMID: 21280195
And this is why I keep my religion and my science separate.
Although, I’ve seen much worse papers having nothing to do with religion (at least they considered some confounders). In fact, measuring anything in the soft sciences is difficult, so I don’t fault them for trying.
That is an admirable sentiment, Catherine, and thank you for commenting. Everyone’s religion (or lack thereof) is essentially a matter of personal realization, choice or preference. But trying to fit (‘shoehorn’ is the word I’d use) scientific research and clinical studies to conform to one’s worldview borne out of religious faith is one of the worst forms of confirmation bias, and is incompatible with science. This is why I do fault the authors for trying; as scientists, they should have known better.
Also, admittedly there are “much worse papers having nothing to do with religion“, but don’t you think this is a rather tu quoque argument given the context?
Organ? Surely, this was intended as a pun.
Well, that’s what they said… 😀