SEARCH

SEARCH BY CITATION

To the Editors:

I read with great interest the article by Pons-Estel et al, recently published in Arthritis Care & Research, on hydroxychloroquine (HCQ) use and its effect on renal damage in persons with underlying lupus renal disease (1). The authors found that HCQ use (classified as ever used versus never used) by persons with renal disease, but not yet any renal damage as defined by the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index renal domain, reduced the risk of subsequent renal damage with an adjusted hazard ratio (HR) of 0.29. As presented, this article has 2 major limitations in study design that make this finding questionable.

The first limitation is confounding by indication. In this set of patients, who already had evidence of kidney disease, it is likely that those who were felt to be at lowest risk of permanent kidney damage were those treated with HCQ, while those at higher risk were likely placed on more aggressive therapies. Therefore, part of the protective effect seen may actually be the result of the underlying lack of severe disease in those patients taking HCQ. Disease severity would, in this case, be a classic uncontrolled confounder (2). Although Pons-Estel et al adjusted for disease severity and other variables that they felt could account for HCQ prescribing at baseline, this would not necessarily account for changes in disease activity over time, nor would it account for unmeasured confounders. Even the use of propensity scores will not adjust for confounding by indication if all relevant variables are not included in the propensity score model (3).

The second potential flaw is that of immortal person-time bias. Immortal person-time occurs when a period of time is included in the potential at-risk followup time, even though that period precedes the last event required to satisfy an exposure definition. In this study, it appears that subjects were classified as being HCQ users or nonusers as if they had been exposed to HCQ at the beginning of followup time, though in fact exposure must have taken place at various times over the course of followup. The time between the beginning of followup time and the first HCQ prescription is immortal, i.e., a subject must have survived without renal damage until this time in order to have received HCQ. It is likely that if the subject did not survive without renal damage, a more aggressive agent than HCQ would be administered. Therefore, subjects classified as taking HCQ were guaranteed to have a survival advantage over those who did not. Use of a time-dependent variable for HCQ exposure in the Cox proportional hazards models would have prevented this bias by classifying persons as unexposed before they had actually received HCQ and as exposed afterward (4).

Finally, the HR given indicates a very strong effect. It suggests that persons with existing lupus renal disease who take HCQ have a 70% reduction in the risk of any renal damage compared with those who do not take HCQ, after adjusting for other risk factors, including severity and other drugs received. This would suggest that a patient with World Health Organization class IV lupus nephritis and treated with HCQ alone would have a 70% reduction in the risk of damage compared with someone receiving no drug treatment. Such a magnitude of effect does not seem to be consistent with clinical experience and suggests that a significant bias could be present.

  • 1
    Pons-Estel GJ, Alarcon GS, McGwin G Jr, Danila MI, Zhang J, Bastian HM, et al. Protective effect of hydroxychloroquine on renal damage in patients with lupus nephritis: LXV, data from a multiethnic US cohort. Arthritis Rheum 2009; 61: 8309.
  • 2
    Rothman KJ, Greenland S, Lash T. Modern epidemiology. Philadelphia: Lipincott Williams & Wilkins; 2008. p. 650.
  • 3
    Sturmer T. Adjusting effect estimates for unmeasured confounding with validation data using propensity score Calibration. Am J Epidemiol 2005; 162: 27989.
  • 4
    Suissa S. Immortal time bias in pharmaco-epidemiology. Am J Epidemiol 2008; 167: 4929.

Steven C. Vlad MD*, * Boston University School of Medicine, Boston, Massachusetts.