1. Home
  2. Issue 6-3
  3. Report

advertisement

Topcon

Interpreting Relative Risks and Possible Implications for the current 'Mantra': 10% Risk Reduction for Every Millimeter IOP Reduction

Ravi Thomas, Chandra Sekhar and Rajul Parikh

When considering the application of clinical research to our patients, it is our personal preference to use the common sense (aka clinical epidemiology) concepts of NNT (number needed to treat), Relative Risk (RR) and Relative Risk Reduction (RRR).

The NNT reflects the number of patients we need to treat in order to obtain one benefit or desirable result. It is calculated from the absolute risks in the treatment and control groups.1 All else being equal, the lower the NNT, the better. The NNT allows us to identify high-risk or high-response sub-groups of patients who have a small NNT and stand the most to gain from the intervention in question. It also allows us to compare interventions. The relative risk (RR) tells us the risk of an outcome in one group with the risk factor (eg increased IOP) compared to the risk of an outcome in a group without the risk factor (eg normal IOP).2 A derivation of the RR, the relative risk reduction (RRR) tells us the proportion of the baseline risk that can potentially be removed by treating the risk factor in question. While we consider all the above measures in the interpretation of data, NNT is our preferred method of translating clinical research to patient care. Relative Risk Reduction (RRR) and excess risk (that we will encounter) are especially useful when intervention effects are small.

In the Early Manifest Glaucoma Treatment Study (EMGT), over a 5- year period 62% of untreated subjects (absolute risk in controls) progressed, compared to 45% in the treated group (absolute risk in the treated group).3 The RR of progression in the EMGT was therefore (62/45) = 1.38. How do we interpret this RR? Or, in other words, what is a 'good' RR? Our epidemiologist friend tells us that RR's below 2 are rarely significant. According to the author of our clinical Bible,1 depending on the type of study RR's of 3-4 are likely significant, and 20 is probably high enough to attribute causality. As an example, an RR of 5.7, with confidence intervals of 1.9-17.1 ascribed to diurnal variation is impressive on its own.4 It doesn't require any manipulation to make it look more impressive. Especially since the lower end of the confidence interval (CI; 1.9-17.1) is around the ball park figure, we would be interested anyway.

What if the relative risk is lower than what would usually be considered significant? Like 1.38 for EMGT? A good way to present a small RR, especially to decision makers, is to use the Relative Risk Reduction:
 

Absolute Risk in the control group - Absolute risk in the treated group  
 × 100 		Absolute Risk in the control group  

For the EMGT: (62-45/62) × 100 = 27.42%. This is more easily calculated as RR-1/RR, and provides the same result.
This RRR implies that if there is a cause and effect relationship between IOP and progression, we can potentially remove 27.4% of that risk by lowering the IOP. Which is very important. However RRR does not provide much useful clinical information. A relative risk reduction (or relative risk) of 50% could mean an 'absolute' reduction from 100% to 50% or from 1% to 0.5%. These figures translate into an NNT of 2 for the first example, versus 200 for the second.1,5 The RRR in the two examples is the same, but the clinical connotation and potential clinical decisions are totally different. All else being equal, we would probably always use the therapy that provides an NNT of 2 and very rarely use one with an NNT of 200.

There is little doubt about the role of raised IOP in the causation and progression of glaucoma. However, results from some recent studies have been presented in an isolated manner that might encourage very aggressive lowering of the IOP with possibly detrimental effects.
Both OHTS and the EMGT state that each mmHg-lowering of IOP is associated with a 10% lowering of risk.6,7 First, we must remember that in both these studies this conclusion was the result of 'post hoc' analyses; such analyses are always interpreted with care. In the EMGT, this association was demonstrated using a Cox's hazard model and the hazard ratio was found to be 1.1. Hazard ratio is like RR and is interpreted in the same manner.

A relative risk of 1.1 is small (large numbers can make anything statistically significant). If the relative risk is small, (like 1.1), one way to make it look attractive as we learnt is the RRR. The RRR for a relative risk of 1.1 (RR-1/RR) is 9%. There is another way to make this look even more attractive or 'sellable', so to speak. Instead of the relative risk reduction, we can use 'excess' risk. The formula for excess risk is RR-1 expressed as a percentage = 1.1-1 = 10%. This derivation (with other assumptions) has probably lead to the '10% reduction per mmHg reduction' interpretation. While it is the truth, the statement is perhaps best interpreted keeping the overall picture in mind.

We personally feel that a combination of all the measures, absolute risk, relative risk, relative risk reduction and of course our favorite, the NNT, provides far more useful and clinically useable information than the RR (or RRR) alone.

References

  1. Sackett DL, Haynes RB, Guyatt GH, et al. 1991. Clinical Epidemiology. A Basic Science for Clinical Medicine. New York: Little, Brown & Co., pp. 205-209
  2. Reigelman RK. 2000. Studying a Study and Testing a Test. How to Read the Medical Evidence. Philadelphia: Lippincott, Williams and Wilkins, pp. 53-54
  3. Heijl A, Leske MC, Bengtsson B, et al. 2002. Early Manifest Glaucoma Trial Group. Reduction of intraocular pressure and glaucoma progression: results from the Early Manifest Glaucoma Trial. Arch Ophthalmol 120: 1268-1279
  4. Asrani S, Zeimer R, Wilensky, et al. 2000. Large diurnal fluctuations in intra ocular pressure are an independent risk factor in patients with glaucoma. Journal of Glaucoma 9: 134-142
  5. Thomas R, Padma P, Braganza A, Muliyil J. 1996. Assessment of clinical significance: the number needed to treat. Indian J Ophthalmol 44: 113-115
  6. Gordon MO, Beiser JA, Brandt JD, et al. 2002. The Ocular Hypertension Treatment Study: baseline factors that predict the onset of primary open-angle glaucoma. Arch Ophthalmol 120: 714-720
  7. Leske MC, Heijl A, Hussein M, et al. 2003. Early Manifest Glaucoma Trial Group. Factors for glaucoma progression and the effect of treatment: the early manifest glaucoma trial. Arch Ophthalmol 121: 48-56

Issue 6-3

Table of Contents Editor's Selection

Change Issue

advertisement

Oculus